Citation |

- Permanent Link:
- http://ufdc.ufl.edu/AA00022260/00001
## Material Information- Title:
- Short-term econometric forecasting analysis for Latin America
- Creator:
- Fullerton, Thomas M
- Publication Date:
- 1996
- Language:
- English
- Physical Description:
- vi, 91 leaves : ; 29 cm.
## Subjects- Subjects / Keywords:
- Debt ( jstor )
Economic inflation ( jstor ) Economic models ( jstor ) Exchange rates ( jstor ) Forecasting models ( jstor ) Market prices ( jstor ) Modeling ( jstor ) Prices ( jstor ) Secondary markets ( jstor ) Time series forecasting ( jstor ) Dissertations, Academic -- Economics -- UF Economic Forecasting -- Latin America ( lcsh ) Economic conditions -- Latin America -- 1982- ( lcsh ) Economic policy -- Latin America ( lcsh ) Economics thesis, Ph. D - Genre:
- bibliography ( marcgt )
non-fiction ( marcgt )
## Notes- Thesis:
- Thesis (Ph. D.)--University of Florida, 1996.
- Bibliography:
- Includes bibliographical references (leaves 84-90).
- General Note:
- Typescript.
- General Note:
- Vita.
- Statement of Responsibility:
- by Thomas M. Fullerton, Jr.
## Record Information- Source Institution:
- University of Florida
- Holding Location:
- University of Florida
- Rights Management:
- Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
- Resource Identifier:
- 023809897 ( ALEPH )
35730871 ( OCLC )
## UFDC Membership |

Downloads |

## This item has the following downloads: |

Full Text |

SHORT-TERM ECONOMETRIC FORECASTING ANALYSIS FOR LATIN AMERICA BY THOMAS M. FULLERTON, JR. A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1996 UNIVERSITY OF FLORIDA LIBRARIES ACKNOWLEDGEMENTS Professor Carol Taylor West is probably the best known and most widely published regional economic forecaster in the world. I was acquainted with her research and professional reputation long before applying for the senior economist position on her staff at the Bureau of Economic and Business Research. Over the past five years, we have faced computer crashes, parameter instability, hurricane shocks, business cycle inflection points, submission deadlines, bureaucratic regulations, jet lag, and mind boggling statistical revisions. Forecasting Program work was always accomplished in a responsible manner befitting the tradition of econometric excellence with which the University of Florida is so well known. It has been a tremendous experience to forecast the Florida economy and conduct research with Carol. David Denslow has been a friend, colleague, and coauthor. We have spent many hours discussing different policy issues and business outlooks that one or the other was called upon to discuss in a public forum. Similarly, Terry McCoy and I have participated in a number of conferences on topics relating to policy problems facing Latin America. Terry definitely has keen insights with respect to the entire region. Chunrong Ai helped clarify many of the mysteries of modern econometric theory. He also saved me a lot of time and effort by indicating which estimation strategies were viable with respect to Chapter 4 and subsequent research developed with that data set. Bill Bomberger spent generous amounts of time going over Chapter 3 and making helpful suggestions. Prior to his retirement, Henri Theil was on my dissertation committee. Subsequent to his retirement, Professor Theil has dropped by my office at least once a week to monitor my progress and talk about the post-war history of econometrics. Many other individuals in the College of Business Administration helped me navigate exams, computer systems, and Graduate School requirements. In no particular order, they include Stan Smith, Larry Kenny, Janet Galvez, Tony Tracy, Min Zhu, Yikang Li, Jennifer Cobb, Jon Hamilton, Richard Romano, Janet Fletcher, Pam Middleton, Tony Daniele, Janet Rose, Cindy Houser, Dot Evans, Dian Studstill, David Lenze, and Richard Lutz. My parents, relatives, and friends provided welcome relief on numerous occasions. It takes a lot of friendly advice to get through a doctoral program. iii TABLE OF CONTENTS page ACKNOWLEDGEMENTS ......................................... ABSTRACT ................................................. CHAPTERS 1 ECONOMETRIC FORECASTING IN LATIN AMERICA ....... 2 INFLATIONARY TRENDS IN COLOMBIA ................ 2.1 Introduction ............................... 2.2 Previous Research .......................... 2.3 Methodology ................................ 2.4 Estimation Results ......................... 2.5 Policy Simulation Results.................. 2.6 Conclusion ................................. 3 SHORT-TERM PRICE MOVEMENTS IN ECUADOR .......... 3.1 Introduction ............................... 3.2 Literature Review.......................... 3.3 Theoretical Model .......................... 3.4 Estimation Results ......................... 3.5 Policy Simulation Results.................. 3.6 Conclusion ................................. 4 PREDICTABILITY OF SECONDARY MARKET DEBT PRICES. 4.1 Introduction ............................... 4.2 Earlier Studies ............................ 4.3 Empirical Analysis ......................... 4.4 Conclusion ................................. 5 SUGGESTIONS FOR FUTURE RESEARCH ................ REFERENCES ................................................ BIOGRAPHICALSKETCH ....................................... ii Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SHORT-TERM ECONOMETRIC FORECASTING ANALYSIS FOR LATIN AMERICA By Thomas M. Fullerton, Jr. August 1996 Chair: Professor Carol Taylor West Major Department: Economics Given the short-term nature, one year or less, of many policy goals espoused by decision makers in Latin America, the objective of this study is to examine the applicability of different forecasting techniques to a subset of short-run economic questions. Because quarterly national income and product series are typically not available, most commercial econometric forecasting analysis for the region is conducted utilizing annual data. Monthly time series do, however, exist for a large number of key macroeconomic and financial variables. It is from the latter group of publicly available data sets that the econometric modeling and simulation exercises are drawn. Chapter 2 examines potential impacts associated with Colombian monetary authority efforts to cut the rate of inflation by 10 over a 12-month period. Principal tools relied upon include slower rates of nominal exchange rate devaluation and money supply growth. Empirical analysis in Chapter 2 is carried out utilizing transfer function autoregressive moving average modeling. Model simulations indicate that adherence to such a program can lead to noticeable disinflation over a 24-month period. Similar to other Latin American economies, Ecuador has faced persistently high rates of inflation in recent years. In May 1994, the government signed a stand-by loan agreement with the International Monetary Fund that established a goal of reducing the inflation rate to 15 percent over a 19-month period. Chapter 3 develops an exchange rate augmented monetary model to assess viability of the price stabilization program. In contrast to the time series approach of Chapter 2, short-run inflationary dynamics are modeled using an econometric framework. Economic debates regarding Latin America in recent years have been dominated by the debt crisis. In response to debtor country defaults, many lenders reduced or reshuffled risk exposures by selling debt certificates at discounts from face value. Chapter 4 analyzes the predictability of secondary- market debt prices for Colombia, Ecuador, and Venezuela. Estimation is accomplished via generalized least squares over 24 separate historical periods utilizing monthly data. Model simulations indicate that forecasting these prices is a difficult task. CHAPTER 1 ECONOMETRIC FORECASTING IN LATIN AMERICA Econometric forecasting analysis began developing as a field of research with the initial endeavors of Jan Tinbergen and others in the 1930s (Dhane and Barten, 1989). These relatively small macroeconometric models of the Dutch economy were developed using annual data. Research on business cycles in the United States began replicating and extending the Dutch modeling examples in the 1940s. Increasing interest in short- term economic fluctuations eventually led to the development of quarterly forecasting models (Barger and Klein, 1954). Along with the expansion of structural econometric modeling and research with respect to forecasting and policy analysis, time series statistics also became increasingly sophisticated in the realm of predictive modeling. Most of these efforts occurred with respect to high frequency monthly data, especially univariate autoregressive moving average models (Box and Jenkins, 1976). While often regarded as competitors, time series models are frequently utilized as complements to econometric equation sets and can also be imbedded within a variety of forecasting systems (Clemen, 1989; Fullerton, 1989a; Zellner, 1994; West, 1996). 2 Latin American macroeconomic forecasting models began to appear in the 1960s (Beltrdn del Rio, 1991). Similar to the first models for the Netherlands and the United States, the early Latin American models utilized systems of simultaneous equations designed around national income and product account (NIPA) data. Unlike most industrialized economies, however, NIPA data in Latin America during this period tended to be published only at an annual frequency. This constraint precluded the development of Latin American quarterly forecasting models in a manner analogous to what occurred in many industrialized economies. Although quarterly forecasting models have not been widely disseminated in Latin America, large-scale forecasting models built using annual NIPA data abound. Representative examples include the CIEMEX-WEFA model for Mexico (Beltran del Rio, 1991), the CIEPLAN model for Chile (Vial, 1988), the Metroecon6mica model for Venezuela (Palma and Fontiveros, 1988), and the WEFA models for Colombia and Ecuador (Fullerton, 1993a, 1993b). Among the most distinguishing characteristics shared by these models are continuous maintenance and enhancement over sustained periods of time. All of the aforementioned studies provide detailed references to the history of macroeconometric modeling in the region. Econometric forecasting analysis using annual data in Latin America has a fairly distinguished history and track record that is well-documented. 3 Given the volatile behaviors of the majority of the economies in Latin America in the 1980s, the absence of quarterly forecasting tools hampered business planning efforts. Forecasting conferences sponsored by commercial entities such as Wharton Econometrics during the late 1980s utilized simulation output from annual structural models for Latin American countries of interest. Client questions at these meetings were generally directed toward the first year of the forecast period, largely ignoring outer period extrapolation results (for example, see Fullerton, 1991). This is not to imply that the traditional macroeconometric Latin American models are regarded as useless. Most analyses of international indebtedness typically rely on annual data in order to examine the consequences of changes in regional balance of payment factors. Not surprisingly, significant effort was put forth in recent years to enhance the current account-capital account linkages and simulation performances in Latin American macroeconomic models (Fullerton, 1993a). The relative lack of forecasting models estimated with higher frequency data, nevertheless, continues to pose an obstacle to corporate, public sector, and multilateral agency planners and economists. Although quarterly national income and product account data are still not widely available to researchers in Latin America, monthly time series for many economic variables do 4 exist. Examples of the latter include inflation indices, money supply measures, currency exchange rates, commodity export prices, and international reserves. A logical step, therefore, is to investigate the econometric properties and predictability of the series presently available at the higher frequency. The centerpiece chapters of this dissertation examine three empirical forecasting questions using monthly data sets from Colombia, Ecuador, and Venezuela. Distinct estimation techniques are employed in each chapter and simulation accuracy is summarized for all of the models presented therein. Chapter 2 utilizes a transfer function autoregressive integrated moving average (transfer ARIMA) modeling framework to analyze movements in consumer prices in Colombia. Forecasting experiments are conducted with the resulting model to shed light on potential impacts associated with an anti- inflationary program enacted by the central bank. Results indicate that this class of time series modeling can provide useful insights with respect to macroeconomic trends in Latin American countries. A revised version of this chapter was published in the Journal of Policy Modeling (Fullerton, 1993c). Chapter 3 also examines the question of forecasting short-term price movements. Consumer prices in Ecuador are modeled using an econometric framework that incorporates domestic macroeconomic factors and international input costs. 5 Parameter estimation is accomplished with a nonlinear procedure that provides sufficient flexibility to handle even mixed error structures. Simulation exercises are also utilized to examine possible consequences associated with a price stabilization program implemented by the central bank. An abridged version of this study was published in Proceedings of the American Statistical Association, Business and Economic Statistics Section (Fullerton, 1995a). Secondary market trades involving sovereign debt certificates became widespread following the outbreak of the international debt crisis in 1982. Chapter 4 employs a generalized least squares modeling strategy to study the predictability of short-run movements in secondary market debt prices in Colombia, Ecuador, and Venezuela. As discussed below, forecast period lengths and information inputs are selected to reflect considerations facing financial market participants. A revised version of this chapter appeared in Applied Economics (Fullerton, 1993d). Additional research extending the initial results on this topic was presented at the 65th annual Southern Economic Association conference (Fullerton, 1995b). CHAPTER 2 INFLATIONARY TRENDS IN COLOMBIA 2.1 Introduction Inflation has long been one of the most serious economic problems facing policymakers in Latin America. Although Colombia has traditionally enjoyed lower rates of inflation than neighboring South American countries, in 1990, a presidential election year, consumer prices rose in excess of 32 percent. In response, Colombian monetary authorities introduced a series of policy innovations designed to lower the inflation rate. Measures adopted include a progressive opening of the economy to greater volumes of international trade, fiscal austerity, tighter credit controls, and a slower rate of currency devaluation (for details, see Fullerton, 1991). This chapter examines some of the potential results associated with the principal tools of the policy adjustment effort. Time series characteristics of the consumer price index (CPI), the money supply (Ml, defined as currency in circulation plus demand deposits), and the peso/dollar exchange rate (REX) are investigated. To measure the short- run relationships among the CPI, Ml, and REX, transfer 7 function autoregressive-moving average (ARIMA) analysis is applied (see Box and Jenkins, 1976, ch. 11). Selection of the nonlinear model estimation methodology was motivated by two factors. First was the usefulness of transfer function ARIMA analysis for short-term forecasting applications. Second was interest expressed in a previous study which utilized this technique to analyze price dynamics in the United States (Fullerton, Hirth, and Smith, 1991). Specifically, economists within the Research Department of the Central Bank in Bogota wished to see whether the same approach would prove useful with respect to the Colombian economy. Subsequent sections of the paper present a brief review of the literature, methodology, and empirical results. Policy simulation exercises designed to reflect exchange rate and monetary policies in Colombia follow. A summary and conclusion form the final section of the chapter. 2.2 Previous Research In one of the earliest time series studies developed for an economy in Latin America, Cabrera and Montes (1978) utilize univariate ARIMA techniques to model the CPI in Colombia. Logarithmic transformation, regular and seasonal differencing of the monthly CPI series are used to induce stationarity. An equation containing an autoregressive term at lag 1 and a seasonal moving average term at lag 12 yields statistically significant parameter estimates. While simple in structure, 8 the model exhibits good statistical traits and is found to simulate historical movements of the CPI successfully. Empirical evidence is provided that Colombian inflation, although high relative to many industrial economies, is stable enough to be modeled and predicted with a fair degree of accuracy. Montes and Candelo (1982) propose a simultaneous system of equations for money, prices, international reserves, and the exchange rate. Full information maximum likelihood estimation is used to calculate model parameters which reflect the hypothesized coefficient restrictions. Although quarterly data are used, the lag structure of the model is fairly simple. Domestic monetary conditions and the rate of devaluation are both found to positively affect consumer prices in a statistically significant manner. The magnitudes of the exchange rate coefficients exceed those of the monetary variables in each of the different sample periods. Leiderman (1984) utilizes vector autoregressions to analyze inflation in Colombia and Mexico. Changes in the rates of growth in the economy and the money supply are included in each model. In the case of Colombia, the results indicate that variations in the rate of change of Ml systematically affect the CPI, but not the converse. From a policy perspective, this implies that monetary authorities do not engage in "accommodative" measures in response to production and inflationary shocks. This result may stem from 9 the usage of monetary policy to achieve goals other than price stability. These include economic growth and export diversification. From an econometric standpoint, this result is also important because it implies that unidirectional Granger causality exists between M1 and the CPI in Colombia. A number of recent studies have examined inflationary dynamics in the United States. Koch, Rosensweig, and Whitt (1988) investigate the relationship between the exchange rate and consumer prices. Cross correlation functions are used to suggest the number of lags to be included in the regression equations. Granger causality tests imply a unidirectional channel of influence from the exchange rate to prices. The inflationary impacts associated with a change in the international value of the dollar are found to extend more than 12 months. Fullerton, Hirth, and Smith (1991) consider the effects of exchange rate and other financial and commodity price variable movements on the CPI in the United States. Transfer function ARIMA analysis indicates that inflationary impacts resulting from variations in the exchange rate generally take more than a year. Credit conditions, as proxied by interest rate spreads, are found to influence consumer prices relatively quickly. The slow speed of adjustment from the exchange rate compared to domestic financial conditions may reflect incomplete pass-through effects which characterize large industrial economies. 10 This is in contrast to what might be expected for a smaller economy such as Colombia's, where pass-through effects, statistically significant relationships between international currency values and inflation, are often strong and relatively quick (see Leith, 1991). Empirical evidence reported elsewhere (Edwards, 1985) indicates that movements in the rate of devaluation are helpful in modeling nominal interest rates in Colombia. Unfortunately, the latter study does not directly test the relationship between prices and the peso/dollar exchange rate. 2.3 Methodology The methodology utilized in this paper is similar to the multiple ARIMA approach applied to the United States by Fullerton, Hirth, and Smith (1991). This technique does not apply any a priori modeling restrictions on the equations estimated. It is a useful procedure for investigating high frequency time series data because it can accommodate different lag structures in a flexible and efficient manner. Equations developed using this approach are also easy to simulate and can help analyze policy proposals. Univariate ARIMA equations are estimated for the stationary components of the CPI, Ml, and REX series. The results are then used to specify and estimate an ARMA transfer function. A weakly stationary series is defined as one whose mean and variance do not change over time. Stationarity in 11 the means of the series is attained through regular and seasonal differencing. Logarithmic transformations are used to induce homoscedasticity (see Pankratz, 1983). The general form of a univariate ARIMA model estimated for the CPI can be written as follows: (2.1) Pt = [Q(B)QS(B)Ut] / [H(B)Hs(B)] where Pt is a stationary series derived from the original CPI series, B is a backshift or lag operator, Q(B) is a moving average polynomial of order q, QS(B) is a seasonal moving average polynomial of order qS, Ut is the disturbance term, H(B) is an autoregressive polynomial of order p, and Hs(B) is a seasonal autoregressive polynomial of order pS. The ARMA model for the CPI is also used in the estimation of the transfer function equations. To examine the effects of other variables on the CPI, transfer function ARIMA models are estimated. These models are used to determine if the input series are incrementally useful in explaining the variation of the CPI beyond the information obtained within the price index itself (see Box and Jenkins, 1976, ch. 11). Prior to estimating a transfer function equation, cross correlation functions (CCF) are used to indicate which lags of an input series may contribute incremental information to the univariate ARMA model already estimated for the CPI. Residuals from the three univariate 12 equations are used to calculate CCFs between the CPI and the other variables (see Chatfield, 1984, p. 173). Statistical, or Granger, causality is assumed to be unidirectional in transfer ARIMA models. If this assumption is reasonable, movements in REX and M1 will chronologically precede statistically significant changes in CPI. The converse will not hold if causality is unidirectional. To examine the possibility that feedback or endogeneity exists between the CPI and the other series, ordinary least squares regression is used to construct Granger causality F-tests. The general form of the transfer function can be written in the following manner: (2.2) Pt = [W(B)Rt + G(B)Mt + Q(B)Qs(B)Ut] / [H(B)Hs(B)] where the univariate model presented in Equation 2.1 is augmented by incorporating general order polynomials, W(B) and G(B), for the respective input variables. Rt and M. represent stationary input series derived from the exchange rate and the money supply data discussed above. Because the analysis is conducted within a dynamic framework, coefficient restrictions are not hypothesized, but the sums of the coefficients associated with each individual input series are expected to be positive. 13 2.4 Estimation Results Data used in this chapter are from the central bank in Bogota. Monthly observations for all three series are published in Revista del Banco de la Rep~blica (for example, see Cabrera and Montes, 1978). The sample period, January 1967 through December 1990, corresponds to the crawling peg era of exchange rate determination in Colombia. Similar to Cabrera and Montes (1978), a logarithmic transformation of the CPI is taken prior to regular and seasonal differencing to obtain stationarity. The same steps are taken with respect to REX and Ml prior to modeling. Results of the stationarity tests for all three series appear in Table 2.1. The unit root tests are performed with both constant and trend terms. Autocorrelation and partial autocorrelation functions suggested the forms of the univariate ARIMA equations reported in Table 2.2. Despite the presence of 11 years of additional data, the results confirm the AR(l), SMA(12) specification employed by Cabrera and Montes (1978, p. 1137) to model the CPI. Parameter estimates reported in Model 2.3 carry the same signs and are similar in magnitude to those estimated in the previous study (0.450 versus 0.645, and -0.924 versus -0.858). Equation 2.3 is incorporated in the estimation of the transfer function equations. Table 2.1: Unit Root Tests for Stationarity Series Aug Dickey-Fuller t-stat MacKinnon crit value P 9.081 (with const, trend, 1 lag) -4.001 (1% lvl) R 5.252 (with const, trend, 1 lag) M -13.195 (with const, trend, 1 lag) Table 2.2: Univariate ARIMA Models Parameters + 0.450*Pti (11.397) - 0.924*Ut.12 (15.900) Q(38) = 52.222 + 0. 795*Rt1 (27.671) - 0.784*Ut12 (13.148) Q(38) = 31.893 - 0.229*Ut.i (4.465) - 0. 634*Ut12 (10.558) Q(38) = 59.339 Model 2.3 Pt CPI 0.001 (10.186) 2.4 Rt REX 0.001 (8.827) 2.5 Mt 0.001 (7.715) The sample period is January 1967 December 1990. Numbers in parentheses are computed t-statistics. Ljung-Box Q-statistics calculated for 38 lags are reported. 16 Ordinary least squares F-tests are used to determine if unidirectional Granger causality exists between consumer prices and the two input series. Results for the F-tests, calculated for 12 and 18 lags, are reported in Table 2.3. The tests constructed to examine the relationship between the CPI and M1 utilize seven years of additional data, but support the conclusions reported by Leiderman (1984). Because monetary policy in Colombia appears to be conducted independently of inflationary shocks, the transfer function methodology can be used to measure the effect of the money supply on inflation. Similar to results reported for the United States (Koch, Rosensweig, and Whitt, 1988), the causality tests for the CPI and REX series indicate that changes in the inflation rate do not precede systematic variations in the exchange rate. From an historical perspective, the result is not surprising. There have been several episodes during the crawling peg era in Colombia when authorities have permitted the exchange rate to become overvalued by failing to devalue the local currency quickly enough to account for inflationary differentials with major trading partners. Typically, this has tended to take place following "coffee bonanzas" when Colombian international reserves are high due to strongly positive merchandise trade surpluses (Kamas, 1986; Ocampo, 1983). Table 2.3: Granger Causality Tests Causality Direction Number of Lags Computed F-stat CPI => REX 12 0.834 CPI => REX 18 0.799 CPI => M1 12 0.009 CPI => M1 18 0.278 Degrees of freedom for the regressions with 12 lags: 12 for the numerator and 263 for the denominator. Degrees of freedom for the regressions with 18 lags: 18 for the numerator and 251 for the denominator. 18 There have also been periods when local price changes have moved in concert with those of Colombia's trading partners, rendering unnecessary any modification in the crawling peg. As a result, the rate of devaluation has not always been adjusted proportionately to variations in the domestic rate of inflation. From an economic policy perspective, the results in Table 2.3 indicate that Colombian monetary authorities take into account goals and variables other than domestic inflation when determining rate of devaluation embodied in the crawling peg for the peso. Econometrically, this implies that transfer function ARIMA analysis can be used to model the influence of the international value of the peso on domestic prices. Similar to the CPI, stationarity in the REX and M1 series is attained after logarithmic transformation, and seasonal and regular differencing. Unit root tests for both variables are reported in Table 2.1. As noted above, univariate equations for the exchange rate and money supply series appear in Table 2.2. These equations also exhibit good statistical traits such as high computed t-statistics and relativley low Q- statistics. Residuals from the three univariate models were used to estimate CCFs containing 18 lags. Both CCFs indicated that the principal effects resulting from a change in either input variable are incorporated in the CPI within a relatively short period. 19 Transfer function equations are reported in Table 2.4. Models 2.6 and 2.7 include only one independent variable, REX and Ml, respectively. Model 2.8 includes both input series. The exchange rate is included with lags of 2 and 10 months. The money supply is included with a 9-month lag. Autoregressive terms at lag 1 and seasonal moving average terms at lag 12 are included in all three equations. Similar to Montes and Candelo (1982), the exchange rate input coefficients are larger than that of Ml. Inclusion of the input series improves the Q-statistic estimated from the residuals associated with each equation. Virtually all of the improvement in the white noise test results from the introduction of the lagged stationary exchange rate data. Coefficients estimated for these series are statistically significant in Model 2.6 and Model 2.8. The lagged stationary component of the money supply exhibits the hypothesized sign, but is not significant at the 5-percent level. Because money is a useful predictor of inflation in Colombian macroeconometric models using annual data (Fullerton, 1993a), the results encountered in this chapter are unexpected and further research is warranted. Another possibility is that M1 may not be the correct money stock measure with respect to Colombian price dynamics. Studies completed for other economies indicate that broader liquidity aggregates such as M2 may be useful (Hallman, Porter, and Table 2.4: ARIMA Transfer Functions Parameters + 0.459Pt.i (11.623) + 0.201*Rt2 (2.215) - 0.890Ut.12 (15.190) + 0.076Rt.10 (4.088) Q(38) = 39.826 + 0.447*Pti (11.334) + 0.030*Mt.9 (1.638) - 0.929*Ut12 (15.974) Q(38) = 51.777 + 0.457*Pt.I (11.570) + 0.216*Rt2 (2.384) + 0.034*Mt9 (1.839) - 0.895Ut.12 (15.235) + 0.077*RtI0 (4.133) Q(38) = 39.389 Model 2.6 Pt CPI 0.018 (8.423) 2.7 Pt CPI 0.018 (9.476) 2.8 Pt CPI 0.018 (8.113) The sample period is January 1967 December 1990. Numbers in parentheses are computed t-statistics. Ljung-Box Q-statistics calculated for 38 lags are reported. 21 Small, 1991, and Ikhide and Fullerton, 1995). Unfortunately, money supply estimates other than M1 currently do not exist for Colombia. This problem is fairly widespread in South America and has long posed difficulties in the analysis of monetary economics in the region (see Fullerton and Kapur, 1991). 2.5 Policy Simulation Results The Gaviria administration announced in December 1990 that it would attempt to lower the inflation rate to 22 percent in only twelve months (for discussion, see Fullerton, 1991). To attain this goal, two principal tools were to be employed. After an extended period of aggressive devaluation, the nominal rate of devaluation for the peso/dollar exchange rate was to be reduced. Policymakers also announced that growth in the money supply would be limited to 22 percent. To examine the potential effects of these policy innovations on the CPI, simulations are created using Equation 2.8. In the first exercise, the 12-month growth rate for Ml is assumed to be immediately limited to 22 percent for an initial 12-month period, and then lowered to 19 percent the following year. Similarly, the 12-month crawling peg rate of devaluation is assumed to be held at 22 percent throughout the first year, and later be revised downward to 19 percent during the following year. The 19 percent rates of change are chosen to reflect longer-term policy goals discussed by the 22 government, including eventual attainment of a 15 percent annual rate of change for consumer prices (Fullerton, 1991). Simulating Model 2.8 under these policy assumptions yields several interesting results that are reproduced in Table 2.5. During the first six months, the inflation rate oscillates near 29 percent. Subsequently, the 12-month rate of change in the CPI declines sharply to 24 percent by year- end. During the next 12 months, disinflation subsides as the severity of the cutbacks in the rates of change of both input variables is moderated. Year-end inflation declines to 22 percent under this scenario. As will be illustrated below, not all of the 10-point improvement can be attributed to the government policy package. Not surprisingly, Colombian monetary authorities did not introduce the exchange rate and credit policy changes in the abrupt manner depicted above. Accordingly, Equation 2.8 is also simulated under an alternative set of assumptions whereby intermediate policy steps are implemented more gradually. Additionally, actual exchange rate and money supply data for the first six months of 1991 are employed. These data reflect slower attainment of the new policy goals espoused by the Finance Ministry and the Central Bank. Subsequent movements in the 12-month rates of change for the exchange rate and the money supply are assumed to steadily decline to 19 percent by December 1992 for both variables. Table 2.5: Immediate Implementation Policy Simulation Results Month PCHYA(CPI) PCHYA(REX) PCHYA(MI) 1 32.1 22.0 22.0 2 30.7 22.0 22.0 3 28.9 22.0 22.0 4 29.1 22.0 22.0 5 29.6 22.0 22.0 6 29.2 22.0 22.0 7 28.9 22.0 22.0 8 27.9 22.0 22.0 9 26.8 22.0 22.0 10 26.2 22.0 22.0 11 25.2 22.0 22.0 12 24.1 22.0 22.0 13 23.5 19.0 19.0 14 23.3 19.0 19.0 15 22.5 19.0 19.0 16 22.5 19.0 19.0 17 22.4 19.0 19.0 18 22.4 19.0 19.0 19 22.4 19.0 19.0 20 22.4 19.0 19.0 21 22.4 19.0 19.0 22 22.3 19.0 19.0 23 22.1 19.0 19.0 24 22.1 19.0 19.0 24 Results of the second simulation exercise are reported in Table 2.6. The more rapid rate of depreciation allows the 12- month inflation rates to remain above 30 percent throughout the first semester of the test period. Steady declines in the rates of change for both REX and M1 force inflation to decline noticeably during the next six months, eventually reaching 25 percent. Year-end inflation slows to 22 percent during the subsequent 12-months of the simulation period, when growth rates for the input variables gradually decline to 19 percent. It is useful to note that the second set of medium-range simulation results reported in Table 2.6 do not vary significantly from those obtained in the first exercise. Despite imperfect implementation efforts, these results indicate that the government can still attain its stated policy goals. Because a 24-month period is still needed to lower the rate of change in the CPI by 10 percentage points, the government policy claims are again shown to be slightly optimistic. Furthermore, as shown in the final simulation exercise presented in Table 2.7, much of the improvement in the inflation rate cannot be attributed to policy design alone. Of course, policy indecision can also result in no progress being made toward achieving either intermediate target. To examine the potential consequences associated with such an eventuality, model simulations were also tested in Table 2.6: Gradual Implementation Policy Simulation Results Month PCHYA(CPI) PCHYA(REX) PCHYA (Ml) 32.1 30.7 30.7 30.6 30.8 30.3 29.8 28.3 27.0 26.5 25.8 24.6 23.8 23.6 23.2 22.9 22.8 22.7 22.7 22.6 22.5 22.4 22.3 22.3 29.9 28.9 27.8 26.8 25.8 23.6 23.0 22.5 22.0 21.5 21.0 20.5 20.0 19.9 19.8 19.7 19.6 19.5 19.4 19.3 19.2 19.1 19.0 18.9 28.2 22.3 25.0 22.6 30.3 26.2 26.0 25.5 25.0 24.5 24.0 23.5 23.0 22.5 22.0 21.5 21.0 20.5 20.0 19.8 19.6 19.4 19.2 19.0 Table 2.7: No Implementation Policy Simulation Results Month PCHYA(CPI) PCHYA(REX) PCHYA (Ml) 1 32.2 30.0 30.0 2 31.9 30.0 30.0 3 31.6 30.0 30.0 4 31.2 30.0 30.0 5 30.9 30.0 30.0 6 30.6 30.0 30.0 7 30.3 30.0 30.0 8 29.9 30.0 30.0 9 29.6 30.0 30.0 10 29.4 30.0 30.0 11 29.2 30.0 30.0 12 28.8 30.0 30.0 13 28.5 30.0 30.0 14 28.3 30.0 30.0 15 28.0 30.0 30.0 16 27.7 30.0 30.0 17 27.4 30.0 30.0 18 27.1 30.0 30.0 19 26.8 30.0 30.0 20 26.6 30.0 30.0 21 26.4 30.0 30.0 22 26.2 30.0 30.0 23 26.3 30.0 30.0 24 26.2 30.0 30.0 27 which the growth rates of the input series were held constant at 30 percent per year. In the absence of any change in the conduct of monetary management and depreciation rate determination, the annual rate of inflation stabilizes at 26 percent by the end of the 24-month simulation period. That rate is well above the announced government target range. More importantly, the 6-point improvement which results in a simulation exercise in which no progress is made with respect to the intermediate policy targets. This result indicates that only 40 percent of policy attainment embodied in Tables 2.5 and 2.6 is by government design. The remaining six-tenths of the 10-point reduction in the annual rate of change in consumer prices would, on the basis of the above modeling and simulation framework, have resulted anyway. 2.6 Conclusion In response to growing inflationary pressures, economic policymakers in Colombia announced that the inflation rate would be slashed by 10 percentage points to 22 percent over a 12-month period. Two principal tools were selected to foster disinflation. The nominal rate of devaluation for the peso/dollar exchange rate was cut and the rate of growth of the money stock was reduced. This chapter examines the empirical relationship between those variables and the consumer price index. Econometric results are similar to those reported in previous studies for 28 Colombia and the United States. Transfer ARIMA functions are estimated and simulated to determine the potential impacts of the new policies. These exercises indicate that substantial progress in the anti-inflationary program may be attained following the implementation of said policy efforts, although not as quickly as stated by government officials. More importantly, over half of the 10-point gain results even if the rates of nominal currency devaluation and money supply expansion are held constant. Because of structural economic and administrative policy changes taking place in Colombia, additional research will eventually be necessary. Future studies may find it useful to consider the effects of other variables such as wage rates, industrial capacity utilization, and commodity prices on the CPI. Given the insignificant, at the 5-percent level, coefficient associated with M1, model estimation utilizing alternative series designed to reflect monetary conditions may also prove helpful. Introduction of new variables such as wage rates may necessitate usage of an estimation technique different from the transfer function methodology described above. This is due to the possibility that feedback effects, or simultaneity, may exist between the CPI and other potential input series. At present, the Colombian economy is being opened to international trade and a free-market exchange rate system is slated to be implemented. These policy innovations could 29 potentially render the above parameter estimates obsolete. ARIMA intervention analysis (Box and Tiao, 1975) may prove beneficial in subsequent empirical research designed to examine this possibility. It is interesting to note, however, that the similarities between the empirical results reported in this paper and those analyzed in earlier studies indicate that agent responses have generally been relatively inelastic with respect to changing monetary and exchange rate policies in Colombia. CHAPTER 3 SHORT-TERM PRICE MOVEMENTS IN ECUADOR 3.1 Introduction Similar to other Latin American economies, Ecuador has faced persistently high rates of inflation in recent years. Although inflation was substantially lower in 1993 and 1994, excessive money supply growth in early 1995 clouded prospects for additional short-term improvements. Prior to the first- quarter 1995 border skirmish with Peru, the government had signed a stand-by loan agreement with the International Monetary Fund that established a goal of reducing the inflation rate to 15 percent over a 19-month time frame (Banco del Pacifico, 1994a). Not withstanding government assurances that inflation would decelerate to its target rate by year-end 1995, very little econometric analysis using short-term forecasting methods appears to have been relied upon in developing the new policy targets. In its attempt to slow price movements, the Duran Ballen- Dahik Garzozi administration introduced a variety of new policy measures. They include import liberalization, fiscal austerity, and a slower rate of currency depreciation. By reducing price pressures, the government hopes to improve economic welfare by enabling the Ecuadorian economy to operate 31 more efficiently. This argument is very similar to those aired in advanced economies such as the United States (Motley, 1993) and analyzed in other developing nations (Zind, 1993). What is unique, however, is the magnitude of the disinflationary goals set by the Ecuadorian policymakers. As a result, short-run price stabilization has become the center piece of government policy efforts in Ecuador. This chapter examines potential results associated with the two principal adjustment tools, money supply growth and exchange rate movements. Despite ongoing difficulties with inflationary uncertainty and the ambitious nature of current policy goals, careful econometric analysis of short-run price movements in Ecuador has not previously been conducted. To bridge this gap, a modeling framework is proposed, tested, and used to develop policy simulation exercises for monthly Ecuadorian price data. In contrast to the time series methodology utilized for Colombia, an econometric approach is followed in the analysis conducted for Ecuadorian price movements. Selection of this alternative approach was motivated by two factors. First was a discussion on quantitative analysis of developing country inflation held at the 32nd International Atlantic Economic Conference. Second was interest expressed by economists at the Research Department of the Central Bank in Quito with respect to attempting to develop a short-run inflationary model similar to the long-run monetary-import cost approach 32 incorporated in Fullerton (1993b). Subsequent sections of this chapter offer a review of the literature, theoretical model, and empirical results. Suggestions for future research are summarized in the conclusion. 3.2 Literature Review The seminal research on inflationary dynamics in developing countries was conducted on Chilean data by Harberger (1963). That early paper interestingly points out that analyzing nominal data in level form could result in spurious correlations in equations estimated for highly inflationary economies. To circumvent this problem, percentage rates of change are utilized in a linear regression framework based on the quantity theory of money. What became known as the "Harberger" framework incorporates real income, current and lagged values of the money supply, and the opportunity cost of holding cash balances. The success of this initial effort conducted on Chilean data spurred a series of replicated studies for other developing countries. Vogel (1974) estimates an inflation equation for several Latin American economies, including Ecuador, using annual data. Results confirm the overall usefulness of the Harberger model. Unlike the study at hand, Vogel utilizes a sample period during which inflation averaged less than 4 percent per year in Ecuador and the exchange rate was fixed. 33 Following numerous applied econometric studies utilizing this approach, it became apparent that its reliance on domestic variables alone often provided unsatisfactory results. Bomberger and Makinen (1979) provide a thorough examination of the Harberger model using quarterly data for Korea, Taiwan, and Vietnam. Extensive testing is conducted using quarterly data in order to establish whether a suitable characterization of inflation is provided. Encouragingly, the parameter estimates do not appear sensitive to the time period selected. However, the elasticities with respect to money and real income are not always unitary as hypothesized. Also, the coefficient signs for the cost of holding money variables are sometimes negative. Hanson (1985) extends the Harberger framework in a systematic fashion to incorporate an important missing component, import costs. An implicit cost function is utilized to derive an aggregate supply curve which includes local prices of imported inputs. When the underlying production function is homogeneous of degree one, inflation becomes a weighted sum of money supply changes and import prices. This is important for studies using higher frequency data if the problem of measurement bias engendered by interpolated values of real output, generally published on either a quarterly or annual basis in developing countries, is to be avoided (see Bomberger and Makinen, 1979). The model also implies the elasticity of inflation with respect to money 34 growth is less than one. Empirical results in the Hanson article strongly support the inclusion of import prices or the rate of devaluation in models of inflation. Subsequent research has provided additional evidence in favor of the augmented Harberger-Hanson approach wherein the effect of import prices on inflation is considered. Koch, Rosensweig, and Witt (1988) and Fullerton, Hirth, and Smith (1991) both report positive linkages between the trade- weighted exchange value of the dollar and consumer prices in the United States. These empirical studies indicate a unidirectional channel of influence from the exchange rate to domestic prices exists in the United States economy. As will be discussed below, causality direction has important implications for both model form and estimation technique. Developing country studies have also confirmed the usefulness of an augmented modeling treatment of inflationary dynamics. Sheehey (1976) reports some of the early econometric work along these lines. Sheehey (1980) reaches additional conclusions on the basis of empirical tests that indicate that accurate assessment of austerity policy efforts will likely require explanatory variables representing cost push factors. More recently, Brajer (1992) provides evidence that the latter category of models may offer better specifications than those which rely solely on domestic economic factors. Conclusions in that article are reached on the basis of F-tests for different regressor sets. Similarly, 35 Fullerton (1993b) successfully imbeds a variant of this approach in a large-scale macroeconometric forecasting model for Ecuador using annual data. There have been very few dynamic models estimated on the basis of monthly data for developing economies. Given that most business decisions in highly inflationary countries are reached within a short-range context, this is an area which needs to be addressed. As detailed in Chapter 2, Fullerton (1993c) empirically examines Colombian anti-inflationary efforts utilizing monthly data with an ARIMA transfer function. The estimated model is found to generate realistic simulation scenarios for policy analysis. The results also support the hypothesis of inflation rate inelasticity with respect to monetary growth. As in the Chilean equation reported by Hanson (1985), and the Argentine model presented in Sheehey (1976), exchange rate price effects are found to outweigh the monetary coefficient. The latter is somewhat surprising given that imported goods and services comprise less than 20 percent of Colombian gross domestic product. 3.3 Theoretical Model Harberger's (1963) model is based on the traditional quantity theory of money equation: (3.1) MV = PQ, 36 where M represents some measure of the money stock, V is the velocity of circulation, P is the price level, and Q is real output. Velocity is not assumed to be constant. Instead, it is hypothesized to be a predictable function of other macroeconomic variables such as the cost of holding cash balances. Given the typical variability of velocity in many Latin American economies, this aspect of the theoretical model is potentially important (see Clavijo, 1987). To utilize percentage changes, the variables can be transformed by natural logarithms and first differenced. Introduction of a time subscript, and rearrangement of the terms, yields the basic Harberger equation: (3.2) DPt = DMt DQt + D(DPt.1), where the last term results from substituting for velocity and D represents a difference or backshift lag operator. Usage of the lagged change in the inflation rate to proxy for the implicit cost of holding money is motivated by the fact that developing countries such as Ecuador have frequently imposed government regulations on interest rates. The latter have occasionally caused savings and loan rates to become negative in real terms. Unadjusted interest rates from these periods in Ecuadorian economic history do not, therefore, provide accurate estimates for the cost of holding idle cash balances. 37 Equation 3.2 implies that inflation will vary positively with the money supply and inversely with respect to real output. A statistically significant intercept term will enter the estimated equation if there is a trend in the velocity of circulation. If only contemporaneous lags of M and Q enter in the equation, the parameters for both variables are hypothesized to be unitary. This can be tested empirically with the following specification: (3.3) DPt = ao + a1DMt a2DQt + a3D(DPt.1) + u3, where al and a3 are hypothesized to be positive, and the absolute values of al and a2 should both be statistically indistinguishable from one. The last argument in the expression represents the disturbance term. Hanson (1985) proposes an implicit cost function dual of an aggregate production function which is homogeneous of degree one. Derived output supply functions from this framework will be homogeneous of degree zero in input and output prices. Equation 3.4 expresses this relationship using logarithmic first differences: (3.4) DQt = b0 + bIDPt b2DPIt + u4, where PI represents imported input prices. When the relative prices of imported inputs increase, output is assumed to 38 decline. The standard homogeneity assumptions for production and derived supply relations imply that bI b2 = 0. Equation 3.4 can be substituted into Equation 3.3 to eliminate the output term from the expression to be estimated. As noted in the literature review, this step is useful for avoiding interpolation bias in empirical studies of monthly inflation for countries such as Ecuador where GDP is published at quarterly and/or annual frequencies. The resulting equation can be written as follows: (3.5) (1 + a2bl)DPt = a0 a2b0 + a1DMt + a2b2DPIt + a3D(DPt.1) + u5. Equation 3.5 can be further simplified prior to estimation. Dividing through by the left-hand side constant term and rearranging terms such that the price series remains as the dependent variable yields the following relation: (3.6) DPt = co + cDMt + c2DPIt + c3D(DPt.1) + u6, which also has testable properties. Importantly, the coefficient on the monetary variable, cl, is now hypothesized to be significantly less than one. Also important, with the possible exception of the intercept, all of the regression parameters in Equation 6 are expected to be positive. 39 Several other properties of this model are worth noting. In particular, the theoretical coefficient restrictions described earlier have interesting implications. Namely, a, and a2 are hypothesized as equal to one, and bI and b2 are equal in absolute value in the version of the model developed thus far. Substitution into Equation 3.5 implies cI + c2 = , which can also be tested. As indicated in the literature review, Equation 3.6 has provided a useful framework for analyzing quarterly and annual inflation rates. But because the lag structure in this model is fairly short, it may require additional modification prior to estimation. This possibility does not reflect any deficiencies in the theoretical model as such, but arises due to the fact that short-term models rely upon monthly data. As a result, if the inflationary impact of a change in the money supply is felt over the course of one calendar year, the implied lag structure for a model estimated with data published at a monthly frequency would potentially range up to 12-months in length. Equation 3.7 takes into account this empirical issue which has confronted and confounded researchers for many years (see Laidler, 1993): (3.7) DPt = co + c1DMti + c2DPIt.j + c3D(DPt.l.k) + u7, where lag subscripts i, j, k = 0, ..., n, respectively. 40 The above model provides an attractive starting point for examining inflationary trends in an economy. It is not, however, without potential problems for analyzing price movements in a relatively high inflation country such as Ecuador. A principal concern arises from the fact that Equation 3.7 treats all of the regressors as exogenous or pre- determined. In doing so, it does not allow for the possibility of statistical feedback or endogeneity between the left-hand and right-hand side variables. If a central bank yields to political pressures and engages in accommodative monetary policy in the face of inflation shocks, this assumption would be violated. As noted in Chapter 2, research conducted using higher frequency data for Colombia indicates that the causality paths in that economy are unidirectional as implied by Equation 3.7 (see Fullerton, 1993c, and Leiderman, 1984). While short-term forecasting models for Ecuadorian inflation have not been previously developed, domestic prices, monetary aggregates, and import prices are modeled simultaneously in the macroeconometric model estimated using annual data by Fullerton (1993b). It would not be surprising if the feedback relations encountered in that paper also emerge in the monthly time series utilized below. As noted elsewhere, monetary authorities in Ecuador have occasionally been forced to yield to political pressures (Garlow, 1993). Granger causality 41 tests will be used to test the severity of this potential problem. A second possible concern arises from utilizing first differenced, log-transformed time series data in the equation to be estimated. If the resulting series are stationary, the equation can be estimated without risk of obtaining spurious correlations in the results. As shown in many studies of hyperinflationary economies, however, higher order differencing may be required to induce stationarity during periods in which prices increase rapidly (Engsted, 1993). Because Ecuador has not undergone any hyperinflationary episodes, first differencing should remove nonstationary trends from the variables in question but this assumption must be tested. The latter tests are accomplished below via a battery of unit root tests, not all of which are reported. A third concern arises from the fact that monthly import price deflators do not exist for Ecuador. To circumvent this problem and also avoid interpolation bias, a trade-weighted exchange rate index is used as a proxy for imported input prices. The index utilized was developed econometrically and takes into account export and import volume changes with Ecuador's major trading partners. It also offers a single monthly index for periods when the government has instituted multiple exchange rate systems (Fullerton, 1989b). To construct the currency index, individual currency weights are calculated as the sum of imports and exports with 42 each of ten major trading partners and divided by total international trade in each year. Over the sample period, annual trade with Ecuador's top ten import sources and export destinations accounts for more than 70 percent of its total trade volume in any given year. Products of the bilateral trade coefficients and the respective currencies are then used to construct the exchange rate index using a geometric mean. The latter method is selected to avoid problems which can potentially result for indexes constructed using arithmetic means during periods of inflationary variability (see Batten and Belongia, 1986, Dutton and Grennes, 1987, and Kercheval, 1987). For periods when multiple exchange rates were instituted, a blended index is calculated. Weights for the free-market and official government intervention exchange rates are obtained from the Central Bank publication, Inforuacift Estadistica Mensual (various issues). Econometric results for total and disaggregated imports, and non-petroleum exports indicate that the blended rate provides a more accurate measure of the appropriate currency basket for Ecuador (Fullerton, 1989b). Diagnostic tests were also conducted using currency baskets with different numbers of trading partners. Introduction of the trade-weighted exchange rate index to Equation 3.7 causes the model to be estimated to take the following form: 43 (3.8) DPt = go + g1DMt-i + g2DTWXt- + g3D(DPt-j-) + u8, where TWX stands for logarithmic first differences of the nominal version of the monthly trade-weighted exchange rate index calculated for the sucre by The WEFA Group (formerly Wharton Econometrics). Although incorporation of the monthly exchange rate index avoids the problems associated with implicit price deflator interpolation, it may introduce other problems due to the fact that import prices in Ecuador are affected by global supply and international demand conditions in addition to exchange rate movements (see Fullerton 1993b). A fourth observation regarding potential pitfalls associated with the theoretical specification of the model is worth noting. Only one class of factor input prices, that for imports, is included. While this represents an improvement over the original Harberger framework, it may overlook additional important inputs such as labor. If the version of the model developed herein omits relevant variables to the inflationary process in Ecuador, it is likely that estimated residuals associated with the empirical version will not be randomly distributed. If this is the case, then correction for serial correlation will be necessary. Extension of the model to overcome this problem cannot currently be accomplished due to data availability. 44 3.4 Estimation Results In order to examine whether the working series included in Equation 3.8 are stationary, unit root tests are conducted for each series utilized in the model. Estimation is conducted for the 1964-1994 sample period for which data are available. Applying unit-root tests to what could be considered a relatively short time span may be risky due to the fact that these tests typically have low power unless long-run data sets are used (Hakkio and Rush, 1991). Because time series data in Latin America generally date back to 1957 at most, there is little that can be done to circumvent this potential problem. Augmented Dickey-Fuller t-statistics are estimated for equations with both intercepts and trends. These results, compared against the corresponding MacKinnon critical value, appear in Table 3.1. In all cases, tests for unit roots in the first differenced log transformed series for consumer prices, money, and the trade-weighted exchange rate index are rejected at the 1-percent level. Based on this evidence, the first-order differenced series used to estimate Equation 3.8 are assumed to be stationary. As specified, the model is explicitly built around a set of unidirectional causality relations from movements in the regressors to consumer prices. To examine whether the absence of simultaneity in the model is plausible, a set of Granger causality tests are calculated for the stationary components 45 of the series of interest. These results are reported in Table 3.2 for lags of 6, 12, 18, and 24 months. Tests are conducted for prices and the money supply, as well as prices and the trade-weighted exchange rate. Similar to the Colombian results reported in Chapter 2, movements in M1 do not appear to be systematically preceded by changes in the CPI in a statistically significant manner at the 5-percent level. Essentially, this implies that monetary policy in Ecuador is conducted in a manner that is not accommodative of price shocks. Although central bank linkages to the executive branch of the government are relatively strong, steps have been taken in recent years to increase both monetary policy autonomy and currency stability (Banco del Pacifico, 1994b). Implications based on the Granger causality tests estimated for consumer prices and the exchange rate series are less clear. At shorter lag lengths, the null hypothesis that changes in consumer prices do not lead to subsequent changes in the international value of the sucre is rejected at the 5- percent level. This conclusion, similar to the Colombian results obtained by Fullerton (1993c) and Kamas (1995), is not upheld at longer lag lengths and makes it difficult to reach concrete conclusions regarding potential feedback effects between the two series. However, truncation of the longer lag lengths can bias the hypothesis toward incorrect rejection of the null hypothesis (Feige and Pearce, 1979). Consequently, Table 3.1: Unit Root Tests for Stationarity Series Aug Dickey-Fuller t-stat MacKinnon crit value P 9.778 (with const, trend, 1 lag) -3.995 (1% ivl) M -16.626 (with const, trend, 1 lag) TWX -13.652 (with const, trend, 1 lag) Table 3.2: Granger Causality Tests Causality Direction Number of Lags Computed F-stat CPI => M1 6 1.659 CPI => Ml 12 1.309 CPI => M1 18 1.420 CPI => M1 24 1.509 CPI => TWX 6 2.900 CPI => TWX 12 2.214 CPI => TWX 18 1.664 CPI => TWX 24 1.029 48 a unidirectional channel of influence from exchange rate movements to prices appears to be a reasonable assumption. Given the latter, model estimation is conducted without resorting to instrumental variables, or developing a system of simultaneous equations, and the resulting coefficients are assumed to be unbiased and consistent. Regression results for Equation 3.8 are summarized here: (3.9) DPt = 0.015 + 0.013*DMt + (10.224) (0.797) 0.028*DMt.4 + 0.039*DMt.9 + (1.735) (2.333) 0.060*DTWXt + 0.007*DTWXt.2 + (5.932) (0.642) 0.063*D(DPt.6) + 0.405*Ut2 + (1.638) (8.439) 0. 369*Vt., (7.101) R2 0.379 S.E.R. 0.015 Log likelihood 1013.715 DW 2.087 F-stat 26.798 Prob(F-stat) 0.001, where the numbers in parentheses are computed t-statistics and Vt is the error term associated with the ARMAX model for Ut. Lag lengths of 24 months were used in the initial estimates for Equation 3.9. While a large number of the resulting coefficients were significant at the 5-percent level, serial correlation was present in the residuals. To avoid potentially spurious estimation results (see Hamilton, 1994), a nonlinear ARMAX procedure is utilized to correct for autocorrelation. This estimator (Pagan, 1974) is useful 49 because of its flexibility in handling a variety of different error generating processes. Correcting for serially correlated disturbances caused the computed t-statistics for many coefficients to become insignificant at the 5-percent level. Because inclusion and exclusion of numerous different lags did not yield clear results, the model structure reported in Equation 3.9 was selected on the basis of likelihood ratio tests. Although the relatively short lag components may seem unexpected, they were confirmed by cross correlation function analysis. To handle autocorrelation, an ARMA(2,1) specification is used to characterize the data generating process for the residuals. As would be expected in an inflationary economy, the algebraic sign of the intercept in Equation 3.9 is positive. Because the model is estimated using differenced data, this result indicates that a systematic upward trend exists in the Ecuadorian consumer price index. As hypothesized by Hanson (1985), the sum of the coefficients for the lagged monetary series is significantly less than one. Similar to Fullerton (1993a, 1993c), but unlike Kamas (1995), the exchange rate appears to play an important role in determining price movements. The coefficient for the velocity of money supply circulation proxy was not, however, significant at the 5- percent level. Although the sum of the lagged monetary aggregate parameters is less than one, the sum of those coefficients 50 with the exchange rate variable coefficients does not equal one as implied by the reduced form of the theoretical model specification. These results cast doubt upon the relatively simple version of the Harberger-Hanson framework developed above. As noted previously, this is not completely unexpected and is a potential root cause underlying the presence of autocorrelated residuals in the initial empirical results. To see if the theoretical model can be improved, an alternative version of the approach is currently under development for an economy where data shortages are less severe and a richer input structure can be handled (Kim and Fullerton, 1996). The latter effort will also benefit from the incorporation of an interest rate variable to measure the cost of holding idle cash balances, as well as an import price index rather than the exchange rate proxy utilized herein. In 1983, the Central Bank introduced new exchange rate policies that allowed the sucre to fluctuate more freely. To allow for potential parameter heterogeneity caused by structural change associated with periodic currency devaluations, a shorter sample was also used for estimation purposes. Results from these exercises, not reported here, generally support the empirical estimates presented above. Experimentation with the lag structure over the shorter sample period does not yield strong evidence of parameter instability or any other major shortcomings with Equation 3.9. 51 3.5 Policy Simulation Results From the alternative specification and sample size results, Equation 3.9 does not seem overly fragile. Alternative lag structures and ARMAX processes were compared to the model using likelihood ratio tests to evaluate the results. While it was not possible to reject the specification shown above, it does seem likely, however, that additional time series analysis will be required in order reach firm conclusions regarding inflationary dynamics in Ecuador. For that reason, the modeling and simulation results associated with Equation 3.9 should be regarded as preliminary in nature. They do, however, provide a good starting point for understanding short-term Ecuadorian price movements and assessing government policy objectives. Cognizant of the paucity of comparative research results in this segment of the literature, a variety of simulation experiments are conducted using Equation 3.9. The goal of the simulation tests is to shed light on the feasibility of attaining the inflationary stabilization targets announced by the Durdn Ballen-Dahik Garzozi administration in 1995. Sample data used to estimate Equation 3.9 only includes information available to government policy makers at the time the inflation target was announced. The simulation analyses thus satisfy the Klein (1984) and Christ (1993) criteria for forecast evaluation. Of course, additional policy analysis with different model specifications may also be useful. 52 To examine the feasibility of the government's inflation goals, four simulation exercises are conducted using Equation 3.9. The first exercise assumes immediate implementation of price stabilization plan wherein the annualized rate of growth in the money supply and the rate of devaluation are reduced to 20 percent within one month of the policy announcement. Scenario two examines rapid implementation of the price stabilization whereby growth in the money supply and the rate of devaluation are reduced to 20 percent over a 6-month phase- in period. The third simulation test analyzes the effects of gradual introduction of the anti-inflationary program with money supply growth and the rate of devaluation lowered to 20 percent during a 12-month period. The rates of money supply growth and currency depreciation reflect the revised policy targets adopted following the border conflict with Peru and the aftermath of the "Tequila effect" associated with Mexico's December 1994 peso devaluation (Banco del Pacifico, 1995). Policy simulation results are reported in Table 3.3. The impacts associated with all three implementation scenarios are striking. In each, reducing the rate of money supply growth and slowing the rate of nominal currency depreciation to 20 percent over a one-to-twelve month implementation period causes inflation to decline to less than 25 percent. This is somewhat close to the policy target established by the government in late 1994. If implementation of the stabilization program is immediate, annual consumer prices 53 increases fall to slightly more than 23 percent. In a more likely scenario under which intermediate policy targets are attained more gradually, disinflation is still fairly rapid and overall policy credibility would not appear to be at risk. To actually achieve the announced inflation target, however, does not appear likely. The final column of Table 3.3 contains the results associated with a scenario in which no progress is made with respect to lowering the rate of growth in the money stock. Similarly, nominal depreciation of the sucre is not lowered under this simulation. In contrast to the "no implementation" policy simulation results reported in Table 2.7, the inflation rate remains practically unchanged if no intermediate steps are taken by Ecuadorian monetary authorities. This result is due in large part to the fact that consumer prices in Ecuador rose on average by approximately 26.7 percent per year during the past three decades. As a result, the policy experiments illustrated in Table 3.3 have a common starting point that is almost identical to the sample period mean. On the basis of the empirical evidence obtained in this chapter, it appears that Ecuador's anti-inflationary program is fairly credible. This conclusion is predicated upon eventual deceleration in the rates of liquidity growth and currency depreciation brought about by the central bank. In light of previous monetary policy analysis conducted for Table 3.3: Policy Simulation Results Month Immediate Rapid Gradual None 1 26.1 26.2 26.2 26.2 3 24.9 25.1 25.3 26.0 6 24.3 24.8 25.1 25.8 9 24.0 24.5 25.1 25.9 12 23.2 23.7 24.2 26.0 55 Ecuador, the results obtained are pleasantly surprising (Fullerton, 1990a, offers a negative assessment of an earlier price stabilization effort). That the government's announced targets are not completely attained is in line with other studies of Latin American inflation policies (Fullerton, 1993c, and Kamas, 1995). 3.6 Conclusion An empirical model of Ecuadorian consumer price inflation is developed and estimated in this chapter by incorporating both monetary and import cost effects in a theoretically plausible manner. Specification and simulation of the model are relatively easy to accomplish. Experimentation with the estimated equation indicates that the current anti- inflationary goals of Ecuadorian monetary authorities are in large part attainable. Because the model does not pose stringent data requirements, it may be applicable to other Latin American economies where inflation remains a problem. Examples include Brazil, Colombia, Mexico, and Venezuela where authorities continue to grapple with short-term price stabilization goals. Additional econometric testing should prove useful. Initial results reported above indicate this framework will likely benefit from incorporating a more realistic conceptual model. Because ARMAX treatment of nonrandom movement in the initial model residuals was necessary, expansion of the scope 56 of the model to include additional factors such as labor costs may be required to more completely specify the inflationary process in Latin America. Doing so, however, may necessitate the usage of instrumental variables or the introduction of multiple equations which allow for potential endogeneity between prices, money, import costs, and wage rates. Even if the latter are not required for parameter estimation consistency, they could enrich subsequent policy simulation analyses. These suggested changes represent avenues for refinement to the basic model outlined above. They are not likely to result in wholesale alterations to the general framework. Similarly, it is not clear that policy simulation impacts and conclusions will change markedly due to expanding the scope of the empirical techniques presented above. But given the breadth of economic conditions prevailing across Latin America, steps in these directions may prove helpful to subsequent econometric research of this nature. Given the divergence between the theoretical model parameters and the estimated coefficients, additional empirical testing is certainly warranted. CHAPTER 4 PREDICTABILITY OF SECONDARY MARKET DEBT PRICES 4.1 Introduction While inflation has undoubtedly been one of the most hotly debated items in Latin American policy debates, issues related to external obligations have also been important topics of discussion in the region. Following the outbreak of the international debt crisis in September 1982, a secondary market for sovereign debt instruments became active in the major world financial centers. As the payments crisis spread from Mexico to the rest of Latin America, much of Africa, and parts of Asia, secondary market trades in developing country sovereign debt paper increased. Key issues in the emerging debate regarding potential solutions to the payments problem often involved the treatment of discounts from face value implied by secondary market debt prices. Accordingly, researchers began investigating different aspects of the behavior of secondary market developing country debt prices. Much of this work investigates the applicability of theoretical valuation models to assessing implied discounts from face. Other research has used secondary market prices and other financial data to evaluate the probability of payment rescheduling requests. Authors have also attempted to 58 assess the sensitivity of the secondary market to macroeconomic fundamentals, but these efforts have not investigated whether or not the discounts from face value on these obligations are predictable. Several basic questions are investigated in the chapter at hand. An important issue to be considered is the nature of the time series behavior of these prices. Under certain conditions in an efficient market, each individual series might be expected to follow a random walk. Of course, in a relatively thin secondary market such as that for developing country debt, the perfectly competitive hypothesis may not always be satisfied. If movements in the price series are nonrandom, it may be possible to relate these variations to changes in other domestic and international economic indicators which are generally included in commercial forecasts of the region. Candidate series which help assess creditworthiness include variables such as interest rates, export prices, and international reserves (see Fullerton, 1991, 1993a). Anecdotal evidence indicates that countries with good payments records such as Colombia have seen their access to international commercial credit diminish over the last fifteen years. If the "good debtor in a bad neighborhood" contagion effect is present, this will be manifested in the movements of the discounted prices. Debt prices for individual countries would consequently be affected by developments in neighboring 59 countries and correlated with the prices for the instruments of those economies and the market in general. The generalized least squares estimation procedure used below implicitly takes this possibility into account by allowing for contemporaneously correlated residuals across equations. To examine the predictability of these series, forecasts are generated using jointly estimated individual country models. For comparison purposes, baseline projections are developed using the random walk assumption that the best forecast is one of no deviation from the last observation available in each data set. Modified Theil U-coefficients are then calculated on the basis of root mean squared error (RMSE) ratios estimated for each set of model based and random walk forecasts. The U-statistic for an individual forecast step- length is equal to the ratio of the model based RMSE to that of a no change RMSE. When the resulting coefficient is less than one, the equation forecast has outperformed the random walk prediction. 4.2 Earlier Studies There have been a growing number of studies regarding developing economy external indebtedness and the secondary market in recent years. Gennotte, Kharas, and Sadeq (1987) develop a numerical debt valuation method based on a financial options pricing technique. Using liquid international reserves plus the estimated values of the capital stocks in 60 mining and manufacturing as proxies for collateral in each country, theoretical values for developing country foreign debts are simulated under different scenarios involving interest rate changes, principal due, and front-end fees. Simulation results are found to be positively correlated with secondary market prices reported for 1985. Stone (1991) examines the behavior of implied returns on secondary market sovereign debt instruments. Using an arbitrage pricing model approach, he examines the empirical relationships between secondary market returns and various macroeconomic variables. To control for cross equation disturbance simultaneity, a seemingly unrelated regression estimator is utilized. Poor equation fits and weak t- statistics indicate that movements in implied sovereign debt returns are not readily explained by arbitrage pricing techniques. Other empirical studies of sovereign debt problems have been more successful. Rahnama-Moghadam and Samavati (1991) employ probit models to examine the propensity to default. Ten different macroeconomic and international financial ratios are used to quantify the probability of rescheduling. Among the ratios found most useful in predicting debt moratoria, formal or informal, are the following: international reserves relative to imports of goods and services; international reserves to disbursed debt; disbursed debt relative to exports; disbursed debt to gross domestic product; and 61 interest payments relative to exports of goods and services, also known as the interest service ratio. Parameter estimates are based on annual cross country sample data from around the globe. The probit results discussed in the preceding paragraph were later confirmed by subsequent research which utilized data from Latin American economies (Rahnama-Moghadam, Samavati, and Haber, 1991). Reasons offered for segmenting the data in this fashion include geographic, structural, and institutional similarities among countries in Latin America. The authors also point out that economies in this smaller sample are all middle-income countries which share similar characteristics in terms of overall development. The models exhibit relatively high goodness-of-fit statistics, individual coefficients with expected algebraic signs, statistically significant parameter estimates, and coefficient stability across different specifications. As debt problems arise from a number of different determinants, it would appear that a further refinement from a regional focus to that of an individual economy may be useful. The roles of fundamental economic and financial factors in the evolution of secondary market sovereign debt prices have been directly examined in recent research. As shown by Anayiotos and de Pinies (1990), these types of characterizations are straightforward and intuitively appealing. In a statistical framework, variables selected to 62 capture market fundamentals can also be combined with regressors designed to represent exogenous risks. Using pooled observations and annual data, the econometric results of these authors show that even simple specifications can represent secondary market developing debt prices with a high degree of accuracy. Perasso (1989) also emphasizes economic factors in his study of secondary market prices. To reflect the importance of debt-equity conversions, his pricing model is derived from a profit maximizing framework that includes real interest rate measures, real costs of capital, international wage differentials (assumed to induce manufacturers to invest abroad), and individual economy performance variables. Time series and cross country annual data are pooled prior to estimation. Some coefficients are statistically insignificant or of the wrong sign, but overall empirical results for the estimated equations for secondary market prices are fairly strong. Empirical models developed by investment bankers indicate that secondary market debt prices can be modeled and forecasted (see Purcell and Orlanski, 1988, 1989). Similar to the studies mentioned above, these models also rely upon pooled cross section time series data for different countries. Regressors used to estimate equation parameters include debt to export ratios and per capita incomes. Dummy variables for payment rescheduling programs, principal payment moratoria, 63 and debt retirement agreements are also entered as right-hand side variables. Model simulations using individual country data are used to calculate specific secondary price forecasts. As mentioned previously, not all of the statistical evidence with respect to the behavior of secondary market prices leads to the same conclusions. Laney (1987) concludes that economic factors are more important as explanatory variables than political and structural risk factors. Sachs and Huizinga (1987) report regression results that indicate both economic and political variables have key roles to play in modeling developing country debt discounts. Similar to other studies, the latter also utilize pooled cross country data in calculating equation parameters. 4.3 Empirical Analysis As the literature review indicates, there have been a variety of studies with interesting econometric results published in recent years. However, none of the initial efforts attempted to model secondary market debt prices for individual countries using time series sample observations. Given the differing sources of debt servicing difficulties, a country by country examination of secondary market price movements appears warranted. Similarly, individual forecasting models or equations have not been systematically tested to examine whether debt prices can be predicted with any degree of accuracy. Despite the fact that financial 64 market participants focus almost exclusively on short-term movements in sovereign debt prices, earlier research efforts also failed to study high-frequency data from this market. In this section of the chapter, simple forecasting models are proposed, estimated, and simulated for three individual debtor countries. These equations are modeled jointly using monthly data series. As stated in the introduction, modified Theil inequality coefficients are calculated using random walk forecasts as the benchmarks to which the model projections are compared. Although the individual equation specifications employed are straightforward, recent financial market research underscores the potential success of simple forecasting models (see Granger, 1992, and Christ, 1993). This initial attempt provides a useful starting point for addressing questions regarding the predictability of secondary market developing country debt prices. Debt price data used in the empirical estimates are collected by Salomon Brothers in New York, with monthly averages published by The WEFA Group in Philadelphia. The sample period is March 1986 December 1991. Countries included in the sample are Colombia, Ecuador, and Venezuela. These countries have interesting and differing histories with respect to their external debt management practices and their individual approaches to economic policymaking in general. Colombian debt management practices have traditionally been more conservative than those of either of its neighbors. 65 Government economists have consistently treated international credit markets as sources of financing to bridge domestic savings gaps (Fullerton, 1990b). Colombian debt negotiators have never sought a Brady initiative write-off, arguing that to do so would only impair the nation's creditworthiness. Over the period from 1945 forward, Colombia has never declared an interest payment moratorium and has one of the best debt service records among developing countries which have utilized external financing sources. Ecuador declared principal and interest moratoria in 1987 as a result of the financial aftermath following the earthquake which shattered the country's transAndean oil pipeline. The latter event interrupted Ecuador's principal source of export earnings and destroyed much of its physical infrastructure. In 1989, the Borja administration resumed negotiations with commercial creditors and eventually began honoring 30 percent of the interest coming due on commercial loans (Fullerton, 1989c). Progress regarding the treatment of growing amortization and interest arrears remained elusive through the balance of the Borja government which stepped down in 1992. A new round of discussions with Ecuador's bank advisory committee began after the Duran Ballen-Dahik Garzozi government took office. Venezuela also encountered debt service problems following negative oil price shocks in 1986 and 1988. Eventually, the Perez administration rescheduled commercial 66 credits under a Brady initiative agreement with Venezuela's bank advisory committee (Fullerton, 1990c). The agreement offered bank loan syndicate members five menu options designed to relieve balance of payment pressures faced by this economy. Several Euromoney bond issues were successfully floated in subsequent periods and Venezuela temporarily regained access to international credit markets. Debt instruments for all three countries have traded at substantial discounts from face value in recent years. Given its superior service record, it is not surprising that Colombian paper generally carries a higher price than that of its two Andean neighbors. Similarly, given its higher level of payment arrears and worse economic performance, Ecuadorian paper tends to trade at sharper discounts than those of the other countries in the sample. Table 4.1 presents summary statistics for each secondary market sovereign debt price series. Discounts from face value on Colombian debt certificates were less variable than those of Ecuador and Venezuela during the March 1986 December 1991 sample period. The range and standard deviation for Colombian debt quotes are smaller than the others, while those for Ecuador are the largest of the three. The arithmetic means for each series follow the opposite pattern in terms of ranking. Data in Table 4.1 are in cents per dollar, or percent of face value of the loans, the units in which transactions are conducted at money center trading desks. Table 4.1: Secondary Market Country Average Maximum Colombia 69.990 86.00 Ecuador 30.332 68.00 Venezuela 57.635 78.50 The sample period is March Debt Price Summary Statistics Minimum Standard Deviation 50.00 10.415 10.67 18.728 31.50 14.859 L986 December 1991. 68 Each series is modeled as a function of key economic and financial variables which are easily observed as well as likely to be used by secondary market participants as indicators of the creditworthiness of the individual country. Possible regressors include lagged debt quotes, world interest rates, commodity export prices, international reserves, and domestic price indexes (for discussion, see Wakeman-Linn, 1991). As mentioned above, series such as these are typically included in macroeconometric forecasting models of Latin America due to their usefulness in predicting balance of payment movements and general economic performance (see Fullerton, 1993a, 1993b). A three-stage generalized least squares (3SLS) regression technique is used to jointly estimate model parameters (Zellner and Theil, 1962). Doing so permits incorporating potential cross-equation simultaneity effects of events such as Citibank's decision to unilaterally increase loan loss reserves in 1987. The latter is believed to have reduced overall secondary market liquidity and also reduced the attractiveness sovereign debt paper to most creditors, irrespective of their individual payment records (Gajdeczka and Stone, 1990, and Snowden, 1989). This estimator also allows for potential simultaneity between the dependent variables and the independent variables. In the case of Ecuador, this is important because feedback exists between the secondary market debt price and the 180-day LIBOR. 69 Individual models may be written conceptually as: 4.1 Pt = b0 + blxlt + ... + bnxnt + et, where Pt is the secondary market sovereign debt price series for an individual country at time period t, x1t, ..., Xnt are predetermined domestic and international variables for each economy, et is a random disturbance term, and bl, ..., bn are regression coefficients. From a theoretical modeling perspective, this approach may seem informal. Two points are relevant. First, alternative methodologies were originally considered but deemed inappropriate due to data requirements and difficulties in applying them to forecasting problems (Fullerton, 1990d). Second, the complete absence of other studies of this nature increases the value of an initial attempt to establish whether any regularities at all are present in the data (Christ, 1994). Both points were repeatedly raised by financial economists who participate in the secondary market and attended the Sovereign Debt Conference sponsored by The WEFA Group in New York in 1990. Similar observations are also made by Friscia (1993). As mentioned above, parameter estimation is accomplished using the 3SLS methodology developed by Zellner and Theil (1962). Other procedures were considered, but 70 monthly observations constitutes a fairly small sample for many time 70 series estimators. The series are not differenced prior to estimation, but are logarithmically transformed. Because the data are in levels, it is important to assess whether the series are cointegrated. The latter assumption is tested via unit-root tests on the individual model residuals (for discussion, see Hamilton, 1994). Because monthly data are used, it is not possible to utilize the same regressor variables as have been used in earlier studies incorporating quarterly or annual series from national income and product accounts. There are still a number of potential candidate series which the financial community may use as indicators of a country's creditworthiness and will potentially influence secondary market price quotes. It should be noted that sets of indicators will vary for different countries according to individual economic endowments and performance records. This argument is similar to that previously made for equations used to estimate developing country borrowing levels under different regimes (Eaton and Gersovitz, 1981). For Colombia, Equation 4.2 in Table 4.2, the regressors include a one-month lag of the debt price, the effective annual rate for the 180-day London interbank offer rate (LIBOR), and the one month change in the national consumer price index (CPI). The one-month lag on the debt price is included to provide information on how the market has valued Colombian paper in the most recent period. International 71 loans to sovereign nations typically carry variable interest rates defined in terms of a fixed spread over the 6-month LIBOR. Upward changes in the variable interest rate assessed on such loans will reduce Colombia's current account balance and exert downward pressure on secondary market prices. The rate of inflation is used as a proxy for overall economic conditions. When inflation rises, Colombian monetary authorities generally attempt to tighten credit conditions (Fullerton, Fainboim, and Agudelo, 1992). Parameter estimates for each variable have the expected arithmetic signs, but the t-statistic for the inflation term is not significant at the 5-percent level. In the case of Ecuador, Equation 4.3 in Table 4.2, only two predetermined variables are included in the three-staged least squares equation. The first is a one-period lag of the secondary market debt price series. The second is the 6-month LIBOR rate. These series are included for the same reasons as they were used in the Colombian equation. Both coefficients have the expected signs and are statistically significant. The absence of other balance of payment indicator series such as export prices may seem surprising. Because Ecuador, similar to Colombia, has a relatively diverse commodity export basket, no single price series will suffice (Fullerton, 1993a, 1993b). This is not the case for Venezuela, where petroleum products account for more than 80 percent of total merchandise exports (Fullerton, 1990c). Table 4.2: Three-Stage Least Squares Regression Results Colombia 1.142 (3.467) + 0.818*Pt.i (15.946) - 0.172*LIBOR6M (3.128) 0. 916*CPIt (1.735) R2 0.951 Q(6) 6.118 SER Q(12) 0.034 8.848 F-stat Prob(F-stat) 366.471 0.001 + 0.883*Ptoi (26.893) - 0.476*LIBOR6M (3.227) F-stat Prob(F-stat) 1462.68 0.001 Venezuela 0.972 (2.765) + 0.811*Pt.i (16.132) 0. 061*IRt (1.735) - 0.249*LIBOR6M (2.529) + 0.072*OILt (2.508) F-stat Prob(F-stat) 319.777 0.001 The sample period is March 1986 December 1991. 4.2 Pt = 4.3 Pt = Ecuador 1.352 (3.376) R2 Q(6) 0.981 5.632 SER Q(12) 4.4 Pt = 0.088 7.803 R2 Q (6) 0.959 3.717 SER Q(12) 0.057 5.671 73 Four independent variables are included in the Venezuelan model. As shown in Equation 4.4 in Table 4.2, they include a one-period lag of the debt price, the 180-day LIBOR series, international reserves net of gold, and the average price of petroleum exports. International reserves and the oil export price are employed as proxies for expected economic conditions in Venezuela. Although related, the two series do not always follow parallel paths. More specifically, the level of international reserves serves as an indicator of domestic economic policy success or failure (Fullerton, 1990c). World oil prices provide a measure of global demand for the nation's principal export. All of the regressor coefficients have the expected signs. The t-statistic for international reserves, however, is not significant at the 5-percent level. The 3SLS technique utilized to estimate the parameters reported in Table 4.2 allows for potential cross equation contemporaneous error correlation, a reasonable assumption for the secondary market for sovereign debt paper. To examine whether this assumption is necessary, residuals from the first stage ordinary least squares (OLS) regression are tested for contemporaneous correlation using a standard t-test methodology (see Ostle and Mensing, 1975, or Snedecor, 1956). As shown in Table 4.3, the OLS residuals for Colombia and Venezuela are correlated in a statistically significant manner, as are those for Ecuador and Venezuela. 74 Table 4.3: OLS Cross Equation Correlation Coefficients Equation Residuals Correlation Coefficient Computed t-stat Colombia-Ecuador 0.101 0.772 Colombia-Venezuela 0.340 2.750 Ecuador-Venezuela 0.453 3.872 75 Unit-root cointegration tests performed on the individual country 3SLS residuals are reported in Table 4.4. In all three cases, it appears cointegrating vectors have been obtained. Note that this procedure may be inappropriate for the short data set which currently exists for secondary market developing country debt prices. Previous research has shown that tests associated with unit-root techniques have low power when applied to short-run data sets (Hakkio and Rush, 1991). For purposes of the chapter at hand, the results in Table 4.4 are interpreted as evidence that the 3SLS regression results are not spurious. Given the diagnostic statistics also reported in Table 4.2, the latter conclusion is probably reasonable. To conduct ex ante dynamic simulation exercises, the equations were jointly reestimated and simulated for 24 different historical subperiods. Four month ahead forecasts were produced for each secondary market debt price series from the last subperiod observation forward. A four-month step length is sufficient to incorporate quarterly portfolio accounting considerations faced by international banks. As pointed out by participants at the 1990 Sovereign Debt Conference, quarterly corporate income tax filing requirements often trigger loan certificate swaps between secondary market participants and make a four-month time frame logical for simulation experiments. 76 Table 4.4: Unit Root Cointegration Tests Country Aug Dickey-Fuller t-stat MacKinnon crit value Colombia -4.895 (with const, trend, 1 lag) -4.122 (1% ivl) Ecuador -5.140 (with const, trend, 1 lag) Venezuela -5.320 (with const, trend, 1 lag) 77 To further enhance the realism of the extrapolation scenarios, actual forecast data available to the financial markets during each of the 24 subperiods are incorporated in the simulations. By relying on forecast data which are unconditional upon any information not available prior to the start of any simulation period, Klein's (1984) forecast evaluation criterion is met. These estimates for the independent regressors were compiled from international outlook reports published by The WEFA Group from 1989 to 1991. Forecast results for each price series are summarized in Table 4.5. For Colombia, the results indicate that the secondary market LDC debt price series is predictable using econometric methods. Although the 1-month ahead Colombian projections have a slightly greater than unity U-coefficient, the remaining estimates are all less than one. Interestingly, the Colombian inequality coefficients monotonically decrease as the length of the forecast period increases. This suggests that inclusion of econometric information, in this particular instance, grows in importance as the timeframe under consideration expands. In the case of Ecuador, the results reported here suggest that its secondary market debt discount is unpredictable. Consequently, it appears that financial analysts can do no better than utilize the last available observation for Ecuadorian paper in anticipating future quotes. Additional research with alternative estimators and different Table 4.5: Modified Theil Inequality Coefficients Series 1 Step 2 Steps 3 Steps 4 Steps Colombia 1.070 0.888 0.728 0.611 Ecuador 1.715 2.109 2.408 2.614 Venezuela 1.263 1.207 1.108 1.017 The simulation period is January 1990 December 1991. 79 specifications might obtain superior forecast accuracy, so it may be premature to conclude that the Ecuadorian debt price series is not predictable. As presented above, however, the results obtained in this chapter provide a fairly striking example of the fact that a relatively high coefficient of determination does not guarantee automatic simulation accuracy. Forecasting results for the Venezuelan debt price series are also interesting. Similar to Colombia, the inequality coefficients decline monotonically as the length of the projections increases. Unlike the Colombian example, however, the Venezuelan U-statistics remain at least slightly above unity at each step length. On the basis of the results in Table 4.5, it therefore appears that a random walk approach yields better forecasts. Experimentation with other estimators is probably warranted. Similar to Ecuador, however, the Venezuelan modeling and simulation results provide another example of a case in which a high coefficient of determination does not guarantee prediction accuracy. 4.4 Conclusion A simple questioned is asked in this chapter. Are secondary market debt prices predictable? To shed light on the possible answer, several steps are taken which have not previously been investigated. A key aspect that distinguishes this research from earlier efforts is the modeling of monthly 80 time series data for individual economies, as opposed to cross section annual or quarterly samples used elsewhere. Also developed are forecast exercises designed to meet the needs of participants in international financial markets where developing country debt instruments are traded. Modeling and simulation results reported here provide only limited evidence that it is possible to forecast secondary market developing country debt discounts from face value. All three series in the sample exhibited interesting model characteristics. In particular, the equations and forecasts for Colombia are encouraging. The same cannot be said of the Ecuadorian and Venezuelan prediction tests, as in both cases simple random walk forecasts of the respective debt price series prove more accurate. Because the three economies, and their respective debt price series, differ substantially from one another, it may be useful to apply this modeling approach to other debt price series. Obvious candidates include Argentina, Brazil, Chile, Mexico, and Peru. Further specification enhancements for Ecuador and Venezuela may also provide useful information. Inclusion of institutional variables related to Brady- initiative debt negotiations might prove beneficial, especially if it is possible to construct monthly indices for these factors. As noted by Friscia (1993), however, confidentiality restrictions may preclude this possibility. 81 Similarly, it may be worthwhile to test alternative estimators. The latter will become more feasible as additional observations and information regarding this market become available. On balance, however, it appears that movements in secondary market debt prices are not predictable. CHAPTER 5 SUGGESTIONS FOR FUTURE RESEARCH In Latin America, short-term econometric forecasting analysis is still a largely uncharted area of research. Material presented above indicates that a variety of techniques, methodologies, and modeling approaches may yield interesting insights with respect to both forecasting and policy issues. Not surprisingly, this research has only hinted at a few of the potentially beneficial topics which merit further attention. As discussed elsewhere, the availability of low cost computer hardware and software, plus the development of wide coverage high-frequency data banks, will help encourage additional research of this nature (Fullerton, 1992). With respect to the results presented above, new efforts are already underway in terms of alternative estimators for secondary market debt price forecasting. The underlying theoretical model presented in Chapter 3 has also been extended to include labor costs as part of the price vector on the output side for studying price dynamics. From a business forecasting perspective, a promising area of endeavor is likely to arise from the publication of quarterly national income and product account data. In 83 countries such as Ecuador, this will permit the development of large-scale macroeconometric models such as those pioneered by Barger and Klein (1954). The latter continue to enjoy a central role in business and government planning exercises throughout the world. Given the short-term uncertainties present in Latin America, quarterly modeling and forecasting will likely be welcomed with enthusiasm. At this juncture, much work remains to be done, but initial efforts such as those presented above point to numerous potential successes to be gained from future research efforts. REFERENCES Anayiotos, G., and J. de Pinies, 1990, "The Secondary Market and the International Debt Problem," World Development, 18, 1655-1669. Banco del Pacifico, 1994a, "Economic Report," 16-5 (September-October), 1-4. Banco del Pacifico, 1994b, "Economic Report," 16-6 (November-December), 1-4. Banco del Pacifico, 1995, "Economic Report," 17-2 (March- April), 1-4. Barger, H., and L.R. Klein, 1954, "A Quarterly Model for the United States Economy," Journal of the American Statistical Association, 49, 413-437. Batten, D.S., and M.T. Belongia, 1986, "Monetary Policy, Exchange Rates, and U.S. Agricultural Exports," American Journal of Agricultural Economics, 68, 422-427. Beltr~n del Rio, A., 1991, "Macroeconometric Model- Building of Latin American Countries, 1965-1985," Chapter 12 in A History of Macroeconometric Model-Building, Edited by R.G. Bodkin, L.R. Klein, and K. Marwah, Brookfield, VT: Edward Elgar Publishing Company. Bomberger, W.A., and G.E. Makinen, 1979, "Some Further Tests of the Harberger Inflation Model using Quarterly Data," Economic Development and Cultural Change, 27, 629-644. Box, G.E.P., and G.M. Jenkins, 1976, Time Series Analysis: Forecasting and Control, San Francisco: Holden-Day. Box, G.E.P., and G.C. Tiao, 1975, "Intervention Analysis with Applications to Economic and Environmental Problems," Journal of the American Statistical Association, 70, 70-79. Brajer, V., 1992, "Empirical Inflation Models in Developing Countries," Journal of Development Studies, 28, 717-729. Cabrera, M., and F. Montes, 1978, "La Dindmica de las Series de Tiempo: El Caso del Indice Nacional de Precios al Consumidor-Empleados," Revista del Banco do la Repfiblica, 51, 1122-1142. Chatfield, C., 1984, The Analysis of Time Series, New York: Chapman and Hall. Christ, C.F., 1993, "Assessing Applied Econometric Results," Federal Reserve Bank of St. Louis Review, 75 (March/April), 71-94. Christ, C.F., 1994, "The Cowles Commission's Contributions to Econometrics at Chicago, 1939-1955," Journal of Economic Literature, 32, 30-59. Clemen, R.T., 1989, "Combining Forecasts: A Review and Annotated Bibliography," International Journal of Forecasting, 5, 559-583. Clavijo, S., 1987, "Hacia una Caracterizaci6n del Comportamiento de la Velocidad de Circulaci6n del Dinero," Ensayos Sobre Politica Econ6mica, 12, 59-79. Dhane, G., and A.P. Barten, 1989, "When it all Began: The 1936 Tinbergen Model Revisited," Economic Modelling, 6, 203- 219. Direcci6n General de Estudios, various issues, Informaci6n Estadistica Mensual, Banco Central del Ecuador: Quito. Dutton, J., and T. Grennes, 1987, "Alternative Measures of Effective Exchange Rates for Agricultural Trade," European Review of Agricultural Economics, 14, 427-442. Eaton, J., and M. Gersovitz, 1981, "Debt with Potential Repudiation: Theoretical and Empirical Analysis," Review of Economic Studies, 48, 289-309. Edwards, S., 1985, "Money, the Rate of Devaluation, and Interest Rates in a Semiopen Economy: Colombia, 1968-82," Journal of Money, Credit, and Banking, 17, 59-68. Engsted, T., 1993, "Cointegration and Cagan's Model of Hyperinflation under Rational Expectations," Journal of Money, Credit and Banking, 25, 350-360. Feige, E.L., and D.K. Pearce, 1979, "The Casual Causal Relationship between Money and Income: Some Caveats for Time Series Analysis," Review of Economics and Statistics, 61, 521- 533. Friscia, A.B., 1993, "Comment on Fullerton," Atlantic Economic Journal, 21 (March), 87. Fullerton, T.M., Jr., 1989a, "A Composite Approach to Forecasting State Government Revenues," International Journal of Forecasting, 5, 373-380. Fullerton, T.M., Jr., 1989b, "Exchange Rate Indexes for Ecuador and Venezuela," Latin America Economic Outlook, The WEFA Group (June), 1.3-1.7. Fullerton, T.M., Jr., 1989c, "Ecuador," Latin America Economic Outlook, The WEFA Group (September), 6.1-6.15. Fullerton, T.M., Jr., 1990a, "Ecuador," Latin America Economic Outlook, The WEFA Group (June), 6.1-6.15. Fullerton, T.M., Jr., 1990b, "Colombia," Latin America Economic Outlook, The WEFA Group (March), 5.1-5.16. Fullerton, T.M., Jr., 1990c, "Venezuela," Latin America Economic Outlook, The WEFA Group (September), 10.1-10.14. Fullerton, T.M., Jr., 1990d, "Methodologies for Evaluating Secondary Market Debt Prices," paper presented at The WEFA Group Sovereign Debt Conference, New York, NY. Fullerton, T.M., Jr., 1991, "Colombia," Latin America Economic Outlook, The WEFA Group (March), 5.1-5.16. Fullerton, T.M., Jr., 1992, "Confessions of an International Forecaster," Business Economics, 27 (April), 61-65. Fullerton, T.M., Jr., 1993a, "Un Modelo Macroeconometrico para Pronosticar la Economia Colombiana," Ensayos Sobre Politica Econ6mica, 24, 101-136. Fullerton, T.M., Jr., 1993b, "Un Modelo Macroeconometrico para Pronosticar la Economia Ecuatoriana," Cuestiones Econ6micas, 20, 59-100. Fullerton, T.M., Jr., 1993c, "Inflationary Trends in Colombia," Journal of Policy Modeling, 15, 463-468. Fullerton, T.M., Jr., 1993d, "Predictability of Secondary-Market Developing-Country Debt Prices," Applied Economics, 25, 1365-1371. Fullerton, T.M., Jr., 1995a, "Short-Term Price Movements in Ecuador," Proceedings of the American Statistical Association, Business and Economic Statistics Section, 280- 285. Fullerton, T.M., Jr., 1995b, "Empirical Models for Secondary Market Debt Prices," paper presented at 65th Annual Meeting of the Southern Economic Association, New Orleans, LA. Fullerton, T.M., Jr., I. Fainboim, and M.I. Agudelo, 1992, "Exchange Rate, Money Supply, and Inflationary Dynamics in Colombia," Best Paper Proceedings of the Atlantic Economic Society, 2 (January), 54-58. Fullerton, T.M., Jr., R. Hirth, and M. Smith, 1991, "Inflationary Dynamics and the Angell-Johnson Proposals," Atlantic Economic Journal, 19 (March), 1-14. Fullerton, T.M., Jr., and A. Kapur, 1991, "Prediccion de Multiplicadores Monetarios en Colombia, Ecuador, y Venezuela," Lecturas de Economia, 35, 53-86. Gajdeczka, P., and M.R. Stone, 1990, "The Secondary Market for Developing Country Loans," Finance and Development (December), 22-25. Garlow, D.C., 1993, "Ecuador," Latin America Economic Outlook, The WEFA Group (September), 6.1-6.17. Gennotte, G., H.J. Kharas, and S.Sadeq, 1987, "A Valuation Model for Developing-Country Debt with Endogenous Rescheduling," World Bank Economic Review, 1, 237-271. Granger, C.W.J., 1992, "Forecasting Stock Market Prices," International Journal of Forecasting, 8, 3-13. Hakkio, C., and M. Rush, 1991, "Cointegration: How Short is the Long-Run?" Journal of International Money and Finance, 10, 571-581. Hallman, J.J., R.D. Porter, and D.H. Small, 1991, "Is the Price Level Tied to the M2 Monetary Aggregate in the Long Run?" American Economic Review, 81, 841-858. Hamilton, J.D., 1994, Time Series Analysis, Princeton, NJ: Princeton University Press. Hanson, J.A., 1985, "Inflation and Imported Input Prices in some Inflationary Latin American Economies," Journal of Development Economics, 18, 395-410. Harberger, A., 1963, "The Dynamics of Inflation in Chile," in Measurement in Economics: Studies in Mathematical Economics and Econometrics in Memory of Yehuda Grunfeld, edited by C.F. Christ, Stanford, CA: Stanford University Press. Ikhide, S., and T.M. Fullerton, Jr., 1995, "Inflationary Dynamics in Nigeria," paper presented at Institute for Developing Economies, Tokyo, Japan. Kamas, L., 1986, "Dutch Disease Economics and the Colombian Export Boom," World Development, 14, 1177-1198. Kamas, L., 1995, "Monetary Policy and Inflation under the Crawling Peg: Some Evidence from VARs for Colombia," Journal of Development Economics, 46, 145-161. Kercheval, M.P., 1987, "Another Look at the Dollar's Trade-Weighted Value," Business Economics, 22 (April), 43-47. Kim, C., and T.M. Fullerton, Jr., 1996, "Import Prices, Wage Rates, and Korean Price Dynamics," working paper, Department of Economics, Ajou University, Suwon, Korea. Klein, L.R., 1984, "The Importance of the Forecast," Journal of Forecasting, 3, 1-9. Koch, P., J. Rosensweig, and J. Whitt, 1988, "The Dynamic Relationship between the Dollar and U.S. Prices," Journal of International Money and Finance, 7, 181-204. Laidler, D., 1993, "Commentary on Assessing Applied Econometric Results," Federal Reserve Bank of St. Louis Review, 75 (March/April), 101-102. Laney, L.O., 1987, "The Secondary Market in Developing Country Debt," Federal Reserve Bank of Dallas Economic Review, (July), 1-12. Leiderman, L., 1984, "On the Monetary-Macro Dynamics of Colombia and Mexico", Journal of Development Economics, 14, 183-201. Leith, J.C., 1991, "The Exchange Rate and the Price Level in a Small Open Economy," Journal of Policy Modeling, 13, 309- 315. Montes, G., and R. Candelo, 1982, "El Enfoque Monetario de la Balanza de Pagos: El Caso de Colombia, 1968-1980," Revista do Planeaci6n y Desarrollo, 14, 11-40. Motley, B., 1993, "Inflation and Growth," FRBSF Weekly Letter, 93-44 (December 31), 1-3. Ocampo, J.A., 1983, "En Defensa de la Continuidad del Regimen Cambiario," Coyuntura Econ6mica, 13, 198-214. Ostle, B., and R.W. Mensing, 1975, Statistics in Research, Ames: Iowa State University Press. Pagan, A.R., 1974, "A Generalized Approach to the Treatment of Autocorrelation," Australian Economic Papers, 13, 267-280. Palma, P., and D. Fontiveros, 1988, "A Comparative Sensitivity Analysis of the MODVEN VII Macroeconomic Model for Venezuela," Economic Modelling, 5, 286-345. Pankratz, A., 1983, Forecasting with Univariate Box- Jenkins Models, New York: John Wiley and Sons. Perasso, G., 1989, "The Pricing of LDC Debt in the Secondary Market," Kyklos, 42, 533-555. Purcell, J.F.H., and D.J. Orlanski, 1988, "Developing Country Loans: A New Valuation Model for Secondary Market Trading," International Loan Trading Analysis, Salomon Brothers, 1-20. Purcell, J.F.H., and D.J. Orlanski, 1989, "Developing Country Loans: Refinements to the Relative Value Model for the Secondary Market," International Loan Trading Analysis, Salomon Brothers, 1-16. Rahnama-Moghadam, M., and Hedayeh Samavati, 1991, "Predicting Debt Rescheduling by Less-Developed Countries," Quarterly Review of Economics and Business, 31, 3-14. Rahnama-Moghadam, M., H. Samavati, and L.J. Haber, 1991, "The Determinants of Debt Rescheduling," Southern Economic Journal, 58, 510-517. Sachs, J., and H. Huizinga, 1987, "US Commercial Banks and the Developing-Country Debt Crisis," Brookings Papers on Economic Activity, 2, 555-601. Sheehey, E.J., 1976, "The Dynamics of Inflation in Latin America: Comment," American Economic Review, 66, 692-694. Sheehey, E.J., 1980, "Money, Income, and Prices in Latin America," Journal of Development Economics, 7, 345-357. Snedecor, G., 1956, Statistical Methods, Ames: Iowa State University Press. Snowden, P.N., 1989, "The Interpretation of Market Discounts in Assessing Buy-Back Benefits," IMF Staff Papers, 36, 733-735. Stone, M.R., 1991, "Are Sovereign Debt Secondary Market Returns Sensitive to Macroeconomic Fundamentals?," Journal of International Money and Finance, 10, SI00-S122. Vial, J., 1988, "Comparaci6n de Modelos Macroeconometricos Latinoamericanos," Colecci6n Estudios CIEPLAN, 24, 145-181. Vogel, R.C., 1974, "The Dynamics of Inflation in Latin America, 1950-1969," American Economic Review, 64, 102-114. Wakeman-Linn, J., 1991, "The Market for Developing Country Debt," The World Bank Research Observer, 6, 191-203. West, C.T., 1996, "System-based Weights vs. Series- specific Weights in the Combination of Forecasts," paper presented at the 42nd Regional Science Association International North American Meetings, Cincinnati, OH. Zellner, A., 1994, "Time Series Analysis, Forecasting, and Econometric Modeling: The Structural Econometric Modeling, Time Series Analysis (SEMSTA) Approach," Journal of Forecasting, 13, 215-233. Zellner, A., and H. Theil, 1962, "Three-Stage Least Squares: Simultaneous Estimation of Simultaneous Relations," Econometrica, 30, 54-78. Zind, R.G., 1993, "On Inflation and Growth in the LDCs," Economic Studies Quarterly, 44, 108-116. BIOGRAPHICAL SKETCH Thomas M. Fullerton, Jr. is a senior economist in the Forecasting Program of the Bureau of Economic and Business Research at the University of Florida. Fullerton is co-author of The Florida Outlook, a quarterly forecast of the state and 20 metropolitan economies. He also teaches a course on Latin American political economy. Fullerton previously worked at Wharton Econometrics as international economist in charge of modeling, forecasting, and policy analysis for Colombia, Ecuador, and Venezuela. He also worked as an economist in the Executive Office of the Governor of Idaho where he forecast the state economy and conducted fiscal policy analysis during legislative sessions. He began his career in the Planning Department of El Paso Electric Company. His research has been published in outlets such as Applied Economics, Journal of Forecasting, Public Budgeting & Finance, Atlantic Economic Journal, Journal of Policy Modeling, Business Economics, Applied Economics Letters, and International Journal of Forecasting. Fullerton holds degrees from the University of Texas at El Paso, Iowa State University, and the Wharton School of the University of Pennsylvania. He is a doctoral candidate in economics at the University of Florida. I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Carol Taylor West, Chair Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David A. Denslow Distinguished Service Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Chunrong Ai Assistant Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Terry '. McCoy Professor of Latin American Studies This dissertation was submitted to the Graduate Faculty of the Department of Economics in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August 1996 Karen A. Holbrook Dean, Graduate School |

Full Text |

SHORT-TERM ECONOMETRIC FORECASTING ANALYSIS
FOR LATIN AMERICA BY THOMAS M. FULLERTON, JR. A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1996 UNIVERSITY OF FLORIDA LIBRARES ACKNOWLEDGEMENTS Professor Carol Taylor West is probably the best known and most widely published regional economic forecaster in the world. I was acquainted with her research and professional reputation long before applying for the senior economist position on her staff at the Bureau of Economic and Business Research. Over the past five years, we have faced computer crashes, parameter instability, hurricane shocks, business cycle inflection points, submission deadlines, bureaucratic regulations, jet lag, and mind boggling statistical revisions. Forecasting Program work was always accomplished in a responsible manner befitting the tradition of econometric excellence with which the University of Florida is so well known. It has been a tremendous experience to forecast the Florida economy and conduct research with Carol. David Denslow has been a friend, colleague, and coauthor. We have spent many hours discussing different policy issues and business outlooks that one or the other was called upon to discuss in a public forum. Similarly, Terry McCoy and I have participated in a number of conferences on topics relating to policy problems facing Latin America. Terry definitely has keen insights with respect to the entire region. ii Chunrong Ai helped clarify many of the mysteries of modern econometric theory. He also saved me a lot of time and effort by indicating which estimation strategies were viable with respect to Chapter 4 and subsequent research developed with that data set. Bill Bomberger spent generous amounts of time going over Chapter 3 and making helpful suggestions. Prior to his retirement, Henri Theil was on my dissertation committee. Subsequent to his retirement, Professor Theil has dropped by my office at least once a week to monitor my progress and talk about the post-war history of econometrics. Many other individuals in the College of Business Administration helped me navigate exams, computer systems, and Graduate School requirements. In no particular order, they include Stan Smith, Larry Kenny, Janet Galvez, Tony Tracy, Min Zhu, Yikang Li, Jennifer Cobb, Jon Hamilton, Richard Romano, Janet Fletcher, Pam Middleton, Tony Daniele, Janet Rose, Cindy Houser, Dot Evans, Dian Studstill, David Lenze, and Richard Lutz. My parents, relatives, and friends provided welcome relief on numerous occasions. It takes a lot of friendly advice to get through a doctoral program. iii TABLE OF CONTENTS Page ACKNOWLEDGEMENTS ii ABSTRACT V CHAPTERS 1 ECONOMETRIC FORECASTING IN LATIN AMERICA 1 2 INFLATIONARY TRENDS IN COLOMBIA 6 2.1 Introduction 6 2.2 Previous Research 7 2.3 Methodology 10 2.4 Estimation Results 13 2.5 Policy Simulation Results 21 2.6 Conclusion 27 3 SHORT-TERM PRICE MOVEMENTS IN ECUADOR 30 3.1 Introduction 30 3.2 Literature Review 32 3.3 Theoretical Model 35 3.4 Estimation Results 44 3.5 Policy Simulation Results 51 3.6 Conclusion 55 4 PREDICTABILITY OF SECONDARY MARKET DEBT PRICES. 57 4.1 Introduction 57 4.2 Earlier Studies 59 4.3 Empirical Analysis 63 4.4 Conclusion 79 5 SUGGESTIONS FOR FUTURE RESEARCH 82 REFERENCES 84 BIOGRAPHICAL SKETCH 91 iv Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SHORT-TERM ECONOMETRIC FORECASTING ANALYSIS FOR LATIN AMERICA By Thomas M. Fullerton, Jr. August 1996 Chair: Professor Carol Taylor West Major Department: Economics Given the short-term nature, one year or less, of many policy goals espoused by decision makers in Latin America, the objective of this study is to examine the applicability of different forecasting techniques to a subset of short-run economic questions. Because quarterly national income and product series are typically not available, most commercial econometric forecasting analysis for the region is conducted utilizing annual data. Monthly time series do, however, exist for a large number of key macroeconomic and financial variables. It is from the latter group of publicly available data sets that the econometric modeling and simulation exercises are drawn. Chapter 2 examines potential impacts associated with Colombian monetary authority efforts to cut the rate of inflation by 10 over a 12-month period. Principal tools v relied upon include slower rates of nominal exchange rate devaluation and money supply growth. Empirical analysis in Chapter 2 is carried out utilizing transfer function autoregressive moving average modeling. Model simulations indicate that adherence to such a program can lead to noticeable disinflation over a 24-month period. Similar to other Latin American economies, Ecuador has faced persistently high rates of inflation in recent years. In May 1994, the government signed a stand-by loan agreement with the International Monetary Fund that established a goal of reducing the inflation rate to 15 percent over a 19-month period. Chapter 3 develops an exchange rate augmented monetary model to assess viability of the price stabilization program. In contrast to the time series approach of Chapter 2, short-run inflationary dynamics are modeled using an econometric framework. Economic debates regarding Latin America in recent years have been dominated by the debt crisis. In response to debtor country defaults, many lenders reduced or reshuffled risk exposures by selling debt certificates at discounts from face value. Chapter 4 analyzes the predictability of secondary- market debt prices for Colombia, Ecuador, and Venezuela. Estimation is accomplished via generalized least sguares over 24 separate historical periods utilizing monthly data. Model simulations indicate that forecasting these prices is a difficult task. vi CHAPTER 1 ECONOMETRIC FORECASTING IN LATIN AMERICA Econometric forecasting analysis began developing as a field of research with the initial endeavors of Jan Tinbergen and others in the 1930s (Dhane and Barten, 1989). These relatively small macroeconometric models of the Dutch economy were developed using annual data. Research on business cycles in the United States began replicating and extending the Dutch modeling examples in the 1940s. Increasing interest in shortÂ¬ term economic fluctuations eventually led to the development of quarterly forecasting models (Barger and Klein, 1954). Along with the expansion of structural econometric modeling and research with respect to forecasting and policy analysis, time series statistics also became increasingly sophisticated in the realm of predictive modeling. Most of these efforts occurred with respect to high frequency monthly data, especially univariate autoregressive moving average models (Box and Jenkins, 1976). While often regarded as competitors, time series models are frequently utilized as complements to econometric equation sets and can also be imbedded within a variety of forecasting systems (Clemen, 1989; Fullerton, 1989a; Zellner, 1994; West, 1996). 1 2 Latin American macroeconomic forecasting models began to appear in the 1960s (BeltrÃ¡n del Rio, 1991). Similar to the first models for the Netherlands and the United States, the early Latin American models utilized systems of simultaneous equations designed around national income and product account (NIPA) data. Unlike most industrialized economies, however, NIPA data in Latin America during this period tended to be published only at an annual frequency. This constraint precluded the development of Latin American quarterly forecasting models in a manner analogous to what occurred in many industrialized economies. Although quarterly forecasting models have not been widely disseminated in Latin America, large-scale forecasting models built using annual NIPA data abound. Representative examples include the CIEMEX-WEFA model for Mexico (BeltrÃ¡n del Rio, 1991), the CIEPLAN model for Chile (Vial, 1988), the MetroeconÃ³mica model for Venezuela (Palma and Fontiveros, 1988), and the WEFA models for Colombia and Ecuador (Fullerton, 1993a, 1993b). Among the most distinguishing characteristics shared by these models are continuous maintenance and enhancement over sustained periods of time. All of the aforementioned studies provide detailed references to the history of macroeconometric modeling in the region. Econometric forecasting analysis using annual data in Latin America has a fairly distinguished history and track record that is well-documented. 3 Given the volatile behaviors of the majority of the economies in Latin America in the 1980s, the absence of quarterly forecasting tools hampered business planning efforts. Forecasting conferences sponsored by commercial entities such as Wharton Econometrics during the late 1980s utilized simulation output from annual structural models for Latin American countries of interest. Client questions at these meetings were generally directed toward the first year of the forecast period, largely ignoring outer period extrapolation results (for example, see Fullerton, 1991). This is not to imply that the traditional macroeconometric Latin American models are regarded as useless. Most analyses of international indebtedness typically rely on annual data in order to examine the consequences of changes in regional balance of payment factors. Not surprisingly, significant effort was put forth in recent years to enhance the current account-capital account linkages and simulation performances in Latin American macroeconomic models (Fullerton, 1993a). The relative lack of forecasting models estimated with higher frequency data, nevertheless, continues to pose an obstacle to corporate, public sector, and multilateral agency planners and economists. Although quarterly national income and product account data are still not widely available to researchers in Latin America, monthly time series for many economic variables do 4 exist. Examples of the latter include inflation indices, money supply measures, currency exchange rates, commodity export prices, and international reserves. A logical step, therefore, is to investigate the econometric properties and predictability of the series presently available at the higher frequency. The centerpiece chapters of this dissertation examine three empirical forecasting questions using monthly data sets from Colombia, Ecuador, and Venezuela. Distinct estimation techniques are employed in each chapter and simulation accuracy is summarized for all of the models presented therein. Chapter 2 utilizes a transfer function autoregressive integrated moving average (transfer ARIMA) modeling framework to analyze movements in consumer prices in Colombia. Forecasting experiments are conducted with the resulting model to shed light on potential impacts associated with an anti- inflationary program enacted by the central bank. Results indicate that this class of time series modeling can provide useful insights with respect to macroeconomic trends in Latin American countries. A revised version of this chapter was published in the Journal of Policy Modeling (Fullerton, 1993c). Chapter 3 also examines the question of forecasting short-term price movements. Consumer prices in Ecuador are modeled using an econometric framework that incorporates domestic macroeconomic factors and international input costs. 5 Parameter estimation is accomplished with a nonlinear procedure that provides sufficient flexibility to handle even mixed error structures. Simulation exercises are also utilized to examine possible consequences associated with a price stabilization program implemented by the central bank. An abridged version of this study was published in Proceedings of the American Statistical Association, Business and Economic Statistics Section (Fullerton, 1995a). Secondary market trades involving sovereign debt certificates became widespread following the outbreak of the international debt crisis in 1982. Chapter 4 employs a generalized least squares modeling strategy to study the predictability of short-run movements in secondary market debt prices in Colombia, Ecuador, and Venezuela. As discussed below, forecast period lengths and information inputs are selected to reflect considerations facing financial market participants. A revised version of this chapter appeared in Applied Economics (Fullerton, 1993d). Additional research extending the initial results on this topic was presented at the 65th annual Southern Economic Association conference (Fullerton, 1995b). CHAPTER 2 INFLATIONARY TRENDS IN COLOMBIA 2.1 Introduction Inflation has long been one of the most serious economic problems facing policymakers in Latin America. Although Colombia has traditionally enjoyed lower rates of inflation than neighboring South American countries, in 1990, a presidential election year, consumer prices rose in excess of 32 percent. In response, Colombian monetary authorities introduced a series of policy innovations designed to lower the inflation rate. Measures adopted include a progressive opening of the economy to greater volumes of international trade, fiscal austerity, tighter credit controls, and a slower rate of currency devaluation (for details, see Fullerton, 1991). This chapter examines some of the potential results associated with the principal tools of the policy adjustment effort. Time series characteristics of the consumer price index (CPI), the money supply (Ml, defined as currency in circulation plus demand deposits), and the peso/dollar exchange rate (REX) are investigated. To measure the short- run relationships among the CPI, Ml, and REX, transfer 6 7 function autoregressive-moving average (ARIMA) analysis is applied (see Box and Jenkins, 1976, ch. 11). Selection of the nonlinear model estimation methodology was motivated by two factors. First was the usefulness of transfer function ARIMA analysis for short-term forecasting applications. Second was interest expressed in a previous study which utilized this technique to analyze price dynamics in the United States (Fullerton, Hirth, and Smith, 1991). Specifically, economists within the Research Department of the Central Bank in BogotÃ¡ wished to see whether the same approach would prove useful with respect to the Colombian economy. Subsequent sections of the paper present a brief review of the literature, methodology, and empirical results. Policy simulation exercises designed to reflect exchange rate and monetary policies in Colombia follow. A summary and conclusion form the final section of the chapter. 2.2 Previous Research In one of the earliest time series studies developed for an economy in Latin America, Cabrera and Montes (1978) utilize univariate ARIMA techniques to model the CPI in Colombia. Logarithmic transformation, regular and seasonal differencing of the monthly CPI series are used to induce stationarity. An equation containing an autoregressive term at lag 1 and a seasonal moving average term at lag 12 yields statistically significant parameter estimates. While simple in structure, 8 the model exhibits good statistical traits and is found to simulate historical movements of the CPI successfully. Empirical evidence is provided that Colombian inflation, although high relative to many industrial economies, is stable enough to be modeled and predicted with a fair degree of accuracy. Montes and CÃ¡ndelo (1982) propose a simultaneous system of eguations for money, prices, international reserves, and the exchange rate. Full information maximum likelihood estimation is used to calculate model parameters which reflect the hypothesized coefficient restrictions. Although quarterly data are used, the lag structure of the model is fairly simple. Domestic monetary conditions and the rate of devaluation are both found to positively affect consumer prices in a statistically significant manner. The magnitudes of the exchange rate coefficients exceed those of the monetary variables in each of the different sample periods. Leiderman (1984) utilizes vector autoregressions to analyze inflation in Colombia and Mexico. Changes in the rates of growth in the economy and the money supply are included in each model. In the case of Colombia, the results indicate that variations in the rate of change of Ml systematically affect the CPI, but not the converse. From a policy perspective, this implies that monetary authorities do not engage in "accommodative" measures in response to production and inflationary shocks. This result may stem from 9 the usage of monetary policy to achieve goals other than price stability. These include economic growth and export diversification. From an econometric standpoint, this result is also important because it implies that unidirectional Granger causality exists between Ml and the CPI in Colombia. A number of recent studies have examined inflationary dynamics in the United States. Koch, Rosensweig, and Whitt (1988) investigate the relationship between the exchange rate and consumer prices. Cross correlation functions are used to suggest the number of lags to be included in the regression equations. Granger causality tests imply a unidirectional channel of influence from the exchange rate to prices. The inflationary impacts associated with a change in the international value of the dollar are found to extend more than 12 months. Fullerton, Hirth, and Smith (1991) consider the effects of exchange rate and other financial and commodity price variable movements on the CPI in the United States. Transfer function ARIMA analysis indicates that inflationary impacts resulting from variations in the exchange rate generally take more than a year. Credit conditions, as proxied by interest rate spreads, are found to influence consumer prices relatively quickly. The slow speed of adjustment from the exchange rate compared to domestic financial conditions may reflect incomplete pass-through effects which characterize large industrial economies. 10 This is in contrast to what might be expected for a smaller economy such as Colombia's, where pass-through effects, statistically significant relationships between international currency values and inflation, are often strong and relatively guick (see Leith, 1991). Empirical evidence reported elsewhere (Edwards, 1985) indicates that movements in the rate of devaluation are helpful in modeling nominal interest rates in Colombia. Unfortunately, the latter study does not directly test the relationship between prices and the peso/dollar exchange rate. 2.3 Methodology The methodology utilized in this paper is similar to the multiple ARIMA approach applied to the United States by Fullerton, Hirth, and Smith (1991). This technique does not apply any a priori modeling restrictions on the equations estimated. It is a useful procedure for investigating high frequency time series data because it can accommodate different lag structures in a flexible and efficient manner. Equations developed using this approach are also easy to simulate and can help analyze policy proposals. Univariate ARIMA equations are estimated for the stationary components of the CPI, Ml, and REX series. The results are then used to specify and estimate an ARMA transfer function. A weakly stationary series is defined as one whose mean and variance do not change over time. Stationarity in 11 the means of the series is attained through regular and seasonal differencing. Logarithmic transformations are used to induce homoscedasticity (see Pankratz, 1983). The general form of a univariate ARIMA model estimated for the CPI can be written as follows: (2.1) Pt = [Q(B)Qs(B)Ut] / [H (B) Hs (B) ] , where Pt is a stationary series derived from the original CPI series, B is a backshift or lag operator, Q(B) is a moving average polynomial of order g, QS(B) is a seasonal moving average polynomial of order gs, Ut is the disturbance term, H(B) is an autoregressive polynomial of order p, and HS(B) is a seasonal autoregressive polynomial of order ps. The ARMA model for the CPI is also used in the estimation of the transfer function eguations. To examine the effects of other variables on the CPI, transfer function ARIMA models are estimated. These models are used to determine if the input series are incrementally useful in explaining the variation of the CPI beyond the information obtained within the price index itself (see Box and Jenkins, 1976, ch. 11). Prior to estimating a transfer function eguation, cross correlation functions (CCF) are used to indicate which lags of an input series may contribute incremental information to the univariate ARMA model already estimated for the CPI. Residuals from the three univariate 12 equations are used to calculate CCFs between the CPI and the other variables (see Chatfield, 1984, p. 173). Statistical, or Granger, causality is assumed to be unidirectional in transfer ARIMA models. If this assumption is reasonable, movements in REX and Ml will chronologically precede statistically significant changes in CPI. The converse will not hold if causality is unidirectional. To examine the possibility that feedback or endogeneity exists between the CPI and the other series, ordinary least squares regression is used to construct Granger causality F-tests. The general form of the transfer function can be written in the following manner: (2.2) Pt = [W(B)Rt + G(B)Mt + Q(B)Qs(B)Ut] / [H(B)HS(B)] , where the univariate model presented in Equation 2.1 is augmented by incorporating general order polynomials, W(B) and G(B), for the respective input variables. Rt and Mt represent stationary input series derived from the exchange rate and the money supply data discussed above. Because the analysis is conducted within a dynamic framework, coefficient restrictions are not hypothesized, but the sums of the coefficients associated with each individual input series are expected to be positive. 13 2.4 Estimation Results Data used in this chapter are from the central bank in BogotÃ¡. Monthly observations for all three series are published in Revista del Banco de la RepÃºblica (for example, see Cabrera and Montes, 1978). The sample period, January 1967 through December 1990, corresponds to the crawling peg era of exchange rate determination in Colombia. Similar to Cabrera and Montes (1978), a logarithmic transformation of the CPI is taken prior to regular and seasonal differencing to obtain stationarity. The same steps are taken with respect to REX and Ml prior to modeling. Results of the stationarity tests for all three series appear in Table 2.1. The unit root tests are performed with both constant and trend terms. Autocorrelation and partial autocorrelation functions suggested the forms of the univariate ARIMA eguations reported in Table 2.2. Despite the presence of 11 years of additional data, the results confirm the AR(1), SMA(12) specification employed by Cabrera and Montes (1978, p. 1137) to model the CPI. Parameter estimates reported in Model 2.3 carry the same signs and are similar in magnitude to those estimated in the previous study (0.450 versus 0.645, and -0.924 versus -0.858). Eguation 2.3 is incorporated in the estimation of the transfer function equations. 14 Table 2.1: Unit Root Tests for Stationarity Series Aug Dickey-Fuller t-stat MacKinnon crit value P - 9.081 (with const, trend, 1 lag) -4.001 (1% lvl) R - 5.252 (with const, trend, 1 lag) M -13.195 (with const, trend, 1 lag) 15 Table 2.2: Univariate ARIMA Models Model Parameters 2.3 CPI pt = 0.001 + 0.450*Pt,1 0.924 *Ut_12 (10.186) (11.397) (15.900) Q(38) = 52.222 2.4 REX Rt = 0.001 + 0.795*Rt,1 0.784*Ut_12 (8.827) (27.671) (13.148) Q(38) = 31.893 2.5 Ml Mt = 0.001 0.229*Ut-1 (4.465) 0.634*Ut,12 (7.715) (10.558) Q(38) = 59.339 The sample period is January 1967 - December 1990. Numbers in parentheses are computed t-statistics. Ljung-Box Q-statistics calculated for 38 lags are reported. 16 Ordinary least squares F-tests are used to determine if unidirectional Granger causality exists between consumer prices and the two input series. Results for the F-tests, calculated for 12 and 18 lags, are reported in Table 2.3. The tests constructed to examine the relationship between the CPI and Ml utilize seven years of additional data, but support the conclusions reported by Leiderman (1984) . Because monetary policy in Colombia appears to be conducted independently of inflationary shocks, the transfer function methodology can be used to measure the effect of the money supply on inflation. Similar to results reported for the United States (Koch, Rosensweig, and Whitt, 1988), the causality tests for the CPI and REX series indicate that changes in the inflation rate do not precede systematic variations in the exchange rate. From an historical perspective, the result is not surprising. There have been several episodes during the crawling peg era in Colombia when authorities have permitted the exchange rate to become overvalued by failing to devalue the local currency quickly enough to account for inflationary differentials with major trading partners. Typically, this has tended to take place following "coffee bonanzas" when Colombian international reserves are high due to strongly positive merchandise trade surpluses (Kamas, 1986; Ocampo, 1983). 17 Table 2.3: Granger Causality Tests Causality Direction Number of Lags Computed F-stat CPI => REX 12 0.834 CPI => REX 18 0.799 CPI => Ml 12 0.009 CPI => Ml 18 0.278 Degrees of freedom for the regressions with 12 lags: 12 for the numerator and 263 for the denominator. Degrees of freedom for the regressions with 18 lags: 18 for the numerator and 251 for the denominator. 18 There have also been periods when local price changes have moved in concert with those of Colombia's trading partners, rendering unnecessary any modification in the crawling peg. As a result, the rate of devaluation has not always been adjusted proportionately to variations in the domestic rate of inflation. From an economic policy perspective, the results in Table 2.3 indicate that Colombian monetary authorities take into account goals and variables other than domestic inflation when determining rate of devaluation embodied in the crawling peg for the peso. Econometrically, this implies that transfer function ARIMA analysis can be used to model the influence of the international value of the peso on domestic prices. Similar to the CPI, stationarity in the REX and Ml series is attained after logarithmic transformation, and seasonal and regular differencing. Unit root tests for both variables are reported in Table 2.1. As noted above, univariate equations for the exchange rate and money supply series appear in Table 2.2. These equations also exhibit good statistical traits such as high computed t-statistics and relativley low Q- statistics. Residuals from the three univariate models were used to estimate CCFs containing 18 lags. Both CCFs indicated that the principal effects resulting from a change in either input variable are incorporated in the CPI within a relatively short period. 19 Transfer function equations are reported in Table 2.4. Models 2.6 and 2.7 include only one independent variable, REX and Ml, respectively. Model 2.8 includes both input series. The exchange rate is included with lags of 2 and 10 months. The money supply is included with a 9-month lag. Autoregressive terms at lag 1 and seasonal moving average terms at lag 12 are included in all three equations. Similar to Montes and CÃ¡ndelo (1982), the exchange rate input coefficients are larger than that of Ml. Inclusion of the input series improves the Q-statistic estimated from the residuals associated with each equation. Virtually all of the improvement in the white noise test results from the introduction of the lagged stationary exchange rate data. Coefficients estimated for these series are statistically significant in Model 2.6 and Model 2.8. The lagged stationary component of the money supply exhibits the hypothesized sign, but is not significant at the 5-percent level. Because money is a useful predictor of inflation in Colombian macroeconometric models using annual data (Fullerton, 1993a), the results encountered in this chapter are unexpected and further research is warranted. Another possibility is that Ml may not be the correct money stock measure with respect to Colombian price dynamics. Studies completed for other economies indicate that broader liquidity aggregates such as M2 may be useful (Hallman, Porter, and 20 Table 2.4: ARIMA Transfer Functions Model Parameters 2.6 CPI Pt = 0.018 + 0.459*Pt,1 (8.423) (11.623) + 0.201*Rt_2 (2.215) Q (38) = 39.826 2.7 CPI Pt = 0.018 + 0.447*Pt,1 (9.476) (11.334) + 0.030*Mt,9 (1.638) Q(38) = 51.777 2.8 CPI Pt = 0.018 (8.113) + 0.457*Pt.1 (11.570) + 0.216*Rt_2 (2.384) + 0.034*Mt,9 (1.839) Q(38) = 39.389 0.890*Ut,12 (15.190) O.O76*Rt,10 (4.088) 0.929*Ut,12 (15.974) 0.895*Ut,12 (15.235) O.O77*Rt,10 (4.133) The sample period is January 1967 - December 1990. Numbers in parentheses are computed t-statistics. Ljung-Box Q-statistics calculated for 38 lags are reported. 21 Small, 1991, and Ikhide and Fullerton, 1995). Unfortunately, money supply estimates other than Ml currently do not exist for Colombia. This problem is fairly widespread in South America and has long posed difficulties in the analysis of monetary economics in the region (see Fullerton and Kapur, 1991). 2.5 Policy Simulation Results The Gaviria administration announced in December 1990 that it would attempt to lower the inflation rate to 22 percent in only twelve months (for discussion, see Fullerton, 1991). To attain this goal, two principal tools were to be employed. After an extended period of aggressive devaluation, the nominal rate of devaluation for the peso/dollar exchange rate was to be reduced. Policymakers also announced that growth in the money supply would be limited to 22 percent. To examine the potential effects of these policy innovations on the CPI, simulations are created using Equation 2.8. In the first exercise, the 12-month growth rate for Ml is assumed to be immediately limited to 22 percent for an initial 12-month period, and then lowered to 19 percent the following year. Similarly, the 12-month crawling peg rate of devaluation is assumed to be held at 22 percent throughout the first year, and later be revised downward to 19 percent during the following year. The 19 percent rates of change are chosen to reflect longer-term policy goals discussed by the 22 government, including eventual attainment of a 15 percent annual rate of change for consumer prices (Fullerton, 1991). Simulating Model 2.8 under these policy assumptions yields several interesting results that are reproduced in Table 2.5. During the first six months, the inflation rate oscillates near 29 percent. Subsequently, the 12-month rate of change in the CPI declines sharply to 24 percent by year- end. During the next 12 months, disinflation subsides as the severity of the cutbacks in the rates of change of both input variables is moderated. Year-end inflation declines to 22 percent under this scenario. As will be illustrated below, not all of the 10-point improvement can be attributed to the government policy package. Not surprisingly, Colombian monetary authorities did not introduce the exchange rate and credit policy changes in the abrupt manner depicted above. Accordingly, Equation 2.8 is also simulated under an alternative set of assumptions whereby intermediate policy steps are implemented more gradually. Additionally, actual exchange rate and money supply data for the first six months of 1991 are employed. These data reflect slower attainment of the new policy goals espoused by the Finance Ministry and the Central Bank. Subsequent movements in the 12-month rates of change for the exchange rate and the money supply are assumed to steadily decline to 19 percent by December 1992 for both variables. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 23 Immediate Implementation Policy Simulation Results PCHYA(CPI) PCHYA(REX) PCHYA(Ml) 32.1 22.0 22.0 30.7 22.0 22.0 28.9 22.0 22.0 29.1 22.0 22.0 29.6 22.0 22.0 29.2 22.0 22.0 28.9 22.0 22.0 27.9 22.0 22.0 26.8 22.0 22.0 26.2 22.0 22.0 25.2 22.0 22.0 24.1 22.0 22.0 23.5 19.0 19.0 23.3 19.0 19.0 22.5 19.0 19.0 22.5 19.0 19.0 22.4 19.0 19.0 22.4 19.0 19.0 22.4 19.0 19.0 22.4 19.0 19.0 22.4 19.0 19.0 22.3 19.0 19.0 22.1 19.0 19.0 22.1 19.0 19.0 24 Results of the second simulation exercise are reported in Table 2.6. The more rapid rate of depreciation allows the 12- month inflation rates to remain above 30 percent throughout the first semester of the test period. Steady declines in the rates of change for both REX and Ml force inflation to decline noticeably during the next six months, eventually reaching 25 percent. Year-end inflation slows to 22 percent during the subsequent 12-months of the simulation period, when growth rates for the input variables gradually decline to 19 percent. It is useful to note that the second set of medium-range simulation results reported in Table 2.6 do not vary significantly from those obtained in the first exercise. Despite imperfect implementation efforts, these results indicate that the government can still attain its stated policy goals. Because a 24-month period is still needed to lower the rate of change in the CPI by 10 percentage points, the government policy claims are again shown to be slightly optimistic. Furthermore, as shown in the final simulation exercise presented in Table 2.7, much of the improvement in the inflation rate cannot be attributed to policy design alone. Of course, policy indecision can also result in no progress being made toward achieving either intermediate target. To examine the potential consequences associated with such an eventuality, model simulations were also tested in 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Gradual Implementation Policy Simulation Results PCHYA(CPI) PCHYA(REX) PCHYA(Ml) 32.1 29.9 28.2 30.7 28.9 22.3 30.7 27.8 25.0 30.6 26.8 22.6 30.8 25.8 30.3 30.3 23.6 26.2 29.8 23.0 26.0 28.3 22.5 25.5 27.0 22.0 25.0 26.5 21.5 24.5 25.8 21.0 24.0 24.6 20.5 23.5 23.8 20.0 23.0 23.6 19.9 22.5 23.2 19.8 22.0 22.9 19.7 21.5 22.8 19.6 21.0 22.7 19.5 20.5 22.7 19.4 20.0 22.6 19.3 19.8 22.5 19.2 19.6 22.4 19.1 19.4 22.3 19.0 19.2 22.3 18.9 19.0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 26 No Implementation Policy Simulation Results PCHYA(CPI) PCHYA(REX) PCHYA(Ml) 32.2 30.0 30.0 31.9 30.0 30.0 31.6 30.0 30.0 31.2 30.0 30.0 30.9 30.0 30.0 30.6 30.0 30.0 30.3 30.0 30.0 29.9 30.0 30.0 29.6 30.0 30.0 29.4 30.0 30.0 29.2 30.0 30.0 28.8 30.0 30.0 28.5 30.0 30.0 28.3 30.0 30.0 28.0 30.0 30.0 27.7 30.0 30.0 27.4 30.0 30.0 27.1 30.0 30.0 26.8 30.0 30.0 26.6 30.0 30.0 26.4 30.0 30.0 26.2 30.0 30.0 26.3 30.0 30.0 26.2 30.0 30.0 27 which the growth rates of the input series were held constant at 30 percent per year. In the absence of any change in the conduct of monetary management and depreciation rate determination, the annual rate of inflation stabilizes at 26 percent by the end of the 24-month simulation period. That rate is well above the announced government target range. More importantly, the 6-point improvement which results in a simulation exercise in which no progress is made with respect to the intermediate policy targets. This result indicates that only 40 percent of policy attainment embodied in Tables 2.5 and 2.6 is by government design. The remaining six-tenths of the 10-point reduction in the annual rate of change in consumer prices would, on the basis of the above modeling and simulation framework, have resulted anyway. 2.6 Conclusion In response to growing inflationary pressures, economic policymakers in Colombia announced that the inflation rate would be slashed by 10 percentage points to 22 percent over a 12-month period. Two principal tools were selected to foster disinflation. The nominal rate of devaluation for the peso/dollar exchange rate was cut and the rate of growth of the money stock was reduced. This chapter examines the empirical relationship between those variables and the consumer price index. Econometric results are similar to those reported in previous studies for 28 Colombia and the United States. Transfer ARIMA functions are estimated and simulated to determine the potential impacts of the new policies. These exercises indicate that substantial progress in the anti-inflationary program may be attained following the implementation of said policy efforts, although not as quickly as stated by government officials. More importantly, over half of the 10-point gain results even if the rates of nominal currency devaluation and money supply expansion are held constant. Because of structural economic and administrative policy changes taking place in Colombia, additional research will eventually be necessary. Future studies may find it useful to consider the effects of other variables such as wage rates, industrial capacity utilization, and commodity prices on the CPI. Given the insignificant, at the 5-percent level, coefficient associated with Ml, model estimation utilizing alternative series designed to reflect monetary conditions may also prove helpful. Introduction of new variables such as wage rates may necessitate usage of an estimation technique different from the transfer function methodology described above. This is due to the possibility that feedback effects, or simultaneity, may exist between the CPI and other potential input series. At present, the Colombian economy is being opened to international trade and a free-market exchange rate system is slated to be implemented. These policy innovations could 29 potentially render the above parameter estimates obsolete. ARIMA intervention analysis (Box and Tiao, 1975) may prove beneficial in subsequent empirical research designed to examine this possibility. It is interesting to note, however, that the similarities between the empirical results reported in this paper and those analyzed in earlier studies indicate that agent responses have generally been relatively inelastic with respect to changing monetary and exchange rate policies in Colombia. CHAPTER 3 SHORT-TERM PRICE MOVEMENTS IN ECUADOR 3.1 Introduction Similar to other Latin American economies, Ecuador has faced persistently high rates of inflation in recent years. Although inflation was substantially lower in 1993 and 1994, excessive money supply growth in early 1995 clouded prospects for additional short-term improvements. Prior to the first- quarter 1995 border skirmish with Peru, the government had signed a stand-by loan agreement with the International Monetary Fund that established a goal of reducing the inflation rate to 15 percent over a 19-month time frame (Banco del Pacifico, 1994a). Not withstanding government assurances that inflation would decelerate to its target rate by year-end 1995, very little econometric analysis using short-term forecasting methods appears to have been relied upon in developing the new policy targets. In its attempt to slow price movements, the DurÃ¡n Ballen- Dahik Garzozi administration introduced a variety of new policy measures. They include import liberalization, fiscal austerity, and a slower rate of currency depreciation. By reducing price pressures, the government hopes to improve economic welfare by enabling the Ecuadorian economy to operate 30 31 more efficiently. This argument is very similar to those aired in advanced economies such as the United States (Motley, 1993) and analyzed in other developing nations (Zind, 1993). What is unigue, however, is the magnitude of the disinflationary goals set by the Ecuadorian policymakers. As a result, short-run price stabilization has become the center piece of government policy efforts in Ecuador. This chapter examines potential results associated with the two principal adjustment tools, money supply growth and exchange rate movements. Despite ongoing difficulties with inflationary uncertainty and the ambitious nature of current policy goals, careful econometric analysis of short-run price movements in Ecuador has not previously been conducted. To bridge this gap, a modeling framework is proposed, tested, and used to develop policy simulation exercises for monthly Ecuadorian price data. In contrast to the time series methodology utilized for Colombia, an econometric approach is followed in the analysis conducted for Ecuadorian price movements. Selection of this alternative approach was motivated by two factors. First was a discussion on guantitative analysis of developing country inflation held at the 32nd International Atlantic Economic Conference. Second was interest expressed by economists at the Research Department of the Central Bank in Quito with respect to attempting to develop a short-run inflationary model similar to the long-run monetary-import cost approach 32 incorporated in Fullerton (1993b). Subsequent sections of this chapter offer a review of the literature, theoretical model, and empirical results. Suggestions for future research are summarized in the conclusion. 3.2 Literature Review The seminal research on inflationary dynamics in developing countries was conducted on Chilean data by Harberger (1963). That early paper interestingly points out that analyzing nominal data in level form could result in spurious correlations in equations estimated for highly inflationary economies. To circumvent this problem, percentage rates of change are utilized in a linear regression framework based on the quantity theory of money. What became known as the "Harberger" framework incorporates real income, current and lagged values of the money supply, and the opportunity cost of holding cash balances. The success of this initial effort conducted on Chilean data spurred a series of replicated studies for other developing countries. Vogel (1974) estimates an inflation equation for several Latin American economies, including Ecuador, using annual data. Results confirm the overall usefulness of the Harberger model. Unlike the study at hand, Vogel utilizes a sample period during which inflation averaged less than 4 percent per year in Ecuador and the exchange rate was fixed. 33 Following numerous applied econometric studies utilizing this approach, it became apparent that its reliance on domestic variables alone often provided unsatisfactory results. Bomberger and Makinen (1979) provide a thorough examination of the Harberger model using quarterly data for Korea, Taiwan, and Vietnam. Extensive testing is conducted using quarterly data in order to establish whether a suitable characterization of inflation is provided. Encouragingly, the parameter estimates do not appear sensitive to the time period selected. However, the elasticities with respect to money and real income are not always unitary as hypothesized. Also, the coefficient signs for the cost of holding money variables are sometimes negative. Hanson (1985) extends the Harberger framework in a systematic fashion to incorporate an important missing component, import costs. An implicit cost function is utilized to derive an aggregate supply curve which includes local prices of imported inputs. When the underlying production function is homogeneous of degree one, inflation becomes a weighted sum of money supply changes and import prices. This is important for studies using higher frequency data if the problem of measurement bias engendered by interpolated values of real output, generally published on either a quarterly or annual basis in developing countries, is to be avoided (see Bomberger and Makinen, 1979). The model also implies the elasticity of inflation with respect to money 34 growth is less than one. Empirical results in the Hanson article strongly support the inclusion of import prices or the rate of devaluation in models of inflation. Subsequent research has provided additional evidence in favor of the augmented Harberger-Hanson approach wherein the effect of import prices on inflation is considered. Koch, Rosensweig, and Witt (1988) and Fullerton, Hirth, and Smith (1991) both report positive linkages between the trade- weighted exchange value of the dollar and consumer prices in the United States. These empirical studies indicate a unidirectional channel of influence from the exchange rate to domestic prices exists in the United States economy. As will be discussed below, causality direction has important implications for both model form and estimation technique. Developing country studies have also confirmed the usefulness of an augmented modeling treatment of inflationary dynamics. Sheehey (1976) reports some of the early econometric work along these lines. Sheehey (1980) reaches additional conclusions on the basis of empirical tests that indicate that accurate assessment of austerity policy efforts will likely require explanatory variables representing cost push factors. More recently, Brajer (1992) provides evidence that the latter category of models may offer better specifications than those which rely solely on domestic economic factors. Conclusions in that article are reached on the basis of F-tests for different regressor sets. Similarly, 35 Fullerton (1993b) successfully imbeds a variant of this approach in a large-scale macroeconometric forecasting model for Ecuador using annual data. There have been very few dynamic models estimated on the basis of monthly data for developing economies. Given that most business decisions in highly inflationary countries are reached within a short-range context, this is an area which needs to be addressed. As detailed in Chapter 2, Fullerton (1993c) empirically examines Colombian anti-inflationary efforts utilizing monthly data with an ARIMA transfer function. The estimated model is found to generate realistic simulation scenarios for policy analysis. The results also support the hypothesis of inflation rate inelasticity with respect to monetary growth. As in the Chilean equation reported by Hanson (1985) , and the Argentine model presented in Sheehey (1976), exchange rate price effects are found to outweigh the monetary coefficient. The latter is somewhat surprising given that imported goods and services comprise less than 20 percent of Colombian gross domestic product. 3.3 Theoretical Model Harberger's (1963) model is based on the traditional quantity theory of money equation: (3.1) MV = PQ, 36 where M represents some measure of the money stock, V is the velocity of circulation, P is the price level, and Q is real output. Velocity is not assumed to be constant. Instead, it is hypothesized to be a predictable function of other macroeconomic variables such as the cost of holding cash balances. Given the typical variability of velocity in many Latin American economies, this aspect of the theoretical model is potentially important (see Clavijo, 1987) . To utilize percentage changes, the variables can be transformed by natural logarithms and first differenced. Introduction of a time subscript, and rearrangement of the terms, yields the basic Harberger eguation: (3.2) DPt = DMt - DQt + D(DPt.1), where the last term results from substituting for velocity and D represents a difference or backshift lag operator. Usage of the lagged change in the inflation rate to proxy for the implicit cost of holding money is motivated by the fact that developing countries such as Ecuador have freguently imposed government regulations on interest rates. The latter have occasionally caused savings and loan rates to become negative in real terms. Unadjusted interest rates from these periods in Ecuadorian economic history do not, therefore, provide accurate estimates for the cost of holding idle cash balances. 37 Equation 3.2 implies that inflation will vary positively with the money supply and inversely with respect to real output. A statistically significant intercept term will enter the estimated equation if there is a trend in the velocity of circulation. If only contemporaneous lags of M and Q enter in the equation, the parameters for both variables are hypothesized to be unitary. This can be tested empirically with the following specification: (3.3) DPt = aQ + a1DMt - a2DQt + ajDÃDP^) + u3, where a1 and a3 are hypothesized to be positive, and the absolute values of a1 and a2 should both be statistically indistinguishable from one. The last argument in the expression represents the disturbance term. Hanson (1985) proposes an implicit cost function dual of an aggregate production function which is homogeneous of degree one. Derived output supply functions from this framework will be homogeneous of degree zero in input and output prices. Equation 3.4 expresses this relationship using logarithmic first differences: (3.4) DQt = b0 + b1DPt - b2DPIt + u4, where PI represents imported input prices. When the relative prices of imported inputs increase, output is assumed to 38 decline. The standard homogeneity assumptions for production and derived supply relations imply that b1 - fc>2 = 0. Equation 3.4 can be substituted into Equation 3.3 to eliminate the output term from the expression to be estimated. As noted in the literature review, this step is useful for avoiding interpolation bias in empirical studies of monthly inflation for countries such as Ecuador where GDP is published at quarterly and/or annual frequencies. The resulting equation can be written as follows: (3.5) (1 + agb^DPj. = aQ - a2bQ + a.,DMt + a2b2DPIt + a3D ( Dpt-1 ) + U5* Equation 3.5 can be further simplified prior to estimation. Dividing through by the left-hand side constant term and rearranging terms such that the price series remains as the dependent variable yields the following relation: (3.6) DPt = C0 + c1DMt + c2DPIt + CjD(DPt.1) + u6, which also has testable properties. Importantly, the coefficient on the monetary variable, c.,, is now hypothesized to be significantly less than one. Also important, with the possible exception of the intercept, all of the regression parameters in Equation 6 are expected to be positive. 39 Several other properties of this model are worth noting. In particular, the theoretical coefficient restrictions described earlier have interesting implications. Namely, a1 and a2 are hypothesized as equal to one, and b1 and b2 are equal in absolute value in the version of the model developed thus far. Substitution into Equation 3.5 implies c., + c2 = 1, which can also be tested. As indicated in the literature review, Equation 3.6 has provided a useful framework for analyzing quarterly and annual inflation rates. But because the lag structure in this model is fairly short, it may require additional modification prior to estimation. This possibility does not reflect any deficiencies in the theoretical model as such, but arises due to the fact that short-term models rely upon monthly data. As a result, if the inflationary impact of a change in the money supply is felt over the course of one calendar year, the implied lag structure for a model estimated with data published at a monthly frequency would potentially range up to 12-months in length. Equation 3.7 takes into account this empirical issue which has confronted and confounded researchers for many years (see Laidler, 1993): (3.7) DPt = c0 + C1DMt.). + C2DPIt_. + CjD(DPt_1k) + u7, where lag subscripts i, j, k=0, ..., n, respectively. 40 The above model provides an attractive starting point for examining inflationary trends in an economy. It is not, however, without potential problems for analyzing price movements in a relatively high inflation country such as Ecuador. A principal concern arises from the fact that Equation 3.7 treats all of the regressors as exogenous or preÂ¬ determined. In doing so, it does not allow for the possibility of statistical feedback or endogeneity between the left-hand and right-hand side variables. If a central bank yields to political pressures and engages in accommodative monetary policy in the face of inflation shocks, this assumption would be violated. As noted in Chapter 2, research conducted using higher frequency data for Colombia indicates that the causality paths in that economy are unidirectional as implied by Equation 3.7 (see Fullerton, 1993c, and Leiderman, 1984) . While short-term forecasting models for Ecuadorian inflation have not been previously developed, domestic prices, monetary aggregates, and import prices are modeled simultaneously in the macroeconometric model estimated using annual data by Fullerton (1993b). It would not be surprising if the feedback relations encountered in that paper also emerge in the monthly time series utilized below. As noted elsewhere, monetary authorities in Ecuador have occasionally been forced to yield to political pressures (Garlow, 1993). Granger causality 41 tests will be used to test the severity of this potential problem. A second possible concern arises from utilizing first differenced, log-transformed time series data in the equation to be estimated. If the resulting series are stationary, the equation can be estimated without risk of obtaining spurious correlations in the results. As shown in many studies of hyperinflationary economies, however, higher order differencing may be required to induce stationarity during periods in which prices increase rapidly (Engsted, 1993). Because Ecuador has not undergone any hyperinflationary episodes, first differencing should remove nonstationary trends from the variables in question but this assumption must be tested. The latter tests are accomplished below via a battery of unit root tests, not all of which are reported. A third concern arises from the fact that monthly import price deflators do not exist for Ecuador. To circumvent this problem and also avoid interpolation bias, a trade-weighted exchange rate index is used as a proxy for imported input prices. The index utilized was developed econometrically and takes into account export and import volume changes with Ecuador's major trading partners. It also offers a single monthly index for periods when the government has instituted multiple exchange rate systems (Fullerton, 1989b). To construct the currency index, individual currency weights are calculated as the sum of imports and exports with 42 each of ten major trading partners and divided by total international trade in each year. Over the sample period, annual trade with Ecuador's top ten import sources and export destinations accounts for more than 70 percent of its total trade volume in any given year. Products of the bilateral trade coefficients and the respective currencies are then used to construct the exchange rate index using a geometric mean. The latter method is selected to avoid problems which can potentially result for indexes constructed using arithmetic means during periods of inflationary variability (see Batten and Belongia, 1986, Dutton and Grennes, 1987, and Kercheval, 1987) . For periods when multiple exchange rates were instituted, a blended index is calculated. Weights for the free-market and official government intervention exchange rates are obtained from the Central Bank publication, InformaciÃ³n EstadÃstica Mensual (various issues). Econometric results for total and disaggregated imports, and non-petroleum exports indicate that the blended rate provides a more accurate measure of the appropriate currency basket for Ecuador (Fullerton, 1989b). Diagnostic tests were also conducted using currency baskets with different numbers of trading partners. Introduction of the trade-weighted exchange rate index to Eguation 3.7 causes the model to be estimated to take the following form: 43 (3.8) DPt = g0 + g1DMt.l. + g2DTWXt.. + g3D (DPt.1.k) + U8, where TWX stands for logarithmic first differences of the nominal version of the monthly trade-weighted exchange rate index calculated for the sucre by The WEFA Group (formerly Wharton Econometrics). Although incorporation of the monthly exchange rate index avoids the problems associated with implicit price deflator interpolation, it may introduce other problems due to the fact that import prices in Ecuador are affected by global supply and international demand conditions in addition to exchange rate movements (see Fullerton 1993b). A fourth observation regarding potential pitfalls associated with the theoretical specification of the model is worth noting. Only one class of factor input prices, that for imports, is included. While this represents an improvement over the original Harberger framework, it may overlook additional important inputs such as labor. If the version of the model developed herein omits relevant variables to the inflationary process in Ecuador, it is likely that estimated residuals associated with the empirical version will not be randomly distributed. If this is the case, then correction for serial correlation will be necessary. Extension of the model to overcome this problem cannot currently be accomplished due to data availability. 44 3.4 Estimation Results In order to examine whether the working series included in Equation 3.8 are stationary, unit root tests are conducted for each series utilized in the model. Estimation is conducted for the 1964-1994 sample period for which data are available. Applying unit-root tests to what could be considered a relatively short time span may be risky due to the fact that these tests typically have low power unless long-run data sets are used (Hakkio and Rush, 1991). Because time series data in Latin America generally date back to 1957 at most, there is little that can be done to circumvent this potential problem. Augmented Dickey-Fuller t-statistics are estimated for equations with both intercepts and trends. These results, compared against the corresponding MacKinnon critical value, appear in Table 3.1. In all cases, tests for unit roots in the first differenced log transformed series for consumer prices, money, and the trade-weighted exchange rate index are rejected at the 1-percent level. Based on this evidence, the first-order differenced series used to estimate Equation 3.8 are assumed to be stationary. As specified, the model is explicitly built around a set of unidirectional causality relations from movements in the regressors to consumer prices. To examine whether the absence of simultaneity in the model is plausible, a set of Granger causality tests are calculated for the stationary components 45 of the series of interest. These results are reported in Table 3.2 for lags of 6, 12, 18, and 24 months. Tests are conducted for prices and the money supply, as well as prices and the trade-weighted exchange rate. Similar to the Colombian results reported in Chapter 2, movements in Ml do not appear to be systematically preceded by changes in the CPI in a statistically significant manner at the 5-percent level. Essentially, this implies that monetary policy in Ecuador is conducted in a manner that is not accommodative of price shocks. Although central bank linkages to the executive branch of the government are relatively strong, steps have been taken in recent years to increase both monetary policy autonomy and currency stability (Banco del Pacifico, 1994b). Implications based on the Granger causality tests estimated for consumer prices and the exchange rate series are less clear. At shorter lag lengths, the null hypothesis that changes in consumer prices do not lead to subsequent changes in the international value of the sucre is rejected at the 5- percent level. This conclusion, similar to the Colombian results obtained by Fullerton (1993c) and Kamas (1995), is not upheld at longer lag lengths and makes it difficult to reach concrete conclusions regarding potential feedback effects between the two series. However, truncation of the longer lag lengths can bias the hypothesis toward incorrect rejection of the null hypothesis (Feige and Pearce, 1979). Consequently, 46 Table 3.1: Unit Root Tests for Stationarity Series Aug Dickey-Fuller t-stat MacKinnon crit value P - 9.778 (with const, trend, 1 lag) -3.995 (1% lvl) M -16.626 (with const, trend, 1 lag) TWX -13.652 (with const, trend, 1 lag) 47 Table 3.2: Granger Causality Tests Causality Direction Number of Lags Computed F-stat CPI => Ml 6 1.659 CPI => Ml 12 1.309 CPI => Ml 18 1.420 CPI => Ml 24 1.509 CPI => TWX 6 2.900 CPI => TWX 12 2.214 CPI => TWX 18 1.664 CPI => TWX 24 1.029 48 a unidirectional channel of influence from exchange rate movements to prices appears to be a reasonable assumption. Given the latter, model estimation is conducted without resorting to instrumental variables, or developing a system of simultaneous equations, and the resulting coefficients are assumed to be unbiased and consistent. Regression results for Equation 3.8 are summarized here: (3.9) DPt = 0.015 (10.224) 0.028*DMt.4 (1.735) 0.060*DTWXt (5.932) + 0.013* DMt (0.797) + 0.039*DMt,9 (2.333) + 0.007 *DTWX (0.642) t-2 0.063*D(DPt_6) + 0.405*Ut_2 (1.638) (8.439) 0.3 69*V (7.101) t-1' R2 0.379 S.E.R. 0.015 Log likelihood 1013.715 DW 2.087 F-stat 26.798 Prob(F-stat) 0.001, where the numbers in parentheses are computed t-statistics and Vt is the error term associated with the ARMAX model for Ut. Lag lengths of 24 months were used in the initial estimates for Equation 3.9. While a large number of the resulting coefficients were significant at the 5-percent level, serial correlation was present in the residuals. To avoid potentially spurious estimation results (see Hamilton, 1994), a nonlinear ARMAX procedure is utilized to correct for autocorrelation. This estimator (Pagan, 1974) is useful 49 because of its flexibility in handling a variety of different error generating processes. Correcting for serially correlated disturbances caused the computed t-statistics for many coefficients to become insignificant at the 5-percent level. Because inclusion and exclusion of numerous different lags did not yield clear results, the model structure reported in Equation 3.9 was selected on the basis of likelihood ratio tests. Although the relatively short lag components may seem unexpected, they were confirmed by cross correlation function analysis. To handle autocorrelation, an ARMA(2,1) specification is used to characterize the data generating process for the residuals. As would be expected in an inflationary economy, the algebraic sign of the intercept in Equation 3.9 is positive. Because the model is estimated using differenced data, this result indicates that a systematic upward trend exists in the Ecuadorian consumer price index. As hypothesized by Hanson (1985), the sum of the coefficients for the lagged monetary series is significantly less than one. Similar to Fullerton (1993a, 1993c), but unlike Kamas (1995), the exchange rate appears to play an important role in determining price movements. The coefficient for the velocity of money supply circulation proxy was not, however, significant at the 5- percent level. Although the sum of the lagged monetary aggregate parameters is less than one, the sum of those coefficients 50 with the exchange rate variable coefficients does not equal one as implied by the reduced form of the theoretical model specification. These results cast doubt upon the relatively simple version of the Harberger-Hanson framework developed above. As noted previously, this is not completely unexpected and is a potential root cause underlying the presence of autocorrelated residuals in the initial empirical results. To see if the theoretical model can be improved, an alternative version of the approach is currently under development for an economy where data shortages are less severe and a richer input structure can be handled (Kim and Fullerton, 1996). The latter effort will also benefit from the incorporation of an interest rate variable to measure the cost of holding idle cash balances, as well as an import price index rather than the exchange rate proxy utilized herein. In 1983, the Central Bank introduced new exchange rate policies that allowed the sucre to fluctuate more freely. To allow for potential parameter heterogeneity caused by structural change associated with periodic currency devaluations, a shorter sample was also used for estimation purposes. Results from these exercises, not reported here, generally support the empirical estimates presented above. Experimentation with the lag structure over the shorter sample period does not yield strong evidence of parameter instability or any other major shortcomings with Equation 3.9. 51 3.5 Policy Simulation Results From the alternative specification and sample size results, Equation 3.9 does not seem overly fragile. Alternative lag structures and ARMAX processes were compared to the model using likelihood ratio tests to evaluate the results. While it was not possible to reject the specification shown above, it does seem likely, however, that additional time series analysis will be required in order reach firm conclusions regarding inflationary dynamics in Ecuador. For that reason, the modeling and simulation results associated with Equation 3.9 should be regarded as preliminary in nature. They do, however, provide a good starting point for understanding short-term Ecuadorian price movements and assessing government policy objectives. Cognizant of the paucity of comparative research results in this segment of the literature, a variety of simulation experiments are conducted using Equation 3.9. The goal of the simulation tests is to shed light on the feasibility of attaining the inflationary stabilization targets announced by the DurÃ¡n Ballen-Dahik Garzozi administration in 1995. Sample data used to estimate Equation 3.9 only includes information available to government policy makers at the time the inflation target was announced. The simulation analyses thus satisfy the Klein (1984) and Christ (1993) criteria for forecast evaluation. Of course, additional policy analysis with different model specifications may also be useful. 52 To examine the feasibility of the government's inflation goals, four simulation exercises are conducted using Equation 3.9. The first exercise assumes immediate implementation of price stabilization plan wherein the annualized rate of growth in the money supply and the rate of devaluation are reduced to 20 percent within one month of the policy announcement. Scenario two examines rapid implementation of the price stabilization whereby growth in the money supply and the rate of devaluation are reduced to 20 percent over a 6-month phase- in period. The third simulation test analyzes the effects of gradual introduction of the anti-inflationary program with money supply growth and the rate of devaluation lowered to 20 percent during a 12-month period. The rates of money supply growth and currency depreciation reflect the revised policy targets adopted following the border conflict with PerÃº and the aftermath of the "Tequila effect" associated with Mexico's December 1994 peso devaluation (Banco del PacÃfico, 1995). Policy simulation results are reported in Table 3.3. The impacts associated with all three implementation scenarios are striking. In each, reducing the rate of money supply growth and slowing the rate of nominal currency depreciation to 20 percent over a one-to-twelve month implementation period causes inflation to decline to less than 25 percent. This is somewhat close to the policy target established by the government in late 1994. If implementation of the stabilization program is immediate, annual consumer prices 53 increases fall to slightly more than 23 percent. In a more likely scenario under which intermediate policy targets are attained more gradually, disinflation is still fairly rapid and overall policy credibility would not appear to be at risk. To actually achieve the announced inflation target, however, does not appear likely. The final column of Table 3.3 contains the results associated with a scenario in which no progress is made with respect to lowering the rate of growth in the money stock. Similarly, nominal depreciation of the sucre is not lowered under this simulation. In contrast to the "no implementation" policy simulation results reported in Table 2.7, the inflation rate remains practically unchanged if no intermediate steps are taken by Ecuadorian monetary authorities. This result is due in large part to the fact that consumer prices in Ecuador rose on average by approximately 26.7 percent per year during the past three decades. As a result, the policy experiments illustrated in Table 3.3 have a common starting point that is almost identical to the sample period mean. On the basis of the empirical evidence obtained in this chapter, it appears that Ecuador's anti-inflationary program is fairly credible. This conclusion is predicated upon eventual deceleration in the rates of liquidity growth and currency depreciation brought about by the central bank. In light of previous monetary policy analysis conducted for 54 Table 3.3: Policy Simulation Results Month Immediate Rapid Gradual None 26.1 26.2 26.2 26.2 24.9 25.1 25.3 26.0 24.3 24.8 25.1 25.8 24.0 24.5 25.1 25.9 23.2 23.7 24.2 26.0 12 55 Ecuador, the results obtained are pleasantly surprising (Fullerton, 1990a, offers a negative assessment of an earlier price stabilization effort). That the government's announced targets are not completely attained is in line with other studies of Latin American inflation policies (Fullerton, 1993c, and Kamas, 1995). 3.6 Conclusion An empirical model of Ecuadorian consumer price inflation is developed and estimated in this chapter by incorporating both monetary and import cost effects in a theoretically plausible manner. Specification and simulation of the model are relatively easy to accomplish. Experimentation with the estimated equation indicates that the current anti- inflationary goals of Ecuadorian monetary authorities are in large part attainable. Because the model does not pose stringent data requirements, it may be applicable to other Latin American economies where inflation remains a problem. Examples include Brazil, Colombia, Mexico, and Venezuela where authorities continue to grapple with short-term price stabilization goals. Additional econometric testing should prove useful. Initial results reported above indicate this framework will likely benefit from incorporating a more realistic conceptual model. Because ARMAX treatment of nonrandom movement in the initial model residuals was necessary, expansion of the scope 56 of the model to include additional factors such as labor costs may be required to more completely specify the inflationary process in Latin America. Doing so, however, may necessitate the usage of instrumental variables or the introduction of multiple equations which allow for potential endogeneity between prices, money, import costs, and wage rates. Even if the latter are not required for parameter estimation consistency, they could enrich subsequent policy simulation analyses. These suggested changes represent avenues for refinement to the basic model outlined above. They are not likely to result in wholesale alterations to the general framework. Similarly, it is not clear that policy simulation impacts and conclusions will change markedly due to expanding the scope of the empirical techniques presented above. But given the breadth of economic conditions prevailing across Latin America, steps in these directions may prove helpful to subsequent econometric research of this nature. Given the divergence between the theoretical model parameters and the estimated coefficients, additional empirical testing is certainly warranted. CHAPTER 4 PREDICTABILITY OF SECONDARY MARKET DEBT PRICES 4.1 Introduction While inflation has undoubtedly been one of the most hotly debated items in Latin American policy debates, issues related to external obligations have also been important topics of discussion in the region. Following the outbreak of the international debt crisis in September 1982, a secondary market for sovereign debt instruments became active in the major world financial centers. As the payments crisis spread from Mexico to the rest of Latin America, much of Africa, and parts of Asia, secondary market trades in developing country sovereign debt paper increased. Key issues in the emerging debate regarding potential solutions to the payments problem often involved the treatment of discounts from face value implied by secondary market debt prices. Accordingly, researchers began investigating different aspects of the behavior of secondary market developing country debt prices. Much of this work investigates the applicability of theoretical valuation models to assessing implied discounts from face. Other research has used secondary market prices and other financial data to evaluate the probability of payment rescheduling requests. Authors have also attempted to 57 58 assess the sensitivity of the secondary market to macroeconomic fundamentals, but these efforts have not investigated whether or not the discounts from face value on these obligations are predictable. Several basic questions are investigated in the chapter at hand. An important issue to be considered is the nature of the time series behavior of these prices. Under certain conditions in an efficient market, each individual series might be expected to follow a random walk. Of course, in a relatively thin secondary market such as that for developing country debt, the perfectly competitive hypothesis may not always be satisfied. If movements in the price series are nonrandom, it may be possible to relate these variations to changes in other domestic and international economic indicators which are generally included in commercial forecasts of the region. Candidate series which help assess creditworthiness include variables such as interest rates, export prices, and international reserves (see Fullerton, 1991, 1993a). Anecdotal evidence indicates that countries with good payments records such as Colombia have seen their access to international commercial credit diminish over the last fifteen years. If the "good debtor in a bad neighborhood" contagion effect is present, this will be manifested in the movements of the discounted prices. Debt prices for individual countries would consequently be affected by developments in neighboring 59 countries and correlated with the prices for the instruments of those economies and the market in general. The generalized least squares estimation procedure used below implicitly takes this possibility into account by allowing for contemporaneously correlated residuals across equations. To examine the predictability of these series, forecasts are generated using jointly estimated individual country models. For comparison purposes, baseline projections are developed using the random walk assumption that the best forecast is one of no deviation from the last observation available in each data set. Modified Theil U-coefficients are then calculated on the basis of root mean squared error (RMSE) ratios estimated for each set of model based and random walk forecasts. The U-statistic for an individual forecast step- length is equal to the ratio of the model based RMSE to that of a no change RMSE. When the resulting coefficient is less than one, the equation forecast has outperformed the random walk prediction. 4.2 Earlier Studies There have been a growing number of studies regarding developing economy external indebtedness and the secondary market in recent years. Gennotte, Kharas, and Sadeq (1987) develop a numerical debt valuation method based on a financial options pricing technique. Using liquid international reserves plus the estimated values of the capital stocks in 60 mining and manufacturing as proxies for collateral in each country, theoretical values for developing country foreign debts are simulated under different scenarios involving interest rate changes, principal due, and front-end fees. Simulation results are found to be positively correlated with secondary market prices reported for 1985. Stone (1991) examines the behavior of implied returns on secondary market sovereign debt instruments. Using an arbitrage pricing model approach, he examines the empirical relationships between secondary market returns and various macroeconomic variables. To control for cross equation disturbance simultaneity, a seemingly unrelated regression estimator is utilized. Poor equation fits and weak t- statistics indicate that movements in implied sovereign debt returns are not readily explained by arbitrage pricing techniques. Other empirical studies of sovereign debt problems have been more successful. Rahnama-Moghadam and Samavati (1991) employ probit models to examine the propensity to default. Ten different macroeconomic and international financial ratios are used to quantify the probability of rescheduling. Among the ratios found most useful in predicting debt moratoria, formal or informal, are the following: international reserves relative to imports of goods and services; international reserves to disbursed debt; disbursed debt relative to exports; disbursed debt to gross domestic product; and 61 interest payments relative to exports of goods and services, also known as the interest service ratio. Parameter estimates are based on annual cross country sample data from around the globe. The probit results discussed in the preceding paragraph were later confirmed by subseguent research which utilized data from Latin American economies (Rahnama-Moghadam, Samavati, and Haber, 1991). Reasons offered for segmenting the data in this fashion include geographic, structural, and institutional similarities among countries in Latin America. The authors also point out that economies in this smaller sample are all middle-income countries which share similar characteristics in terms of overall development. The models exhibit relatively high goodness-of-fit statistics, individual coefficients with expected algebraic signs, statistically significant parameter estimates, and coefficient stability across different specifications. As debt problems arise from a number of different determinants, it would appear that a further refinement from a regional focus to that of an individual economy may be useful. The roles of fundamental economic and financial factors in the evolution of secondary market sovereign debt prices have been directly examined in recent research. As shown by Anayiotos and de PiniÃ©s (1990), these types of characterizations are straightforward and intuitively appealing. In a statistical framework, variables selected to 62 capture market fundamentals can also be combined with regressors designed to represent exogenous risks. Using pooled observations and annual data, the econometric results of these authors show that even simple specifications can represent secondary market developing debt prices with a high degree of accuracy. Perasso (1989) also emphasizes economic factors in his study of secondary market prices. To reflect the importance of debt-equity conversions, his pricing model is derived from a profit maximizing framework that includes real interest rate measures, real costs of capital, international wage differentials (assumed to induce manufacturers to invest abroad), and individual economy performance variables. Time series and cross country annual data are pooled prior to estimation. Some coefficients are statistically insignificant or of the wrong sign, but overall empirical results for the estimated equations for secondary market prices are fairly strong. Empirical models developed by investment bankers indicate that secondary market debt prices can be modeled and forecasted (see Purcell and Orlanski, 1988, 1989). Similar to the studies mentioned above, these models also rely upon pooled cross section time series data for different countries. Regressors used to estimate equation parameters include debt to export ratios and per capita incomes. Dummy variables for payment rescheduling programs, principal payment moratoria, 63 and debt retirement agreements are also entered as right-hand side variables. Model simulations using individual country data are used to calculate specific secondary price forecasts. As mentioned previously, not all of the statistical evidence with respect to the behavior of secondary market prices leads to the same conclusions. Laney (1987) concludes that economic factors are more important as explanatory variables than political and structural risk factors. Sachs and Huizinga (1987) report regression results that indicate both economic and political variables have key roles to play in modeling developing country debt discounts. Similar to other studies, the latter also utilize pooled cross country data in calculating eguation parameters. 4.3 Empirical Analysis As the literature review indicates, there have been a variety of studies with interesting econometric results published in recent years. However, none of the initial efforts attempted to model secondary market debt prices for individual countries using time series sample observations. Given the differing sources of debt servicing difficulties, a country by country examination of secondary market price movements appears warranted. Similarly, individual forecasting models or eguations have not been systematically tested to examine whether debt prices can be predicted with any degree of accuracy. Despite the fact that financial 64 market participants focus almost exclusively on short-term movements in sovereign debt prices, earlier research efforts also failed to study high-frequency data from this market. In this section of the chapter, simple forecasting models are proposed, estimated, and simulated for three individual debtor countries. These equations are modeled jointly using monthly data series. As stated in the introduction, modified Theil inequality coefficients are calculated using random walk forecasts as the benchmarks to which the model projections are compared. Although the individual equation specifications employed are straightforward, recent financial market research underscores the potential success of simple forecasting models (see Granger, 1992, and Christ, 1993). This initial attempt provides a useful starting point for addressing questions regarding the predictability of secondary market developing country debt prices. Debt price data used in the empirical estimates are collected by Salomon Brothers in New York, with monthly averages published by The WEFA Group in Philadelphia. The sample period is March 1986 - December 1991. Countries included in the sample are Colombia, Ecuador, and Venezuela. These countries have interesting and differing histories with respect to their external debt management practices and their individual approaches to economic policymaking in general. Colombian debt management practices have traditionally been more conservative than those of either of its neighbors. 65 Government economists have consistently treated international credit markets as sources of financing to bridge domestic savings gaps (Fullerton, 1990b). Colombian debt negotiators have never sought a Brady initiative write-off, arguing that to do so would only impair the nation's creditworthiness. Over the period from 1945 forward, Colombia has never declared an interest payment moratorium and has one of the best debt service records among developing countries which have utilized external financing sources. Ecuador declared principal and interest moratoria in 1987 as a result of the financial aftermath following the earthquake which shattered the country's transAndean oil pipeline. The latter event interrupted Ecuador's principal source of export earnings and destroyed much of its physical infrastructure. In 1989, the Borja administration resumed negotiations with commercial creditors and eventually began honoring 30 percent of the interest coming due on commercial loans (Fullerton, 1989c). Progress regarding the treatment of growing amortization and interest arrears remained elusive through the balance of the Borja government which stepped down in 1992. A new round of discussions with Ecuador's bank advisory committee began after the DurÃ¡n Ballen-Dahik Garzozi government took office. Venezuela also encountered debt service problems following negative oil price shocks in 1986 and 1988. Eventually, the Perez administration rescheduled commercial 66 credits under a Brady initiative agreement with Venezuela's bank advisory committee (Fullerton, 1990c). The agreement offered bank loan syndicate members five menu options designed to relieve balance of payment pressures faced by this economy. Several Euromoney bond issues were successfully floated in subseguent periods and Venezuela temporarily regained access to international credit markets. Debt instruments for all three countries have traded at substantial discounts from face value in recent years. Given its superior service record, it is not surprising that Colombian paper generally carries a higher price than that of its two Andean neighbors. Similarly, given its higher level of payment arrears and worse economic performance, Ecuadorian paper tends to trade at sharper discounts than those of the other countries in the sample. Table 4.1 presents summary statistics for each secondary market sovereign debt price series. Discounts from face value on Colombian debt certificates were less variable than those of Ecuador and Venezuela during the March 1986 - December 1991 sample period. The range and standard deviation for Colombian debt guotes are smaller than the others, while those for Ecuador are the largest of the three. The arithmetic means for each series follow the opposite pattern in terms of ranking. Data in Table 4.1 are in cents per dollar, or percent of face value of the loans, the units in which transactions are conducted at money center trading desks. 67 Table 4.1 : Secondary Market Debt Price Summary Statistics Country Average Maximum Minimum Standard Deviation Colombia 69.990 86.00 50.00 10.415 Ecuador 30.332 68.00 10.67 18.728 Venezuela 57.635 78.50 31.50 14.859 The sample period is March 1986 - December 1991. 68 Each series is modeled as a function of key economic and financial variables which are easily observed as well as likely to be used by secondary market participants as indicators of the creditworthiness of the individual country. Possible regressors include lagged debt quotes, world interest rates, commodity export prices, international reserves, and domestic price indexes (for discussion, see Wakeman-Linn, 1991). As mentioned above, series such as these are typically included in macroeconometric forecasting models of Latin America due to their usefulness in predicting balance of payment movements and general economic performance (see Fullerton, 1993a, 1993b). A three-stage generalized least squares (3SLS) regression technique is used to jointly estimate model parameters (Zellner and Theil, 1962) . Doing so permits incorporating potential cross-equation simultaneity effects of events such as Citibank's decision to unilaterally increase loan loss reserves in 1987. The latter is believed to have reduced overall secondary market liquidity and also reduced the attractiveness sovereign debt paper to most creditors, irrespective of their individual payment records (Gajdeczka and Stone, 1990, and Snowden, 1989). This estimator also allows for potential simultaneity between the dependent variables and the independent variables. In the case of Ecuador, this is important because feedback exists between the secondary market debt price and the 180-day LIBOR. 69 Individual models may be written conceptually as: 4.1 Pt = b0 + b,x1t + ... + bnxnt + et, where pt is the secondary market sovereign debt price series for an individual country at time period t, x1t, . .., xnt are predetermined domestic and international variables for each economy, et is a random disturbance term, and b1# ..., bn are regression coefficients. From a theoretical modeling perspective, this approach may seem informal. Two points are relevant. First, alternative methodologies were originally considered but deemed inappropriate due to data requirements and difficulties in applying them to forecasting problems (Fullerton, 1990d). Second, the complete absence of other studies of this nature increases the value of an initial attempt to establish whether any regularities at all are present in the data (Christ, 1994) . Both points were repeatedly raised by financial economists who participate in the secondary market and attended the Sovereign Debt Conference sponsored by The WEFA Group in New York in 1990. Similar observations are also made by Friscia (1993). As mentioned above, parameter estimation is accomplished using the 3SLS methodology developed by Zellner and Theil (1962). Other procedures were considered, but 70 monthly observations constitutes a fairly small sample for many time 70 series estimators. The series are not differenced prior to estimation, but are logarithmically transformed. Because the data are in levels, it is important to assess whether the series are cointegrated. The latter assumption is tested via unit-root tests on the individual model residuals (for discussion, see Hamilton, 1994) . Because monthly data are used, it is not possible to utilize the same regressor variables as have been used in earlier studies incorporating quarterly or annual series from national income and product accounts. There are still a number of potential candidate series which the financial community may use as indicators of a country's creditworthiness and will potentially influence secondary market price quotes. It should be noted that sets of indicators will vary for different countries according to individual economic endowments and performance records. This argument is similar to that previously made for equations used to estimate developing country borrowing levels under different regimes (Eaton and Gersovitz, 1981). For Colombia, Equation 4.2 in Table 4.2, the regressors include a one-month lag of the debt price, the effective annual rate for the 180-day London interbank offer rate (LIBOR), and the one month change in the national consumer price index (CPI) . The one-month lag on the debt price is included to provide information on how the market has valued Colombian paper in the most recent period. International 71 loans to sovereign nations typically carry variable interest rates defined in terms of a fixed spread over the 6-month LIBOR. Upward changes in the variable interest rate assessed on such loans will reduce Colombia's current account balance and exert downward pressure on secondary market prices. The rate of inflation is used as a proxy for overall economic conditions. When inflation rises, Colombian monetary authorities generally attempt to tighten credit conditions (Fullerton, Fainboim, and Agudelo, 1992). Parameter estimates for each variable have the expected arithmetic signs, but the t-statistic for the inflation term is not significant at the 5-percent level. In the case of Ecuador, Equation 4.3 in Table 4.2, only two predetermined variables are included in the three-staged least squares equation. The first is a one-period lag of the secondary market debt price series. The second is the 6-month LIBOR rate. These series are included for the same reasons as they were used in the Colombian equation. Both coefficients have the expected signs and are statistically significant. The absence of other balance of payment indicator series such as export prices may seem surprising. Because Ecuador, similar to Colombia, has a relatively diverse commodity export basket, no single price series will suffice (Fullerton, 1993a, 1993b). This is not the case for Venezuela, where petroleum products account for more than 80 percent of total merchandise exports (Fullerton, 1990c). 72 Table 4.2: Three-Stage Least Squares Regression Results 4.2 Colombia Pt = 1.142 + 0.8l8*Pt.1 - 0.172*LIBOR6M (3.467) (15.946) (3.128) 0.916*CPIt (1.735) R2 0.951 Q (6) 6.118 SER 0.034 Q(12) 8.848 F-stat 366.471 Prob(F-stat) 0.001 4.3 Ecuador Pt = 1.352 + 0.883*Pt,1 (3.376) (26.893) 0.476*LIBOR6M (3.227) R2 0.981 Q(6) 5.632 SER 0.088 F-stat 1462.68 Q(12) 7.803 Prob(F-stat) 0.001 4.4 Venezuela Pt = 0.972 + 0.8ll*Pt,1 (2.765) (16.132) 0.2 4 9 * LIBOR6M + (2.529) 0.061*IRt (1.735) + 0.072*OILt (2.508) R2 0.959 SER 0.057 F-stat 319.777 Q(6) 3.717 Q(12) 5.671 Prob(F-stat) 0.001 The sample period is March 1986 - December 1991 73 Four independent variables are included in the Venezuelan model. As shown in Equation 4.4 in Table 4.2, they include a one-period lag of the debt price, the 180-day LIBOR series, international reserves net of gold, and the average price of petroleum exports. International reserves and the oil export price are employed as proxies for expected economic conditions in Venezuela. Although related, the two series do not always follow parallel paths. More specifically, the level of international reserves serves as an indicator of domestic economic policy success or failure (Fullerton, 1990c). World oil prices provide a measure of global demand for the nation's principal export. All of the regressor coefficients have the expected signs. The t-statistic for international reserves, however, is not significant at the 5-percent level. The 3SLS technique utilized to estimate the parameters reported in Table 4.2 allows for potential cross equation contemporaneous error correlation, a reasonable assumption for the secondary market for sovereign debt paper. To examine whether this assumption is necessary, residuals from the first stage ordinary least squares (OLS) regression are tested for contemporaneous correlation using a standard t-test methodology (see Ostle and Mensing, 1975, or Snedecor, 1956). As shown in Table 4.3, the OLS residuals for Colombia and Venezuela are correlated in a statistically significant manner, as are those for Ecuador and Venezuela. 74 Table 4.3: OLS Cross Equation Correlation Coefficients Equation Residuals Correlation Coefficient Computed t-stat Colombia-Ecuador 0.101 0.772 Colombia-Venezuela 0.340 2.750 Ecuador-Venezuela 0.453 3.872 75 Unit-root cointegration tests performed on the individual country 3SLS residuals are reported in Table 4.4. In all three cases, it appears cointegrating vectors have been obtained. Note that this procedure may be inappropriate for the short data set which currently exists for secondary market developing country debt prices. Previous research has shown that tests associated with unit-root techniques have low power when applied to short-run data sets (Hakkio and Rush, 1991). For purposes of the chapter at hand, the results in Table 4.4 are interpreted as evidence that the 3SLS regression results are not spurious. Given the diagnostic statistics also reported in Table 4.2, the latter conclusion is probably reasonable. To conduct ex ante dynamic simulation exercises, the equations were jointly reestimated and simulated for 24 different historical subperiods. Four month ahead forecasts were produced for each secondary market debt price series from the last subperiod observation forward. A four-month step length is sufficient to incorporate quarterly portfolio accounting considerations faced by international banks. As pointed out by participants at the 1990 Sovereign Debt Conference, quarterly corporate income tax filing requirements often trigger loan certificate swaps between secondary market participants and make a four-month time frame logical for simulation experiments. 76 Table 4.4: Unit Root Cointegration Tests Country Aug Dickey-Fuller t-stat MacKinnon crit value Colombia Ecuador Venezuela -4.895 (with const, trend, -5.140 (with const, trend, -5.320 (with const, trend, 1 lag) -4.122 (1% lvl) 1 lag) 1 lag) 77 To further enhance the realism of the extrapolation scenarios, actual forecast data available to the financial markets during each of the 24 subperiods are incorporated in the simulations. By relying on forecast data which are unconditional upon any information not available prior to the start of any simulation period, Klein's (1984) forecast evaluation criterion is met. These estimates for the independent regressors were compiled from international outlook reports published by The WEFA Group from 1989 to 1991. Forecast results for each price series are summarized in Table 4.5. For Colombia, the results indicate that the secondary market LDC debt price series is predictable using econometric methods. Although the 1-month ahead Colombian projections have a slightly greater than unity U-coefficient, the remaining estimates are all less than one. Interestingly, the Colombian inequality coefficients monotonically decrease as the length of the forecast period increases. This suggests that inclusion of econometric information, in this particular instance, grows in importance as the timeframe under consideration expands. In the case of Ecuador, the results reported here suggest that its secondary market debt discount is unpredictable. Consequently, it appears that financial analysts can do no better than utilize the last available observation for Ecuadorian paper in anticipating future quotes. Additional research with alternative estimators and different 78 Table 4.5: Modified Theil Inequality Coefficients Series 1 Step 2 Steps 3 Steps 4 Steps Colombia 1.070 0.888 0.728 0.611 Ecuador 1.715 2.109 2.408 2.614 Venezuela 1.263 1.207 1.108 1.017 The simulation period is January 1990 - December 1991. 79 specifications might obtain superior forecast accuracy, so it may be premature to conclude that the Ecuadorian debt price series is not predictable. As presented above, however, the results obtained in this chapter provide a fairly striking example of the fact that a relatively high coefficient of determination does not guarantee automatic simulation accuracy. Forecasting results for the Venezuelan debt price series are also interesting. Similar to Colombia, the inequality coefficients decline monotonically as the length of the projections increases. Unlike the Colombian example, however, the Venezuelan U-statistics remain at least slightly above unity at each step length. On the basis of the results in Table 4.5, it therefore appears that a random walk approach yields better forecasts. Experimentation with other estimators is probably warranted. Similar to Ecuador, however, the Venezuelan modeling and simulation results provide another example of a case in which a high coefficient of determination does not guarantee prediction accuracy. 4.4 Conclusion A simple questioned is asked in this chapter. Are secondary market debt prices predictable? To shed light on the possible answer, several steps are taken which have not previously been investigated. A key aspect that distinguishes this research from earlier efforts is the modeling of monthly 80 time series data for individual economies, as opposed to cross section annual or quarterly samples used elsewhere. Also developed are forecast exercises designed to meet the needs of participants in international financial markets where developing country debt instruments are traded. Modeling and simulation results reported here provide only limited evidence that it is possible to forecast secondary market developing country debt discounts from face value. All three series in the sample exhibited interesting model characteristics. In particular, the equations and forecasts for Colombia are encouraging. The same cannot be said of the Ecuadorian and Venezuelan prediction tests, as in both cases simple random walk forecasts of the respective debt price series prove more accurate. Because the three economies, and their respective debt price series, differ substantially from one another, it may be useful to apply this modeling approach to other debt price series. Obvious candidates include Argentina, Brazil, Chile, Mexico, and PerÃº. Further specification enhancements for Ecuador and Venezuela may also provide useful information. Inclusion of institutional variables related to Brady- initiative debt negotiations might prove beneficial, especially if it is possible to construct monthly indices for these factors. As noted by Friscia (1993), however, confidentiality restrictions may preclude this possibility. 81 Similarly, it may be worthwhile to test alternative estimators. The latter will become more feasible as additional observations and information regarding this market become available. On balance, however, it appears that movements in secondary market debt prices are not predictable. CHAPTER 5 SUGGESTIONS FOR FUTURE RESEARCH In Latin America, short-term econometric forecasting analysis is still a largely uncharted area of research. Material presented above indicates that a variety of techniques, methodologies, and modeling approaches may yield interesting insights with respect to both forecasting and policy issues. Not surprisingly, this research has only hinted at a few of the potentially beneficial topics which merit further attention. As discussed elsewhere, the availability of low cost computer hardware and software, plus the development of wide coverage high-frequency data banks, will help encourage additional research of this nature (Fullerton, 1992). With respect to the results presented above, new efforts are already underway in terms of alternative estimators for secondary market debt price forecasting. The underlying theoretical model presented in Chapter 3 has also been extended to include labor costs as part of the price vector on the output side for studying price dynamics. From a business forecasting perspective, a promising area of endeavor is likely to arise from the publication of quarterly national income and product account data. In 82 83 countries such as Ecuador, this will permit the development of large-scale macroeconometric models such as those pioneered by Barger and Klein (1954). The latter continue to enjoy a central role in business and government planning exercises throughout the world. Given the short-term uncertainties present in Latin America, guarterly modeling and forecasting will likely be welcomed with enthusiasm. At this juncture, much work remains to be done, but initial efforts such as those presented above point to numerous potential successes to be gained from future research efforts. REFERENCES Anayiotos, G. , and J. de PiniÃ©s, 1990, "The Secondary Market and the International Debt Problem," World Development, 18, 1655-1669. Banco del PacÃfico, 1994a, "Economic Report," 16-5 (September-October), 1-4. Banco del PacÃfico, 1994b, "Economic Report," 16-6 (November-December), 1-4. Banco del PacÃfico, 1995, "Economic Report," 17-2 (March- April), 1-4. Barger, H., and L.R. Klein, 1954, "A Quarterly Model for the United States Economy," Journal of the American Statistical Association, 49, 413-437. Batten, D.S., and M.T. Belongia, 1986, "Monetary Policy, Exchange Rates, and U.S. Agricultural Exports," American Journal of Agricultural Economics, 68, 422-427. BeltrÃ¡n del Rio, A., 1991, "Macroeconometric Model- Building of Latin American Countries, 1965-1985," Chapter 12 in A History of Macroeconometric Model-Building, Edited by R.G. Bodkin, L.R. Klein, and K. Marwah, Brookfield, VT: Edward Elgar Publishing Company. Bomberger, W.A., and G.E. Makinen, 1979, "Some Further Tests of the Harberger Inflation Model using Quarterly Data," Economic Development and Cultural Change, 27, 629-644. Box, G.E.P., and G.M. Jenkins, 1976, Time Series Analysis: Forecasting and Control, San Francisco: Holden-Day. Box, G.E.P., and G.C. Tiao, 1975, "Intervention Analysis with Applications to Economic and Environmental Problems," Journal of the American Statistical Association, 70, 70-79. Brajer, V., 1992, "Empirical Inflation Models in Developing Countries," Journal of Development Studies, 28, 717-729. 84 85 Cabrera, M., and F. Montes, 1978, "La DinÃ¡mica de las Series de Tiempo: El Caso del Indice Nacional de Precios al Consumidor-Empleados," Revista del Banco de la RepÃºblica, 51, 1122-1142. Chatfield, C., 1984, The Analysis of Time Series, New York: Chapman and Hall. Christ, C.F., 1993, "Assessing Applied Econometric Results," Federal Reserve Bank of St. Louis Review, 75 (March/April), 71-94. Christ, C.F., 1994, "The Cowles Commission's Contributions to Econometrics at Chicago, 1939-1955," Journal of Economic Literature, 32, 30-59. Clemen, R.T., 1989, "Combining Forecasts: A Review and Annotated Bibliography," International Journal of Forecasting, 5, 559-583. Clavijo, S., 1987, "Hacia una CaracterizaciÃ³n del Comportamiento de la Velocidad de CirculaciÃ³n del Dinero," Ensayos Sobre PolÃtica EconÃ³mica, 12, 59-79. Dhane, G., and A.P. Barten, 1989, "When it all Began: The 1936 Tinbergen Model Revisited," Economic Modelling, 6, 203- 219. DirecciÃ³n General de Estudios, various issues, InformaciÃ³n EstadÃstica Mensual, Banco Central del Ecuador: Quito. Dutton, J., and T. Grennes, 1987, "Alternative Measures of Effective Exchange Rates for Agricultural Trade," European Review of Agricultural Economics, 14, 427-442. Eaton, J., and M. Gersovitz, 1981, "Debt with Potential Repudiation: Theoretical and Empirical Analysis," Review of Economic Studies, 48, 289-309. Edwards, S., 1985, "Money, the Rate of Devaluation, and Interest Rates in a Semiopen Economy: Colombia, 1968-82," Journal of Money, Credit, and Banking, 17, 59-68. Engsted, T. , 1993, "Cointegration and Cagan's Model of Hyperinflation under Rational Expectations," Journal of Money, Credit and Banking, 25, 350-360. Feige, E.L., and D.K. Pearce, 1979, "The Casual Causal Relationship between Money and Income: Some Caveats for Time Series Analysis," Review of Economics and Statistics, 61, 521- 533. 86 Friscia, A.B., 1993, "Comment on Fullerton," Atlantic Economic Journal, 21 (March), 87. Fullerton, T.M., Jr., 1989a, "A Composite Approach to Forecasting State Government Revenues," International Journal of Forecasting, 5, 373-380. Fullerton, T.M., Jr., 1989b, "Exchange Rate Indexes for Ecuador and Venezuela," Latin America Economic Outlook, The WEFA Group (June), 1.3-1.7. Fullerton, T.M., Jr., 1989c, "Ecuador," Latin America Economic Outlook, The WEFA Group (September), 6.1-6.15. Fullerton, T.M., Jr., 1990a, "Ecuador," Latin America Economic Outlook, The WEFA Group (June), 6.1-6.15. Fullerton, T.M., Jr., 1990b, "Colombia," Latin America Economic Outlook, The WEFA Group (March), 5.1-5.16. Fullerton, T.M., Jr., 1990c, "Venezuela," Latin America Economic Outlook, The WEFA Group (September), 10.1-10.14. Fullerton, T.M., Jr., 1990d, "Methodologies for Evaluating Secondary Market Debt Prices," paper presented at The WEFA Group Sovereign Debt Conference, New York, NY. Fullerton, T.M., Jr., 1991, "Colombia," Latin America Economic Outlook, The WEFA Group (March), 5.1-5.16. Fullerton, T.M., Jr., 1992, "Confessions of an International Forecaster," Business Economics, 27 (April), 61-65. Fullerton, T.M. , Jr., 1993a, "Un Modelo MacroeconomÃ©trico para Pronosticar la EconomÃa Colombiana," Ensayos Sobre PolÃtica EconÃ³mica, 24, 101-136. Fullerton, T.M., Jr. , 1993b, "Un Modelo MacroeconomÃ©trico para Pronosticar la EconomÃa Ecuatoriana," Cuestiones EconÃ³micas, 20, 59-100. Fullerton, T.M., Jr. , 1993c, "Inflationary Trends in Colombia," Journal of Policy Modeling, 15, 463-468. Fullerton, T.M., Jr., 1993d, "Predictability of Secondary-Market Developing-Country Debt Prices," Applied Economics, 25, 1365-1371. 87 Fullerton, T.M., Jr., 1995a, "Short-Term Price Movements in Ecuador," Proceedings of the American Statistical Association, Business and Economic Statistics Section, 280- 285. Fullerton, T.M., Jr., 1995b, "Empirical Models for Secondary Market Debt Prices," paper presented at 65th Annual Meeting of the Southern Economic Association, New Orleans, LA. Fullerton, T.M., Jr., I. Fainboim, and M.I. Agudelo, 1992, "Exchange Rate, Money Supply, and Inflationary Dynamics in Colombia," Best Paper Proceedings of the Atlantic Economic Society, 2 (January), 54-58. Fullerton, T.M. , Jr., R. Hirth, and M. Smith, 1991, "Inflationary Dynamics and the Angell-Johnson Proposals," Atlantic Economic Journal, 19 (March), 1-14. Fullerton, T.M., Jr., and A. Kapur, 1991, "PredicciÃ³n de Multiplicadores Monetarios en Colombia, Ecuador, y Venezuela," Lecturas de EconomÃa, 35, 53-86. Gajdeczka, P. , and M.R. Stone, 1990, "The Secondary Market for Developing Country Loans," Finance and Development (December), 22-25. Garlow, D.C., 1993, "Ecuador," Latin America Economic Outlook, The WEFA Group (September), 6.1-6.17. Gennotte, G., H.J. Kharas, and S.Sadeq, 1987, "A Valuation Model for Developing-Country Debt with Endogenous Rescheduling," World Bank Economic Review, 1, 237-271. Granger, C.W.J., 1992, "Forecasting Stock Market Prices, " International Journal of Forecasting, 8, 3-13. Hakkio, C., and M. Rush, 1991, "Cointegration: How Short is the Long-Run?" Journal of International Money and Finance, 10, 571-581. Hallman, J.J., R.D. Porter, and D.H. Small, 1991, "Is the Price Level Tied to the M2 Monetary Aggregate in the Long Run?" American Economic Review, 81, 841-858. Hamilton, J.D., 1994, Time Series Analysis, Princeton, NJ: Princeton University Press. Hanson, J.A., 1985, "Inflation and Imported Input Prices in some Inflationary Latin American Economies," Journal of Development Economics, 18, 395-410. 88 Harberger, A., 1963, "The Dynamics of Inflation in Chile," in Measurement in Economics: Studies in Mathematical Economics and Econometrics in Memory of Yehuda Grunfeld, edited by C.F. Christ, Stanford, CA: Stanford University Press. Ikhide, S., and T.M. Fullerton, Jr., 1995, "Inflationary Dynamics in Nigeria," paper presented at Institute for Developing Economies, Tokyo, Japan. Kamas, L., 1986, "Dutch Disease Economics and the Colombian Export Boom," World Development, 14, 1177-1198. Kamas, L., 1995, "Monetary Policy and Inflation under the Crawling Peg: Some Evidence from VARs for Colombia," Journal of Development Economics, 46, 145-161. Kercheval, M.P., 1987, "Another Look at the Dollar's Trade-Weighted Value," Business Economics, 22 (April), 43-47. Kim, C., and T.M. Fullerton, Jr., 1996, "Import Prices, Wage Rates, and Korean Price Dynamics," working paper, Department of Economics, Ajou University, Suwon, Korea. Klein, L.R., 1984, "The Importance of the Forecast," Journal of Forecasting, 3, 1-9. Koch, P. , J. Rosensweig, and J. Whitt, 1988, "The Dynamic Relationship between the Dollar and U.S. Prices," Journal of International Money and Finance, 7, 181-204. Laidler, D., 1993, "Commentary on Assessing Applied Econometric Results," Federal Reserve Bank of St. Louis Review, 75 (March/April), 101-102. Laney, L.O., 1987, "The Secondary Market in Developing Country Debt," Federal Reserve Bank of Dallas Economic Review, (July), 1-12. Leiderman, L., 1984, "On the Monetary-Macro Dynamics of Colombia and Mexico", Journal of Development Economics, 14, 183-201. Leith, J.C., 1991, "The Exchange Rate and the Price Level in a Small Open Economy," Journal of Policy Modeling, 13, 309- 315. Montes, G., and R. CÃ¡ndelo, 1982, "El Enfoque Monetario de la Balanza de Pagos: El Caso de Colombia, 1968-1980," Revista de PlaneaciÃ³n y Desarrollo, 14, 11-40. 89 Motley, B., 1993, "Inflation and Growth," FRBSF Weekly Letter, 93-44 (December 31), 1-3. Ocampo, J.A., 1983, "En Defensa de la Continuidad del Regimen Cambiarlo," Coyuntura EconÃ³mica, 13, 198-214. Ostle, B., and R.W. Mensing, 1975, Statistics in Research, Ames: Iowa State University Press. Pagan, A.R., 1974, "A Generalized Approach to the Treatment of Autocorrelation," Australian Economic Papers, 13, 267-280. Palma, P., and D. Fontiveros, 1988, "A Comparative Sensitivity Analysis of the MODVEN VII Macroeconomic Model for Venezuela," Economic Modelling, 5, 286-345. Pankratz, A., 1983, Forecasting with Univariate Box- Jenkins Models, New York: John Wiley and Sons. Perasso, G. , 1989, "The Pricing of LDC Debt in the Secondary Market," Kyklos, 42, 533-555. Purcell, J.F.H., and D.J. Orlanski, 1988, "Developing Country Loans: A New Valuation Model for Secondary Market Trading," International Loan Trading Analysis, Salomon Brothers, 1-20. Purcell, J.F.H., and D.J. Orlanski, 1989, "Developing Country Loans: Refinements to the Relative Value Model for the Secondary Market," International Loan Trading Analysis, Salomon Brothers, 1-16. Rahnama-Moghadam, M., and Hedayeh Samavati, 1991, "Predicting Debt Rescheduling by Less-Developed Countries," Quarterly Review of Economics and Business, 31, 3-14. Rahnama-Moghadam, M., H. Samavati, and L.J. Haber, 1991, "The Determinants of Debt Rescheduling," Southern Economic Journal, 58, 510-517. Sachs, J., and H. Huizinga, 1987, "US Commercial Banks and the Developing-Country Debt Crisis," Brookings Papers on Economic Activity, 2, 555-601. Sheehey, E.J., 1976, "The Dynamics of Inflation in Latin America: Comment," American Economic Review, 66, 692-694. Sheehey, E.J., 1980, "Money, Income, and Prices in Latin America," Journal of Development Economics, 7, 345-357. 90 Snedecor, G., 1956, Statistical Methods, Ames: Iowa State University Press. Snowden, P.N., 1989, "The Interpretation of Market Discounts in Assessing Buy-Back Benefits," IMF Staff Papers, 36, 733-735. Stone, M.R., 1991, "Are Sovereign Debt Secondary Market Returns Sensitive to Macroeconomic Fundamentals?," Journal of International Money and Finance, 10, S100-S122. Vial, J., 1988, "ComparaciÃ³n de Modelos MacroeconomÃ©tricos Latinoamericanos," ColecciÃ³n Estudios CIEPLAN, 24, 145-181. Vogel, R.C., 1974, "The Dynamics of Inflation in Latin America, 1950-1969," American Economic Review, 64, 102-114. Wakeman-Linn, J., 1991, "The Market for Developing Country Debt," The World Bank Research Observer, 6, 191-203. West, C.T., 1996, "System-based Weights vs. Series- specific Weights in the Combination of Forecasts," paper presented at the 42nd Regional Science Association International North American Meetings, Cincinnati, OH. Zellner, A., 1994, "Time Series Analysis, Forecasting, and Econometric Modeling: The Structural Econometric Modeling, Time Series Analysis (SEMSTA) Approach," Journal of Forecasting, 13, 215-233. Zellner, A., and H. Theil, 1962, "Three-Stage Least Squares: Simultaneous Estimation of Simultaneous Relations," Econometrica, 30, 54-78. Zind, R.G., 1993, "On Inflation and Growth in the LDCs," Economic Studies Quarterly, 44, 108-116. BIOGRAPHICAL SKETCH Thomas M. Fullerton, Jr. is a senior economist in the Forecasting Program of the Bureau of Economic and Business Research at the University of Florida. Fullerton is co-author of The Florida Outlook, a quarterly forecast of the state and 20 metropolitan economies. He also teaches a course on Latin American political economy. Fullerton previously worked at Wharton Econometrics as international economist in charge of modeling, forecasting, and policy analysis for Colombia, Ecuador, and Venezuela. He also worked as an economist in the Executive Office of the Governor of Idaho where he forecast the state economy and conducted fiscal policy analysis during legislative sessions. He began his career in the Planning Department of El Paso Electric Company. His research has been published in outlets such as Applied Economics, Journal of Forecasting, Public Budgeting & Finance, Atlantic Economic Journal, Journal of Policy Modeling, Business Economics, Applied Economics Letters, and International Journal of Forecasting. Fullerton holds degrees from the University of Texas at El Paso, Iowa State University, and the Wharton School of the University of Pennsylvania. He is a doctoral candidate in economics at the University of Florida. 91 I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the deqree of Doctor of Philosophy. Carol Taylor West, Chair Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David A. Denslow Distinguished Service Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Chunrong Ai v~>' Assistant Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. u Terry L. McCoy Professor of Latin American Studies This dissertation was submitted to the Graduate Faculty of the Department of Economics in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August 1996 Karen A. Holbrook Dean, Graduate School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f 7KHVH UHODWLYHO\ VPDOO PDFURHFRQRPHWULF PRGHOV RI WKH 'XWFK HFRQRP\ ZHUH GHYHORSHG XVLQJ DQQXDO GDWD 5HVHDUFK RQ EXVLQHVV F\FOHV LQ WKH 8QLWHG 6WDWHV EHJDQ UHSOLFDWLQJ DQG H[WHQGLQJ WKH 'XWFK PRGHOLQJ H[DPSOHV LQ WKH V ,QFUHDVLQJ LQWHUHVW LQ VKRUWn WHUP HFRQRPLF IOXFWXDWLRQV HYHQWXDOO\ OHG WR WKH GHYHORSPHQW RI TXDUWHUO\ IRUHFDVWLQJ PRGHOV %DUJHU DQG .OHLQ f $ORQJ ZLWK WKH H[SDQVLRQ RI VWUXFWXUDO HFRQRPHWULF PRGHOLQJ DQG UHVHDUFK ZLWK UHVSHFW WR IRUHFDVWLQJ DQG SROLF\ DQDO\VLV WLPH VHULHV VWDWLVWLFV DOVR EHFDPH LQFUHDVLQJO\ VRSKLVWLFDWHG LQ WKH UHDOP RI SUHGLFWLYH PRGHOLQJ 0RVW RI WKHVH HIIRUWV RFFXUUHG ZLWK UHVSHFW WR KLJK IUHTXHQF\ PRQWKO\ GDWD HVSHFLDOO\ XQLYDULDWH DXWRUHJUHVVLYH PRYLQJ DYHUDJH PRGHOV %R[ DQG -HQNLQV f :KLOH RIWHQ UHJDUGHG DV FRPSHWLWRUV WLPH VHULHV PRGHOV DUH IUHTXHQWO\ XWLOL]HG DV FRPSOHPHQWV WR HFRQRPHWULF HTXDWLRQ VHWV DQG FDQ DOVR EH LPEHGGHG ZLWKLQ D YDULHW\ RI IRUHFDVWLQJ V\VWHPV &OHPHQ )XOOHUWRQ D =HOOQHU :HVW f PAGE 8 /DWLQ $PHULFDQ PDFURHFRQRPLF IRUHFDVWLQJ PRGHOV EHJDQ WR DSSHDU LQ WKH V %HOWUÂ£Q GHO 5LR f 6LPLODU WR WKH ILUVW PRGHOV IRU WKH 1HWKHUODQGV DQG WKH 8QLWHG 6WDWHV WKH HDUO\ /DWLQ $PHULFDQ PRGHOV XWLOL]HG V\VWHPV RI VLPXOWDQHRXV HTXDWLRQV GHVLJQHG DURXQG QDWLRQDO LQFRPH DQG SURGXFW DFFRXQW 1,3$f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Â£Q GHO 5LR f WKH &,(3/$1 PRGHO IRU &KLOH 9LDO f WKH 0HWURHFRQPLFD PRGHO IRU 9HQH]XHOD 3DOPD DQG )RQWLYHURV f DQG WKH :()$ PRGHOV IRU &RORPELD DQG (FXDGRU )XOOHUWRQ D Ef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f 7KLV LV QRW WR LPSO\ WKDW WKH WUDGLWLRQDO PDFURHFRQRPHWULF /DWLQ $PHULFDQ PRGHOV DUH UHJDUGHG DV XVHOHVV 0RVW DQDO\VHV RI LQWHUQDWLRQDO LQGHEWHGQHVV W\SLFDOO\ UHO\ RQ DQQXDO GDWD LQ RUGHU WR H[DPLQH WKH FRQVHTXHQFHV RI FKDQJHV LQ UHJLRQDO EDODQFH RI SD\PHQW IDFWRUV 1RW VXUSULVLQJO\ VLJQLILFDQW HIIRUW ZDV SXW IRUWK LQ UHFHQW \HDUV WR HQKDQFH WKH FXUUHQW DFFRXQWFDSLWDO DFFRXQW OLQNDJHV DQG VLPXODWLRQ SHUIRUPDQFHV LQ /DWLQ $PHULFDQ PDFURHFRQRPLF PRGHOV )XOOHUWRQ Df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f PRGHOLQJ IUDPHZRUN WR DQDO\]H PRYHPHQWV LQ FRQVXPHU SULFHV LQ &RORPELD )RUHFDVWLQJ H[SHULPHQWV DUH FRQGXFWHG ZLWK WKH UHVXOWLQJ PRGHO WR VKHG OLJKW RQ SRWHQWLDO LPSDFWV DVVRFLDWHG ZLWK DQ DQWL LQIODWLRQDU\ SURJUDP HQDFWHG E\ WKH FHQWUDO EDQN 5HVXOWV LQGLFDWH WKDW WKLV FODVV RI WLPH VHULHV PRGHOLQJ FDQ SURYLGH XVHIXO LQVLJKWV ZLWK UHVSHFW WR PDFURHFRQRPLF WUHQGV LQ /DWLQ $PHULFDQ FRXQWULHV $ UHYLVHG YHUVLRQ RI WKLV FKDSWHU ZDV SXEOLVKHG LQ WKH -RXUQDO RI 3ROLF\ 0RGHOLQJ )XOOHUWRQ Ff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f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f $GGLWLRQDO UHVHDUFK H[WHQGLQJ WKH LQLWLDO UHVXOWV RQ WKLV WRSLF ZDV SUHVHQWHG DW WKH WK DQQXDO 6RXWKHUQ (FRQRPLF $VVRFLDWLRQ FRQIHUHQFH )XOOHUWRQ Ef PAGE 12 &+$37(5 ,1)/$7,21$5< 75(1'6 ,1 &2/20%,$ ,QWURGXFWLRQ ,QIODWLRQ KDV ORQJ EHHQ RQH RI WKH PRVW VHULRXV HFRQRPLF SUREOHPV IDFLQJ SROLF\PDNHUV LQ /DWLQ $PHULFD $OWKRXJK &RORPELD KDV WUDGLWLRQDOO\ HQMR\HG ORZHU UDWHV RI LQIODWLRQ WKDQ QHLJKERULQJ 6RXWK $PHULFDQ FRXQWULHV LQ D SUHVLGHQWLDO HOHFWLRQ \HDU FRQVXPHU SULFHV URVH LQ H[FHVV RI SHUFHQW ,Q UHVSRQVH &RORPELDQ PRQHWDU\ DXWKRULWLHV LQWURGXFHG D VHULHV RI SROLF\ LQQRYDWLRQV GHVLJQHG WR ORZHU WKH LQIODWLRQ UDWH 0HDVXUHV DGRSWHG LQFOXGH D SURJUHVVLYH RSHQLQJ RI WKH HFRQRP\ WR JUHDWHU YROXPHV RI LQWHUQDWLRQDO WUDGH ILVFDO DXVWHULW\ WLJKWHU FUHGLW FRQWUROV DQG D VORZHU UDWH RI FXUUHQF\ GHYDOXDWLRQ IRU GHWDLOV VHH )XOOHUWRQ f 7KLV FKDSWHU H[DPLQHV VRPH RI WKH SRWHQWLDO UHVXOWV DVVRFLDWHG ZLWK WKH SULQFLSDO WRROV RI WKH SROLF\ DGMXVWPHQW HIIRUW 7LPH VHULHV FKDUDFWHULVWLFV RI WKH FRQVXPHU SULFH LQGH[ &3,f WKH PRQH\ VXSSO\ 0O GHILQHG DV FXUUHQF\ LQ FLUFXODWLRQ SOXV GHPDQG GHSRVLWVf DQG WKH SHVRGROODU H[FKDQJH UDWH 5(;f DUH LQYHVWLJDWHG 7R PHDVXUH WKH VKRUW UXQ UHODWLRQVKLSV DPRQJ WKH &3, 0O DQG 5(; WUDQVIHU PAGE 13 IXQFWLRQ DXWRUHJUHVVLYHPRYLQJ DYHUDJH $5,0$f DQDO\VLV LV DSSOLHG VHH %R[ DQG -HQNLQV FK f 6HOHFWLRQ RI WKH QRQOLQHDU PRGHO HVWLPDWLRQ PHWKRGRORJ\ ZDV PRWLYDWHG E\ WZR IDFWRUV )LUVW ZDV WKH XVHIXOQHVV RI WUDQVIHU IXQFWLRQ $5,0$ DQDO\VLV IRU VKRUWWHUP IRUHFDVWLQJ DSSOLFDWLRQV 6HFRQG ZDV LQWHUHVW H[SUHVVHG LQ D SUHYLRXV VWXG\ ZKLFK XWLOL]HG WKLV WHFKQLTXH WR DQDO\]H SULFH G\QDPLFV LQ WKH 8QLWHG 6WDWHV )XOOHUWRQ +LUWK DQG 6PLWK f 6SHFLILFDOO\ HFRQRPLVWV ZLWKLQ WKH 5HVHDUFK 'HSDUWPHQW RI WKH &HQWUDO %DQN LQ %RJRWÂ£ ZLVKHG WR VHH ZKHWKHU WKH VDPH DSSURDFK ZRXOG SURYH XVHIXO ZLWK UHVSHFW WR WKH &RORPELDQ HFRQRP\ 6XEVHTXHQW VHFWLRQV RI WKH SDSHU SUHVHQW D EULHI UHYLHZ RI WKH OLWHUDWXUH PHWKRGRORJ\ DQG HPSLULFDO UHVXOWV 3ROLF\ VLPXODWLRQ H[HUFLVHV GHVLJQHG WR UHIOHFW H[FKDQJH UDWH DQG PRQHWDU\ SROLFLHV LQ &RORPELD IROORZ $ VXPPDU\ DQG FRQFOXVLRQ IRUP WKH ILQDO VHFWLRQ RI WKH FKDSWHU 3UHYLRXV 5HVHDUFK ,Q RQH RI WKH HDUOLHVW WLPH VHULHV VWXGLHV GHYHORSHG IRU DQ HFRQRP\ LQ /DWLQ $PHULFD &DEUHUD DQG 0RQWHV f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Â£QGHOR f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f XWLOL]HV YHFWRU DXWRUHJUHVVLRQV WR DQDO\]H LQIODWLRQ LQ &RORPELD DQG 0H[LFR &KDQJHV LQ WKH UDWHV RI JURZWK LQ WKH HFRQRP\ DQG WKH PRQH\ VXSSO\ DUH LQFOXGHG LQ HDFK PRGHO ,Q WKH FDVH RI &RORPELD WKH UHVXOWV LQGLFDWH WKDW YDULDWLRQV LQ WKH UDWH RI FKDQJH RI 0O V\VWHPDWLFDOO\ DIIHFW WKH &3, EXW QRW WKH FRQYHUVH )URP D SROLF\ SHUVSHFWLYH WKLV LPSOLHV WKDW PRQHWDU\ DXWKRULWLHV GR QRW HQJDJH LQ DFFRPPRGDWLYH PHDVXUHV LQ UHVSRQVH WR SURGXFWLRQ DQG LQIODWLRQDU\ VKRFNV 7KLV UHVXOW PD\ VWHP IURP PAGE 15 WKH XVDJH RI PRQHWDU\ SROLF\ WR DFKLHYH JRDOV RWKHU WKDQ SULFH VWDELOLW\ 7KHVH LQFOXGH HFRQRPLF JURZWK DQG H[SRUW GLYHUVLILFDWLRQ )URP DQ HFRQRPHWULF VWDQGSRLQW WKLV UHVXOW LV DOVR LPSRUWDQW EHFDXVH LW LPSOLHV WKDW XQLGLUHFWLRQDO *UDQJHU FDXVDOLW\ H[LVWV EHWZHHQ 0O DQG WKH &3, LQ &RORPELD $ QXPEHU RI UHFHQW VWXGLHV KDYH H[DPLQHG LQIODWLRQDU\ G\QDPLFV LQ WKH 8QLWHG 6WDWHV .RFK 5RVHQVZHLJ DQG :KLWW f LQYHVWLJDWH WKH UHODWLRQVKLS EHWZHHQ WKH H[FKDQJH UDWH DQG FRQVXPHU SULFHV &URVV FRUUHODWLRQ IXQFWLRQV DUH XVHG WR VXJJHVW WKH QXPEHU RI ODJV WR EH LQFOXGHG LQ WKH UHJUHVVLRQ HTXDWLRQV *UDQJHU FDXVDOLW\ WHVWV LPSO\ D XQLGLUHFWLRQDO FKDQQHO RI LQIOXHQFH IURP WKH H[FKDQJH UDWH WR SULFHV 7KH LQIODWLRQDU\ LPSDFWV DVVRFLDWHG ZLWK D FKDQJH LQ WKH LQWHUQDWLRQDO YDOXH RI WKH GROODU DUH IRXQG WR H[WHQG PRUH WKDQ PRQWKV )XOOHUWRQ +LUWK DQG 6PLWK f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nV ZKHUH SDVVWKURXJK HIIHFWV VWDWLVWLFDOO\ VLJQLILFDQW UHODWLRQVKLSV EHWZHHQ LQWHUQDWLRQDO FXUUHQF\ YDOXHV DQG LQIODWLRQ DUH RIWHQ VWURQJ DQG UHODWLYHO\ JXLFN VHH /HLWK f (PSLULFDO HYLGHQFH UHSRUWHG HOVHZKHUH (GZDUGV f LQGLFDWHV WKDW PRYHPHQWV LQ WKH UDWH RI GHYDOXDWLRQ DUH KHOSIXO LQ PRGHOLQJ QRPLQDO LQWHUHVW UDWHV LQ &RORPELD 8QIRUWXQDWHO\ WKH ODWWHU VWXG\ GRHV QRW GLUHFWO\ WHVW WKH UHODWLRQVKLS EHWZHHQ SULFHV DQG WKH SHVRGROODU H[FKDQJH UDWH 0HWKRGRORJ\ 7KH PHWKRGRORJ\ XWLOL]HG LQ WKLV SDSHU LV VLPLODU WR WKH PXOWLSOH $5,0$ DSSURDFK DSSOLHG WR WKH 8QLWHG 6WDWHV E\ )XOOHUWRQ +LUWK DQG 6PLWK f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f 7KH JHQHUDO IRUP RI D XQLYDULDWH $5,0$ PRGHO HVWLPDWHG IRU WKH &3, FDQ EH ZULWWHQ DV IROORZV f 3W >4%f4V%f8W@ >+ %f +V %f @ ZKHUH 3W LV D VWDWLRQDU\ VHULHV GHULYHG IURP WKH RULJLQDO &3, VHULHV % LV D EDFNVKLIW RU ODJ RSHUDWRU 4%f LV D PRYLQJ DYHUDJH SRO\QRPLDO RI RUGHU J 46%f LV D VHDVRQDO PRYLQJ DYHUDJH SRO\QRPLDO RI RUGHU JV 8W LV WKH GLVWXUEDQFH WHUP +%f LV DQ DXWRUHJUHVVLYH SRO\QRPLDO RI RUGHU S DQG +6%f LV D VHDVRQDO DXWRUHJUHVVLYH SRO\QRPLDO RI RUGHU SV 7KH $50$ PRGHO IRU WKH &3, LV DOVR XVHG LQ WKH HVWLPDWLRQ RI WKH WUDQVIHU IXQFWLRQ HJXDWLRQV 7R H[DPLQH WKH HIIHFWV RI RWKHU YDULDEOHV RQ WKH &3, WUDQVIHU IXQFWLRQ $5,0$ PRGHOV DUH HVWLPDWHG 7KHVH PRGHOV DUH XVHG WR GHWHUPLQH LI WKH LQSXW VHULHV DUH LQFUHPHQWDOO\ XVHIXO LQ H[SODLQLQJ WKH YDULDWLRQ RI WKH &3, EH\RQG WKH LQIRUPDWLRQ REWDLQHG ZLWKLQ WKH SULFH LQGH[ LWVHOI VHH %R[ DQG -HQNLQV FK f 3ULRU WR HVWLPDWLQJ D WUDQVIHU IXQFWLRQ HJXDWLRQ FURVV FRUUHODWLRQ IXQFWLRQV &&)f DUH XVHG WR LQGLFDWH ZKLFK ODJV RI DQ LQSXW VHULHV PD\ FRQWULEXWH LQFUHPHQWDO LQIRUPDWLRQ WR WKH XQLYDULDWH $50$ PRGHO DOUHDG\ HVWLPDWHG IRU WKH &3, 5HVLGXDOV IURP WKH WKUHH XQLYDULDWH PAGE 18 HTXDWLRQV DUH XVHG WR FDOFXODWH &&)V EHWZHHQ WKH &3, DQG WKH RWKHU YDULDEOHV VHH &KDWILHOG S f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f 3W >:%f5W *%f0W 4%f4V%f8W@ >+%f+6%f@ ZKHUH WKH XQLYDULDWH PRGHO SUHVHQWHG LQ (TXDWLRQ LV DXJPHQWHG E\ LQFRUSRUDWLQJ JHQHUDO RUGHU SRO\QRPLDOV :%f DQG *%f IRU WKH UHVSHFWLYH LQSXW YDULDEOHV 5W DQG 0W UHSUHVHQW VWDWLRQDU\ LQSXW VHULHV GHULYHG IURP WKH H[FKDQJH UDWH DQG WKH PRQH\ VXSSO\ GDWD GLVFXVVHG DERYH %HFDXVH WKH DQDO\VLV LV FRQGXFWHG ZLWKLQ D G\QDPLF IUDPHZRUN FRHIILFLHQW UHVWULFWLRQV DUH QRW K\SRWKHVL]HG EXW WKH VXPV RI WKH FRHIILFLHQWV DVVRFLDWHG ZLWK HDFK LQGLYLGXDO LQSXW VHULHV DUH H[SHFWHG WR EH SRVLWLYH PAGE 19 (VWLPDWLRQ 5HVXOWV 'DWD XVHG LQ WKLV FKDSWHU DUH IURP WKH FHQWUDO EDQN LQ %RJRWÂ£ 0RQWKO\ REVHUYDWLRQV IRU DOO WKUHH VHULHV DUH SXEOLVKHG LQ 5HYLVWD GHO %DQFR GH OD 5HSEOLFD IRU H[DPSOH VHH &DEUHUD DQG 0RQWHV f 7KH VDPSOH SHULRG -DQXDU\ WKURXJK 'HFHPEHU FRUUHVSRQGV WR WKH FUDZOLQJ SHJ HUD RI H[FKDQJH UDWH GHWHUPLQDWLRQ LQ &RORPELD 6LPLODU WR &DEUHUD DQG 0RQWHV f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f 60$f VSHFLILFDWLRQ HPSOR\HG E\ &DEUHUD DQG 0RQWHV S f WR PRGHO WKH &3, 3DUDPHWHU HVWLPDWHV UHSRUWHG LQ 0RGHO FDUU\ WKH VDPH VLJQV DQG DUH VLPLODU LQ PDJQLWXGH WR WKRVH HVWLPDWHG LQ WKH SUHYLRXV VWXG\ YHUVXV DQG YHUVXV f (JXDWLRQ LV LQFRUSRUDWHG LQ WKH HVWLPDWLRQ RI WKH WUDQVIHU IXQFWLRQ HTXDWLRQV PAGE 20 7DEOH 8QLW 5RRW 7HVWV IRU 6WDWLRQDULW\ 6HULHV $XJ 'LFNH\)XOOHU WVWDW 0DF.LQQRQ FULW YDOXH 3 ZLWK FRQVW WUHQG ODJf b OYOf 5 ZLWK FRQVW WUHQG ODJf 0 ZLWK FRQVW WUHQG ODJf PAGE 21 7DEOH 8QLYDULDWH $5,0$ 0RGHOV 0RGHO 3DUDPHWHUV &3, SW r3W r8WB f f f 4f 5(; 5W r5W r8WB f f f 4f 0O 0W r8W f r8W f f 4f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f %HFDXVH PRQHWDU\ SROLF\ LQ &RORPELD DSSHDUV WR EH FRQGXFWHG LQGHSHQGHQWO\ RI LQIODWLRQDU\ VKRFNV WKH WUDQVIHU IXQFWLRQ PHWKRGRORJ\ FDQ EH XVHG WR PHDVXUH WKH HIIHFW RI WKH PRQH\ VXSSO\ RQ LQIODWLRQ 6LPLODU WR UHVXOWV UHSRUWHG IRU WKH 8QLWHG 6WDWHV .RFK 5RVHQVZHLJ DQG :KLWW f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f PAGE 23 7DEOH *UDQJHU &DXVDOLW\ 7HVWV &DXVDOLW\ 'LUHFWLRQ 1XPEHU RI /DJV &RPSXWHG )VWDW &3, 5(; &3, 5(; &3, 0O &3, 0O 'HJUHHV RI IUHHGRP IRU WKH UHJUHVVLRQV ZLWK ODJV IRU WKH QXPHUDWRU DQG IRU WKH GHQRPLQDWRU 'HJUHHV RI IUHHGRP IRU WKH UHJUHVVLRQV ZLWK ODJV IRU WKH QXPHUDWRU DQG IRU WKH GHQRPLQDWRU PAGE 24 7KHUH KDYH DOVR EHHQ SHULRGV ZKHQ ORFDO SULFH FKDQJHV KDYH PRYHG LQ FRQFHUW ZLWK WKRVH RI &RORPELDn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Â£QGHOR f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f WKH UHVXOWV HQFRXQWHUHG LQ WKLV FKDSWHU DUH XQH[SHFWHG DQG IXUWKHU UHVHDUFK LV ZDUUDQWHG $QRWKHU SRVVLELOLW\ LV WKDW 0O PD\ QRW EH WKH FRUUHFW PRQH\ VWRFN PHDVXUH ZLWK UHVSHFW WR &RORPELDQ SULFH G\QDPLFV 6WXGLHV FRPSOHWHG IRU RWKHU HFRQRPLHV LQGLFDWH WKDW EURDGHU OLTXLGLW\ DJJUHJDWHV VXFK DV 0 PD\ EH XVHIXO +DOOPDQ 3RUWHU DQG PAGE 26 7DEOH $5,0$ 7UDQVIHU )XQFWLRQV 0RGHO 3DUDPHWHUV &3, 3W r3W f f r5WB f 4 f &3, 3W r3W f f r0W f 4f &3, 3W f r3W f r5WB f r0W f 4f r8W f 22r5W f r8W f r8W f 22r5W f 7KH VDPSOH SHULRG LV -DQXDU\ 'HFHPEHU 1XPEHUV LQ SDUHQWKHVHV DUH FRPSXWHG WVWDWLVWLFV /MXQJ%R[ 4VWDWLVWLFV FDOFXODWHG IRU ODJV DUH UHSRUWHG PAGE 27 6PDOO DQG ,NKLGH DQG )XOOHUWRQ f 8QIRUWXQDWHO\ PRQH\ VXSSO\ HVWLPDWHV RWKHU WKDQ 0O FXUUHQWO\ GR QRW H[LVW IRU &RORPELD 7KLV SUREOHP LV IDLUO\ ZLGHVSUHDG LQ 6RXWK $PHULFD DQG KDV ORQJ SRVHG GLIILFXOWLHV LQ WKH DQDO\VLV RI PRQHWDU\ HFRQRPLFV LQ WKH UHJLRQ VHH )XOOHUWRQ DQG .DSXU f 3ROLF\ 6LPXODWLRQ 5HVXOWV 7KH *DYLULD DGPLQLVWUDWLRQ DQQRXQFHG LQ 'HFHPEHU WKDW LW ZRXOG DWWHPSW WR ORZHU WKH LQIODWLRQ UDWH WR SHUFHQW LQ RQO\ WZHOYH PRQWKV IRU GLVFXVVLRQ VHH )XOOHUWRQ f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f 6LPXODWLQJ 0RGHO XQGHU WKHVH SROLF\ DVVXPSWLRQV \LHOGV VHYHUDO LQWHUHVWLQJ UHVXOWV WKDW DUH UHSURGXFHG LQ 7DEOH 'XULQJ WKH ILUVW VL[ PRQWKV WKH LQIODWLRQ UDWH RVFLOODWHV QHDU SHUFHQW 6XEVHTXHQWO\ WKH PRQWK UDWH RI FKDQJH LQ WKH &3, GHFOLQHV VKDUSO\ WR SHUFHQW E\ \HDU HQG 'XULQJ WKH QH[W PRQWKV GLVLQIODWLRQ VXEVLGHV DV WKH VHYHULW\ RI WKH FXWEDFNV LQ WKH UDWHV RI FKDQJH RI ERWK LQSXW YDULDEOHV LV PRGHUDWHG PAGE 29 ,PPHGLDWH ,PSOHPHQWDWLRQ 3ROLF\ 6LPXODWLRQ 5HVXOWV 3&+<$&3,f 3&+<$5(;f 3&+<$0Of PAGE 30 5HVXOWV RI WKH VHFRQG VLPXODWLRQ H[HUFLVH DUH UHSRUWHG LQ 7DEOH 7KH PRUH UDSLG UDWH RI GHSUHFLDWLRQ DOORZV WKH PRQWK LQIODWLRQ UDWHV WR UHPDLQ DERYH SHUFHQW WKURXJKRXW WKH ILUVW VHPHVWHU RI WKH WHVW SHULRG 6WHDG\ GHFOLQHV LQ WKH UDWHV RI FKDQJH IRU ERWK 5(; DQG 0O IRUFH LQIODWLRQ WR GHFOLQH QRWLFHDEO\ GXULQJ WKH QH[W VL[ PRQWKV HYHQWXDOO\ UHDFKLQJ SHUFHQW PAGE 31 *UDGXDO ,PSOHPHQWDWLRQ 3ROLF\ 6LPXODWLRQ 5HVXOWV 3&+<$&3,f 3&+<$5(;f 3&+<$0Of PAGE 32 1R ,PSOHPHQWDWLRQ 3ROLF\ 6LPXODWLRQ 5HVXOWV 3&+<$&3,f 3&+<$5(;f 3&+<$0Of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f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f 1RW ZLWKVWDQGLQJ JRYHUQPHQW DVVXUDQFHV WKDW LQIODWLRQ ZRXOG GHFHOHUDWH WR LWV WDUJHW UDWH E\ \HDUHQG YHU\ OLWWOH HFRQRPHWULF DQDO\VLV XVLQJ VKRUWWHUP IRUHFDVWLQJ PHWKRGV DSSHDUV WR KDYH EHHQ UHOLHG XSRQ LQ GHYHORSLQJ WKH QHZ SROLF\ WDUJHWV ,Q LWV DWWHPSW WR VORZ SULFH PRYHPHQWV WKH 'XUÂ£Q %DOOHQ 'DKLN *DU]R]L DGPLQLVWUDWLRQ LQWURGXFHG D YDULHW\ RI QHZ SROLF\ PHDVXUHV 7KH\ LQFOXGH LPSRUW OLEHUDOL]DWLRQ ILVFDO DXVWHULW\ DQG D VORZHU UDWH RI FXUUHQF\ GHSUHFLDWLRQ %\ UHGXFLQJ SULFH SUHVVXUHV WKH JRYHUQPHQW KRSHV WR LPSURYH HFRQRPLF ZHOIDUH E\ HQDEOLQJ WKH (FXDGRULDQ HFRQRP\ WR RSHUDWH PAGE 37 PRUH HIILFLHQWO\ 7KLV DUJXPHQW LV YHU\ VLPLODU WR WKRVH DLUHG LQ DGYDQFHG HFRQRPLHV VXFK DV WKH 8QLWHG 6WDWHV 0RWOH\ f DQG DQDO\]HG LQ RWKHU GHYHORSLQJ QDWLRQV =LQG f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f 6XEVHTXHQW VHFWLRQV RI WKLV FKDSWHU RIIHU D UHYLHZ RI WKH OLWHUDWXUH WKHRUHWLFDO PRGHO DQG HPSLULFDO UHVXOWV 6XJJHVWLRQV IRU IXWXUH UHVHDUFK DUH VXPPDUL]HG LQ WKH FRQFOXVLRQ /LWHUDWXUH 5HYLHZ 7KH VHPLQDO UHVHDUFK RQ LQIODWLRQDU\ G\QDPLFV LQ GHYHORSLQJ FRXQWULHV ZDV FRQGXFWHG RQ &KLOHDQ GDWD E\ +DUEHUJHU f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f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f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f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f 7KH PRGHO DOVR LPSOLHV WKH HODVWLFLW\ RI LQIODWLRQ ZLWK UHVSHFW WR PRQH\ PAGE 40 JURZWK LV OHVV WKDQ RQH (PSLULFDO UHVXOWV LQ WKH +DQVRQ DUWLFOH VWURQJO\ VXSSRUW WKH LQFOXVLRQ RI LPSRUW SULFHV RU WKH UDWH RI GHYDOXDWLRQ LQ PRGHOV RI LQIODWLRQ 6XEVHTXHQW UHVHDUFK KDV SURYLGHG DGGLWLRQDO HYLGHQFH LQ IDYRU RI WKH DXJPHQWHG +DUEHUJHU+DQVRQ DSSURDFK ZKHUHLQ WKH HIIHFW RI LPSRUW SULFHV RQ LQIODWLRQ LV FRQVLGHUHG .RFK 5RVHQVZHLJ DQG :LWW f DQG )XOOHUWRQ +LUWK DQG 6PLWK f ERWK UHSRUW SRVLWLYH OLQNDJHV EHWZHHQ WKH WUDGH ZHLJKWHG H[FKDQJH YDOXH RI WKH GROODU DQG FRQVXPHU SULFHV LQ WKH 8QLWHG 6WDWHV 7KHVH HPSLULFDO VWXGLHV LQGLFDWH D XQLGLUHFWLRQDO FKDQQHO RI LQIOXHQFH IURP WKH H[FKDQJH UDWH WR GRPHVWLF SULFHV H[LVWV LQ WKH 8QLWHG 6WDWHV HFRQRP\ $V ZLOO EH GLVFXVVHG EHORZ FDXVDOLW\ GLUHFWLRQ KDV LPSRUWDQW LPSOLFDWLRQV IRU ERWK PRGHO IRUP DQG HVWLPDWLRQ WHFKQLTXH 'HYHORSLQJ FRXQWU\ VWXGLHV KDYH DOVR FRQILUPHG WKH XVHIXOQHVV RI DQ DXJPHQWHG PRGHOLQJ WUHDWPHQW RI LQIODWLRQDU\ G\QDPLFV 6KHHKH\ f UHSRUWV VRPH RI WKH HDUO\ HFRQRPHWULF ZRUN DORQJ WKHVH OLQHV 6KHHKH\ f UHDFKHV DGGLWLRQDO FRQFOXVLRQV RQ WKH EDVLV RI HPSLULFDO WHVWV WKDW LQGLFDWH WKDW DFFXUDWH DVVHVVPHQW RI DXVWHULW\ SROLF\ HIIRUWV ZLOO OLNHO\ UHTXLUH H[SODQDWRU\ YDULDEOHV UHSUHVHQWLQJ FRVW SXVK IDFWRUV 0RUH UHFHQWO\ %UDMHU f SURYLGHV HYLGHQFH WKDW WKH ODWWHU FDWHJRU\ RI PRGHOV PD\ RIIHU EHWWHU VSHFLILFDWLRQV WKDQ WKRVH ZKLFK UHO\ VROHO\ RQ GRPHVWLF HFRQRPLF IDFWRUV &RQFOXVLRQV LQ WKDW DUWLFOH DUH UHDFKHG RQ WKH EDVLV RI )WHVWV IRU GLIIHUHQW UHJUHVVRU VHWV 6LPLODUO\ PAGE 41 )XOOHUWRQ Ef VXFFHVVIXOO\ LPEHGV D YDULDQW RI WKLV DSSURDFK LQ D ODUJHVFDOH PDFURHFRQRPHWULF IRUHFDVWLQJ PRGHO IRU (FXDGRU XVLQJ DQQXDO GDWD 7KHUH KDYH EHHQ YHU\ IHZ G\QDPLF PRGHOV HVWLPDWHG RQ WKH EDVLV RI PRQWKO\ GDWD IRU GHYHORSLQJ HFRQRPLHV *LYHQ WKDW PRVW EXVLQHVV GHFLVLRQV LQ KLJKO\ LQIODWLRQDU\ FRXQWULHV DUH UHDFKHG ZLWKLQ D VKRUWUDQJH FRQWH[W WKLV LV DQ DUHD ZKLFK QHHGV WR EH DGGUHVVHG $V GHWDLOHG LQ &KDSWHU )XOOHUWRQ Ff HPSLULFDOO\ H[DPLQHV &RORPELDQ DQWLLQIODWLRQDU\ HIIRUWV XWLOL]LQJ PRQWKO\ GDWD ZLWK DQ $5,0$ WUDQVIHU IXQFWLRQ 7KH HVWLPDWHG PRGHO LV IRXQG WR JHQHUDWH UHDOLVWLF VLPXODWLRQ VFHQDULRV IRU SROLF\ DQDO\VLV 7KH UHVXOWV DOVR VXSSRUW WKH K\SRWKHVLV RI LQIODWLRQ UDWH LQHODVWLFLW\ ZLWK UHVSHFW WR PRQHWDU\ JURZWK $V LQ WKH &KLOHDQ HTXDWLRQ UHSRUWHG E\ +DQVRQ f DQG WKH $UJHQWLQH PRGHO SUHVHQWHG LQ 6KHHKH\ f H[FKDQJH UDWH SULFH HIIHFWV DUH IRXQG WR RXWZHLJK WKH PRQHWDU\ FRHIILFLHQW 7KH ODWWHU LV VRPHZKDW VXUSULVLQJ JLYHQ WKDW LPSRUWHG JRRGV DQG VHUYLFHV FRPSULVH OHVV WKDQ SHUFHQW RI &RORPELDQ JURVV GRPHVWLF SURGXFW 7KHRUHWLFDO 0RGHO +DUEHUJHUnV f PRGHO LV EDVHG RQ WKH WUDGLWLRQDO TXDQWLW\ WKHRU\ RI PRQH\ HTXDWLRQ f 09 34 PAGE 42 ZKHUH 0 UHSUHVHQWV VRPH PHDVXUH RI WKH PRQH\ VWRFN 9 LV WKH YHORFLW\ RI FLUFXODWLRQ 3 LV WKH SULFH OHYHO DQG 4 LV UHDO RXWSXW 9HORFLW\ LV QRW DVVXPHG WR EH FRQVWDQW ,QVWHDG LW LV K\SRWKHVL]HG WR EH D SUHGLFWDEOH IXQFWLRQ RI RWKHU PDFURHFRQRPLF YDULDEOHV VXFK DV WKH FRVW RI KROGLQJ FDVK EDODQFHV *LYHQ WKH W\SLFDO YDULDELOLW\ RI YHORFLW\ LQ PDQ\ /DWLQ $PHULFDQ HFRQRPLHV WKLV DVSHFW RI WKH WKHRUHWLFDO PRGHO LV SRWHQWLDOO\ LPSRUWDQW VHH &ODYLMR f 7R XWLOL]H SHUFHQWDJH FKDQJHV WKH YDULDEOHV FDQ EH WUDQVIRUPHG E\ QDWXUDO ORJDULWKPV DQG ILUVW GLIIHUHQFHG ,QWURGXFWLRQ RI D WLPH VXEVFULSW DQG UHDUUDQJHPHQW RI WKH WHUPV \LHOGV WKH EDVLF +DUEHUJHU HJXDWLRQ f '3W '0W '4W ''3Wf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f '3W D4 D'0W D'4W DM'Â'3Af X ZKHUH D DQG D DUH K\SRWKHVL]HG WR EH SRVLWLYH DQG WKH DEVROXWH YDOXHV RI D DQG D VKRXOG ERWK EH VWDWLVWLFDOO\ LQGLVWLQJXLVKDEOH IURP RQH 7KH ODVW DUJXPHQW LQ WKH H[SUHVVLRQ UHSUHVHQWV WKH GLVWXUEDQFH WHUP +DQVRQ f SURSRVHV DQ LPSOLFLW FRVW IXQFWLRQ GXDO RI DQ DJJUHJDWH SURGXFWLRQ IXQFWLRQ ZKLFK LV KRPRJHQHRXV RI GHJUHH RQH 'HULYHG RXWSXW VXSSO\ IXQFWLRQV IURP WKLV IUDPHZRUN ZLOO EH KRPRJHQHRXV RI GHJUHH ]HUR LQ LQSXW DQG RXWSXW SULFHV (TXDWLRQ H[SUHVVHV WKLV UHODWLRQVKLS XVLQJ ORJDULWKPLF ILUVW GLIIHUHQFHV f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f DJEA'3M D4 DE4 D'0W DE'3,W D' 'SW f 8r (TXDWLRQ FDQ EH IXUWKHU VLPSOLILHG SULRU WR HVWLPDWLRQ 'LYLGLQJ WKURXJK E\ WKH OHIWKDQG VLGH FRQVWDQW WHUP DQG UHDUUDQJLQJ WHUPV VXFK WKDW WKH SULFH VHULHV UHPDLQV DV WKH GHSHQGHQW YDULDEOH \LHOGV WKH IROORZLQJ UHODWLRQ f '3W & F'0W F'3,W F-''3Wf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f f '3W F &O'0WB M &'3,WB &M''3WBNf 8" ZKHUH ODJ VXEVFULSWV L M N Q UHVSHFWLYHO\ PAGE 46 7KH DERYH PRGHO SURYLGHV DQ DWWUDFWLYH VWDUWLQJ SRLQW IRU H[DPLQLQJ LQIODWLRQDU\ WUHQGV LQ DQ HFRQRP\ ,W LV QRW KRZHYHU ZLWKRXW SRWHQWLDO SUREOHPV IRU DQDO\]LQJ SULFH PRYHPHQWV LQ D UHODWLYHO\ KLJK LQIODWLRQ FRXQWU\ VXFK DV (FXDGRU $ SULQFLSDO FRQFHUQ DULVHV IURP WKH IDFW WKDW (TXDWLRQ WUHDWV DOO RI WKH UHJUHVVRUV DV H[RJHQRXV RU SUHn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f :KLOH VKRUWWHUP IRUHFDVWLQJ PRGHOV IRU (FXDGRULDQ LQIODWLRQ KDYH QRW EHHQ SUHYLRXVO\ GHYHORSHG GRPHVWLF SULFHV PRQHWDU\ DJJUHJDWHV DQG LPSRUW SULFHV DUH PRGHOHG VLPXOWDQHRXVO\ LQ WKH PDFURHFRQRPHWULF PRGHO HVWLPDWHG XVLQJ DQQXDO GDWD E\ )XOOHUWRQ Ef ,W ZRXOG QRW EH VXUSULVLQJ LI WKH IHHGEDFN UHODWLRQV HQFRXQWHUHG LQ WKDW SDSHU DOVR HPHUJH LQ WKH PRQWKO\ WLPH VHULHV XWLOL]HG EHORZ $V QRWHG HOVHZKHUH PRQHWDU\ DXWKRULWLHV LQ (FXDGRU KDYH RFFDVLRQDOO\ EHHQ IRUFHG WR \LHOG WR SROLWLFDO SUHVVXUHV *DUORZ f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f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nV PDMRU WUDGLQJ SDUWQHUV ,W DOVR RIIHUV D VLQJOH PRQWKO\ LQGH[ IRU SHULRGV ZKHQ WKH JRYHUQPHQW KDV LQVWLWXWHG PXOWLSOH H[FKDQJH UDWH V\VWHPV )XOOHUWRQ Ef 7R FRQVWUXFW WKH FXUUHQF\ LQGH[ LQGLYLGXDO FXUUHQF\ ZHLJKWV DUH FDOFXODWHG DV WKH VXP RI LPSRUWV DQG H[SRUWV ZLWK PAGE 48 HDFK RI WHQ PDMRU WUDGLQJ SDUWQHUV DQG GLYLGHG E\ WRWDO LQWHUQDWLRQDO WUDGH LQ HDFK \HDU 2YHU WKH VDPSOH SHULRG DQQXDO WUDGH ZLWK (FXDGRUnV WRS WHQ LPSRUW VRXUFHV DQG H[SRUW GHVWLQDWLRQV DFFRXQWV IRU PRUH WKDQ SHUFHQW RI LWV WRWDO WUDGH YROXPH LQ DQ\ JLYHQ \HDU 3URGXFWV RI WKH ELODWHUDO WUDGH FRHIILFLHQWV DQG WKH UHVSHFWLYH FXUUHQFLHV DUH WKHQ XVHG WR FRQVWUXFW WKH H[FKDQJH UDWH LQGH[ XVLQJ D JHRPHWULF PHDQ 7KH ODWWHU PHWKRG LV VHOHFWHG WR DYRLG SUREOHPV ZKLFK FDQ SRWHQWLDOO\ UHVXOW IRU LQGH[HV FRQVWUXFWHG XVLQJ DULWKPHWLF PHDQV GXULQJ SHULRGV RI LQIODWLRQDU\ YDULDELOLW\ VHH %DWWHQ DQG %HORQJLD 'XWWRQ DQG *UHQQHV DQG .HUFKHYDO f )RU SHULRGV ZKHQ PXOWLSOH H[FKDQJH UDWHV ZHUH LQVWLWXWHG D EOHQGHG LQGH[ LV FDOFXODWHG :HLJKWV IRU WKH IUHHPDUNHW DQG RIILFLDO JRYHUQPHQW LQWHUYHQWLRQ H[FKDQJH UDWHV DUH REWDLQHG IURP WKH &HQWUDO %DQN SXEOLFDWLRQ ,QIRUPDFLQ (VWDGVWLFD 0HQVXDO YDULRXV LVVXHVf (FRQRPHWULF UHVXOWV IRU WRWDO DQG GLVDJJUHJDWHG LPSRUWV DQG QRQSHWUROHXP H[SRUWV LQGLFDWH WKDW WKH EOHQGHG UDWH SURYLGHV D PRUH DFFXUDWH PHDVXUH RI WKH DSSURSULDWH FXUUHQF\ EDVNHW IRU (FXDGRU )XOOHUWRQ Ef 'LDJQRVWLF WHVWV ZHUH DOVR FRQGXFWHG XVLQJ FXUUHQF\ EDVNHWV ZLWK GLIIHUHQW QXPEHUV RI WUDGLQJ SDUWQHUV ,QWURGXFWLRQ RI WKH WUDGHZHLJKWHG H[FKDQJH UDWH LQGH[ WR (JXDWLRQ FDXVHV WKH PRGHO WR EH HVWLPDWHG WR WDNH WKH IROORZLQJ IRUP PAGE 49 f '3W J J'0WL J'7:;W J''3WNf X ZKHUH 7:; VWDQGV IRU ORJDULWKPLF ILUVW GLIIHUHQFHV RI WKH QRPLQDO YHUVLRQ RI WKH PRQWKO\ WUDGHZHLJKWHG H[FKDQJH UDWH LQGH[ FDOFXODWHG IRU WKH VXFUH E\ 7KH :()$ *URXS IRUPHUO\ :KDUWRQ (FRQRPHWULFVf $OWKRXJK LQFRUSRUDWLRQ RI WKH PRQWKO\ H[FKDQJH UDWH LQGH[ DYRLGV WKH SUREOHPV DVVRFLDWHG ZLWK LPSOLFLW SULFH GHIODWRU LQWHUSRODWLRQ LW PD\ LQWURGXFH RWKHU SUREOHPV GXH WR WKH IDFW WKDW LPSRUW SULFHV LQ (FXDGRU DUH DIIHFWHG E\ JOREDO VXSSO\ DQG LQWHUQDWLRQDO GHPDQG FRQGLWLRQV LQ DGGLWLRQ WR H[FKDQJH UDWH PRYHPHQWV VHH )XOOHUWRQ Ef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f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f ,PSOLFDWLRQV EDVHG RQ WKH *UDQJHU FDXVDOLW\ WHVWV HVWLPDWHG IRU FRQVXPHU SULFHV DQG WKH H[FKDQJH UDWH VHULHV DUH OHVV FOHDU $W VKRUWHU ODJ OHQJWKV WKH QXOO K\SRWKHVLV WKDW FKDQJHV LQ FRQVXPHU SULFHV GR QRW OHDG WR VXEVHTXHQW FKDQJHV LQ WKH LQWHUQDWLRQDO YDOXH RI WKH VXFUH LV UHMHFWHG DW WKH SHUFHQW OHYHO 7KLV FRQFOXVLRQ VLPLODU WR WKH &RORPELDQ UHVXOWV REWDLQHG E\ )XOOHUWRQ Ff DQG .DPDV f LV QRW XSKHOG DW ORQJHU ODJ OHQJWKV DQG PDNHV LW GLIILFXOW WR UHDFK FRQFUHWH FRQFOXVLRQV UHJDUGLQJ SRWHQWLDO IHHGEDFN HIIHFWV EHWZHHQ WKH WZR VHULHV +RZHYHU WUXQFDWLRQ RI WKH ORQJHU ODJ OHQJWKV FDQ ELDV WKH K\SRWKHVLV WRZDUG LQFRUUHFW UHMHFWLRQ RI WKH QXOO K\SRWKHVLV )HLJH DQG 3HDUFH f &RQVHTXHQWO\ PAGE 52 7DEOH 8QLW 5RRW 7HVWV IRU 6WDWLRQDULW\ 6HULHV $XJ 'LFNH\)XOOHU WVWDW 0DF.LQQRQ FULW YDOXH 3 ZLWK FRQVW WUHQG ODJf b OYOf 0 ZLWK FRQVW WUHQG ODJf 7:; ZLWK FRQVW WUHQG ODJf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f '3W f r'0W f r'7:;W f r '0W f r'0W f r'7:; f W r''3WBf r8WB f f r9 f Wn 5 6(5 /RJ OLNHOLKRRG ': )VWDW 3URE)VWDWf ZKHUH WKH QXPEHUV LQ SDUHQWKHVHV DUH FRPSXWHG WVWDWLVWLFV DQG 9W LV WKH HUURU WHUP DVVRFLDWHG ZLWK WKH $50$; PRGHO IRU 8W /DJ OHQJWKV RI PRQWKV ZHUH XVHG LQ WKH LQLWLDO HVWLPDWHV IRU (TXDWLRQ :KLOH D ODUJH QXPEHU RI WKH UHVXOWLQJ FRHIILFLHQWV ZHUH VLJQLILFDQW DW WKH SHUFHQW OHYHO VHULDO FRUUHODWLRQ ZDV SUHVHQW LQ WKH UHVLGXDOV 7R DYRLG SRWHQWLDOO\ VSXULRXV HVWLPDWLRQ UHVXOWV VHH +DPLOWRQ f D QRQOLQHDU $50$; SURFHGXUH LV XWLOL]HG WR FRUUHFW IRU DXWRFRUUHODWLRQ 7KLV HVWLPDWRU 3DJDQ f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f VSHFLILFDWLRQ LV XVHG WR FKDUDFWHUL]H WKH GDWD JHQHUDWLQJ SURFHVV IRU WKH UHVLGXDOV $V ZRXOG EH H[SHFWHG LQ DQ LQIODWLRQDU\ HFRQRP\ WKH DOJHEUDLF VLJQ RI WKH LQWHUFHSW LQ (JXDWLRQ LV SRVLWLYH %HFDXVH WKH PRGHO LV HVWLPDWHG XVLQJ GLIIHUHQFHG GDWD WKLV UHVXOW LQGLFDWHV WKDW D V\VWHPDWLF XSZDUG WUHQG H[LVWV LQ WKH (FXDGRULDQ FRQVXPHU SULFH LQGH[ $V K\SRWKHVL]HG E\ +DQVRQ f WKH VXP RI WKH FRHIILFLHQWV IRU WKH ODJJHG PRQHWDU\ VHULHV LV VLJQLILFDQWO\ OHVV WKDQ RQH 6LPLODU WR )XOOHUWRQ D Ff EXW XQOLNH .DPDV f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f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Â£Q %DOOHQ'DKLN *DU]R]L DGPLQLVWUDWLRQ LQ 6DPSOH GDWD XVHG WR HVWLPDWH (TXDWLRQ RQO\ LQFOXGHV LQIRUPDWLRQ DYDLODEOH WR JRYHUQPHQW SROLF\ PDNHUV DW WKH WLPH WKH LQIODWLRQ WDUJHW ZDV DQQRXQFHG 7KH VLPXODWLRQ DQDO\VHV WKXV VDWLVI\ WKH .OHLQ f DQG &KULVW f FULWHULD IRU IRUHFDVW HYDOXDWLRQ 2I FRXUVH DGGLWLRQDO SROLF\ DQDO\VLV ZLWK GLIIHUHQW PRGHO VSHFLILFDWLRQV PD\ DOVR EH XVHIXO PAGE 58 7R H[DPLQH WKH IHDVLELOLW\ RI WKH JRYHUQPHQWn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nV 'HFHPEHU SHVR GHYDOXDWLRQ %DQFR GHO 3DFILFR f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nV DQWLLQIODWLRQDU\ SURJUDP LV IDLUO\ FUHGLEOH 7KLV FRQFOXVLRQ LV SUHGLFDWHG XSRQ HYHQWXDO GHFHOHUDWLRQ LQ WKH UDWHV RI OLTXLGLW\ JURZWK DQG FXUUHQF\ GHSUHFLDWLRQ EURXJKW DERXW E\ WKH FHQWUDO EDQN ,Q OLJKW RI SUHYLRXV PRQHWDU\ SROLF\ DQDO\VLV FRQGXFWHG IRU PAGE 60 7DEOH 3ROLF\ 6LPXODWLRQ 5HVXOWV 0RQWK ,PPHGLDWH 5DSLG *UDGXDO 1RQH PAGE 61 (FXDGRU WKH UHVXOWV REWDLQHG DUH SOHDVDQWO\ VXUSULVLQJ )XOOHUWRQ D RIIHUV D QHJDWLYH DVVHVVPHQW RI DQ HDUOLHU SULFH VWDELOL]DWLRQ HIIRUWf 7KDW WKH JRYHUQPHQWnV DQQRXQFHG WDUJHWV DUH QRW FRPSOHWHO\ DWWDLQHG LV LQ OLQH ZLWK RWKHU VWXGLHV RI /DWLQ $PHULFDQ LQIODWLRQ SROLFLHV )XOOHUWRQ F DQG .DPDV f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f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f UDWLRV HVWLPDWHG IRU HDFK VHW RI PRGHO EDVHG DQG UDQGRP ZDON IRUHFDVWV 7KH 8VWDWLVWLF IRU DQ LQGLYLGXDO IRUHFDVW VWHS OHQJWK LV HTXDO WR WKH UDWLR RI WKH PRGHO EDVHG 506( WR WKDW RI D QR FKDQJH 506( :KHQ WKH UHVXOWLQJ FRHIILFLHQW LV OHVV WKDQ RQH WKH HTXDWLRQ IRUHFDVW KDV RXWSHUIRUPHG WKH UDQGRP ZDON SUHGLFWLRQ (DUOLHU 6WXGLHV 7KHUH KDYH EHHQ D JURZLQJ QXPEHU RI VWXGLHV UHJDUGLQJ GHYHORSLQJ HFRQRP\ H[WHUQDO LQGHEWHGQHVV DQG WKH VHFRQGDU\ PDUNHW LQ UHFHQW \HDUV *HQQRWWH .KDUDV DQG 6DGHT f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f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f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f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f WKHVH W\SHV RI FKDUDFWHUL]DWLRQV DUH VWUDLJKWIRUZDUG DQG LQWXLWLYHO\ DSSHDOLQJ ,Q D VWDWLVWLFDO IUDPHZRUN YDULDEOHV VHOHFWHG WR PAGE 68 FDSWXUH PDUNHW IXQGDPHQWDOV FDQ DOVR EH FRPELQHG ZLWK UHJUHVVRUV GHVLJQHG WR UHSUHVHQW H[RJHQRXV ULVNV 8VLQJ SRROHG REVHUYDWLRQV DQG DQQXDO GDWD WKH HFRQRPHWULF UHVXOWV RI WKHVH DXWKRUV VKRZ WKDW HYHQ VLPSOH VSHFLILFDWLRQV FDQ UHSUHVHQW VHFRQGDU\ PDUNHW GHYHORSLQJ GHEW SULFHV ZLWK D KLJK GHJUHH RI DFFXUDF\ 3HUDVVR f DOVR HPSKDVL]HV HFRQRPLF IDFWRUV LQ KLV VWXG\ RI VHFRQGDU\ PDUNHW SULFHV 7R UHIOHFW WKH LPSRUWDQFH RI GHEWHTXLW\ FRQYHUVLRQV KLV SULFLQJ PRGHO LV GHULYHG IURP D SURILW PD[LPL]LQJ IUDPHZRUN WKDW LQFOXGHV UHDO LQWHUHVW UDWH PHDVXUHV UHDO FRVWV RI FDSLWDO LQWHUQDWLRQDO ZDJH GLIIHUHQWLDOV DVVXPHG WR LQGXFH PDQXIDFWXUHUV WR LQYHVW DEURDGf DQG LQGLYLGXDO HFRQRP\ SHUIRUPDQFH YDULDEOHV 7LPH VHULHV DQG FURVV FRXQWU\ DQQXDO GDWD DUH SRROHG SULRU WR HVWLPDWLRQ 6RPH FRHIILFLHQWV DUH VWDWLVWLFDOO\ LQVLJQLILFDQW RU RI WKH ZURQJ VLJQ EXW RYHUDOO HPSLULFDO UHVXOWV IRU WKH HVWLPDWHG HTXDWLRQV IRU VHFRQGDU\ PDUNHW SULFHV DUH IDLUO\ VWURQJ (PSLULFDO PRGHOV GHYHORSHG E\ LQYHVWPHQW EDQNHUV LQGLFDWH WKDW VHFRQGDU\ PDUNHW GHEW SULFHV FDQ EH PRGHOHG DQG IRUHFDVWHG VHH 3XUFHOO DQG 2UODQVNL f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f FRQFOXGHV WKDW HFRQRPLF IDFWRUV DUH PRUH LPSRUWDQW DV H[SODQDWRU\ YDULDEOHV WKDQ SROLWLFDO DQG VWUXFWXUDO ULVN IDFWRUV 6DFKV DQG +XL]LQJD f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f 7KLV LQLWLDO DWWHPSW SURYLGHV D XVHIXO VWDUWLQJ SRLQW IRU DGGUHVVLQJ TXHVWLRQV UHJDUGLQJ WKH SUHGLFWDELOLW\ RI VHFRQGDU\ PDUNHW GHYHORSLQJ FRXQWU\ GHEW SULFHV 'HEW SULFH GDWD XVHG LQ WKH HPSLULFDO HVWLPDWHV DUH FROOHFWHG E\ 6DORPRQ %URWKHUV LQ 1HZ PAGE 71 *RYHUQPHQW HFRQRPLVWV KDYH FRQVLVWHQWO\ WUHDWHG LQWHUQDWLRQDO FUHGLW PDUNHWV DV VRXUFHV RI ILQDQFLQJ WR EULGJH GRPHVWLF VDYLQJV JDSV )XOOHUWRQ Ef &RORPELDQ GHEW QHJRWLDWRUV KDYH QHYHU VRXJKW D %UDG\ LQLWLDWLYH ZULWHRII DUJXLQJ WKDW WR GR VR ZRXOG RQO\ LPSDLU WKH QDWLRQnV FUHGLWZRUWKLQHVV 2YHU WKH SHULRG IURP IRUZDUG &RORPELD KDV QHYHU GHFODUHG DQ LQWHUHVW SD\PHQW PRUDWRULXP DQG KDV RQH RI WKH EHVW GHEW VHUYLFH UHFRUGV DPRQJ GHYHORSLQJ FRXQWULHV ZKLFK KDYH XWLOL]HG H[WHUQDO ILQDQFLQJ VRXUFHV (FXDGRU GHFODUHG SULQFLSDO DQG LQWHUHVW PRUDWRULD LQ DV D UHVXOW RI WKH ILQDQFLDO DIWHUPDWK IROORZLQJ WKH HDUWKTXDNH ZKLFK VKDWWHUHG WKH FRXQWU\nV WUDQV$QGHDQ RLO SLSHOLQH 7KH ODWWHU HYHQW LQWHUUXSWHG (FXDGRUnV SULQFLSDO VRXUFH RI H[SRUW HDUQLQJV DQG GHVWUR\HG PXFK RI LWV SK\VLFDO LQIUDVWUXFWXUH ,Q WKH %RUMD DGPLQLVWUDWLRQ UHVXPHG QHJRWLDWLRQV ZLWK FRPPHUFLDO FUHGLWRUV DQG HYHQWXDOO\ EHJDQ KRQRULQJ SHUFHQW RI WKH LQWHUHVW FRPLQJ GXH RQ FRPPHUFLDO ORDQV )XOOHUWRQ Ff 3URJUHVV UHJDUGLQJ WKH WUHDWPHQW RI JURZLQJ DPRUWL]DWLRQ DQG LQWHUHVW DUUHDUV UHPDLQHG HOXVLYH WKURXJK WKH EDODQFH RI WKH %RUMD JRYHUQPHQW ZKLFK VWHSSHG GRZQ LQ $ QHZ URXQG RI GLVFXVVLRQV ZLWK (FXDGRUnV EDQN DGYLVRU\ FRPPLWWHH EHJDQ DIWHU WKH 'XUÂ£Q %DOOHQ'DKLN *DU]R]L JRYHUQPHQW WRRN RIILFH 9HQH]XHOD DOVR HQFRXQWHUHG GHEW VHUYLFH SUREOHPV IROORZLQJ QHJDWLYH RLO SULFH VKRFNV LQ DQG (YHQWXDOO\ WKH 3HUH] DGPLQLVWUDWLRQ UHVFKHGXOHG FRPPHUFLDO PAGE 72 FUHGLWV XQGHU D %UDG\ LQLWLDWLYH DJUHHPHQW ZLWK 9HQH]XHODnV EDQN DGYLVRU\ FRPPLWWHH )XOOHUWRQ Ff 7KH DJUHHPHQW RIIHUHG EDQN ORDQ V\QGLFDWH PHPEHUV ILYH PHQX RSWLRQV GHVLJQHG WR UHOLHYH EDODQFH RI SD\PHQW SUHVVXUHV IDFHG E\ WKLV HFRQRP\ 6HYHUDO (XURPRQH\ ERQG LVVXHV ZHUH VXFFHVVIXOO\ IORDWHG LQ VXEVHJXHQW SHULRGV DQG 9HQH]XHOD WHPSRUDULO\ UHJDLQHG DFFHVV WR LQWHUQDWLRQDO FUHGLW PDUNHWV 'HEW LQVWUXPHQWV IRU DOO WKUHH FRXQWULHV KDYH WUDGHG DW VXEVWDQWLDO GLVFRXQWV IURP IDFH YDOXH LQ UHFHQW \HDUV *LYHQ LWV VXSHULRU VHUYLFH UHFRUG LW LV QRW VXUSULVLQJ WKDW &RORPELDQ SDSHU JHQHUDOO\ FDUULHV D KLJKHU SULFH WKDQ WKDW RI LWV WZR $QGHDQ QHLJKERUV 6LPLODUO\ JLYHQ LWV KLJKHU OHYHO RI SD\PHQW DUUHDUV DQG ZRUVH HFRQRPLF SHUIRUPDQFH (FXDGRULDQ SDSHU WHQGV WR WUDGH DW VKDUSHU GLVFRXQWV WKDQ WKRVH RI WKH RWKHU FRXQWULHV LQ WKH VDPSOH 7DEOH SUHVHQWV VXPPDU\ VWDWLVWLFV IRU HDFK VHFRQGDU\ PDUNHW VRYHUHLJQ GHEW SULFH VHULHV 'LVFRXQWV IURP IDFH YDOXH RQ &RORPELDQ GHEW FHUWLILFDWHV ZHUH OHVV YDULDEOH WKDQ WKRVH RI (FXDGRU DQG 9HQH]XHOD GXULQJ WKH 0DUFK 'HFHPEHU VDPSOH SHULRG 7KH UDQJH DQG VWDQGDUG GHYLDWLRQ IRU &RORPELDQ GHEW JXRWHV DUH VPDOOHU WKDQ WKH RWKHUV ZKLOH WKRVH IRU (FXDGRU DUH WKH ODUJHVW RI WKH WKUHH 7KH DULWKPHWLF PHDQV IRU HDFK VHULHV IROORZ WKH RSSRVLWH SDWWHUQ LQ WHUPV RI UDQNLQJ 'DWD LQ 7DEOH DUH LQ FHQWV SHU GROODU RU SHUFHQW RI IDFH YDOXH RI WKH ORDQV WKH XQLWV LQ ZKLFK WUDQVDFWLRQV DUH FRQGXFWHG DW PRQH\ FHQWHU WUDGLQJ GHVNV PAGE 73 7DEOH 6HFRQGDU\ 0DUNHW 'HEW 3ULFH 6XPPDU\ 6WDWLVWLFV &RXQWU\ $YHUDJH 0D[LPXP 0LQLPXP 6WDQGDUG 'HYLDWLRQ &RORPELD (FXDGRU 9HQH]XHOD 7KH VDPSOH SHULRG LV 0DUFK 'HFHPEHU PAGE 74 (DFK VHULHV LV PRGHOHG DV D IXQFWLRQ RI NH\ HFRQRPLF DQG ILQDQFLDO YDULDEOHV ZKLFK DUH HDVLO\ REVHUYHG DV ZHOO DV OLNHO\ WR EH XVHG E\ VHFRQGDU\ PDUNHW SDUWLFLSDQWV DV LQGLFDWRUV RI WKH FUHGLWZRUWKLQHVV RI WKH LQGLYLGXDO FRXQWU\ 3RVVLEOH UHJUHVVRUV LQFOXGH ODJJHG GHEW TXRWHV ZRUOG LQWHUHVW UDWHV FRPPRGLW\ H[SRUW SULFHV LQWHUQDWLRQDO UHVHUYHV DQG GRPHVWLF SULFH LQGH[HV IRU GLVFXVVLRQ VHH :DNHPDQ/LQQ f $V PHQWLRQHG DERYH VHULHV VXFK DV WKHVH DUH W\SLFDOO\ LQFOXGHG LQ PDFURHFRQRPHWULF IRUHFDVWLQJ PRGHOV RI /DWLQ $PHULFD GXH WR WKHLU XVHIXOQHVV LQ SUHGLFWLQJ EDODQFH RI SD\PHQW PRYHPHQWV DQG JHQHUDO HFRQRPLF SHUIRUPDQFH VHH )XOOHUWRQ D Ef $ WKUHHVWDJH JHQHUDOL]HG OHDVW VTXDUHV 6/6f UHJUHVVLRQ WHFKQLTXH LV XVHG WR MRLQWO\ HVWLPDWH PRGHO SDUDPHWHUV =HOOQHU DQG 7KHLO f 'RLQJ VR SHUPLWV LQFRUSRUDWLQJ SRWHQWLDO FURVVHTXDWLRQ VLPXOWDQHLW\ HIIHFWV RI HYHQWV VXFK DV &LWLEDQNnV GHFLVLRQ WR XQLODWHUDOO\ LQFUHDVH ORDQ ORVV UHVHUYHV LQ 7KH ODWWHU LV EHOLHYHG WR KDYH UHGXFHG RYHUDOO VHFRQGDU\ PDUNHW OLTXLGLW\ DQG DOVR UHGXFHG WKH DWWUDFWLYHQHVV VRYHUHLJQ GHEW SDSHU WR PRVW FUHGLWRUV LUUHVSHFWLYH RI WKHLU LQGLYLGXDO SD\PHQW UHFRUGV *DMGHF]ND DQG 6WRQH DQG 6QRZGHQ f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f 6HFRQG WKH FRPSOHWH DEVHQFH RI RWKHU VWXGLHV RI WKLV QDWXUH LQFUHDVHV WKH YDOXH RI DQ LQLWLDO DWWHPSW WR HVWDEOLVK ZKHWKHU DQ\ UHJXODULWLHV DW DOO DUH SUHVHQW LQ WKH GDWD &KULVW f %RWK SRLQWV ZHUH UHSHDWHGO\ UDLVHG E\ ILQDQFLDO HFRQRPLVWV ZKR SDUWLFLSDWH LQ WKH VHFRQGDU\ PDUNHW DQG DWWHQGHG WKH 6RYHUHLJQ 'HEW &RQIHUHQFH VSRQVRUHG E\ 7KH :()$ *URXS LQ 1HZ PAGE 76 VHULHV HVWLPDWRUV 7KH VHULHV DUH QRW GLIIHUHQFHG SULRU WR HVWLPDWLRQ EXW DUH ORJDULWKPLFDOO\ WUDQVIRUPHG %HFDXVH WKH GDWD DUH LQ OHYHOV LW LV LPSRUWDQW WR DVVHVV ZKHWKHU WKH VHULHV DUH FRLQWHJUDWHG 7KH ODWWHU DVVXPSWLRQ LV WHVWHG YLD XQLWURRW WHVWV RQ WKH LQGLYLGXDO PRGHO UHVLGXDOV IRU GLVFXVVLRQ VHH +DPLOWRQ f %HFDXVH PRQWKO\ GDWD DUH XVHG LW LV QRW SRVVLEOH WR XWLOL]H WKH VDPH UHJUHVVRU YDULDEOHV DV KDYH EHHQ XVHG LQ HDUOLHU VWXGLHV LQFRUSRUDWLQJ TXDUWHUO\ RU DQQXDO VHULHV IURP QDWLRQDO LQFRPH DQG SURGXFW DFFRXQWV 7KHUH DUH VWLOO D QXPEHU RI SRWHQWLDO FDQGLGDWH VHULHV ZKLFK WKH ILQDQFLDO FRPPXQLW\ PD\ XVH DV LQGLFDWRUV RI D FRXQWU\nV FUHGLWZRUWKLQHVV DQG ZLOO SRWHQWLDOO\ LQIOXHQFH VHFRQGDU\ PDUNHW SULFH TXRWHV ,W VKRXOG EH QRWHG WKDW VHWV RI LQGLFDWRUV ZLOO YDU\ IRU GLIIHUHQW FRXQWULHV DFFRUGLQJ WR LQGLYLGXDO HFRQRPLF HQGRZPHQWV DQG SHUIRUPDQFH UHFRUGV 7KLV DUJXPHQW LV VLPLODU WR WKDW SUHYLRXVO\ PDGH IRU HTXDWLRQV XVHG WR HVWLPDWH GHYHORSLQJ FRXQWU\ ERUURZLQJ OHYHOV XQGHU GLIIHUHQW UHJLPHV (DWRQ DQG *HUVRYLW] f )RU &RORPELD (TXDWLRQ LQ 7DEOH WKH UHJUHVVRUV LQFOXGH D RQHPRQWK ODJ RI WKH GHEW SULFH WKH HIIHFWLYH DQQXDO UDWH IRU WKH GD\ /RQGRQ LQWHUEDQN RIIHU UDWH /,%25f DQG WKH RQH PRQWK FKDQJH LQ WKH QDWLRQDO FRQVXPHU SULFH LQGH[ &3,f 7KH RQHPRQWK ODJ RQ WKH GHEW SULFH LV LQFOXGHG WR SURYLGH LQIRUPDWLRQ RQ KRZ WKH PDUNHW KDV YDOXHG &RORPELDQ SDSHU LQ WKH PRVW UHFHQW SHULRG ,QWHUQDWLRQDO PAGE 77 ORDQV WR VRYHUHLJQ QDWLRQV W\SLFDOO\ FDUU\ YDULDEOH LQWHUHVW UDWHV GHILQHG LQ WHUPV RI D IL[HG VSUHDG RYHU WKH PRQWK /,%25 8SZDUG FKDQJHV LQ WKH YDULDEOH LQWHUHVW UDWH DVVHVVHG RQ VXFK ORDQV ZLOO UHGXFH &RORPELDnV FXUUHQW DFFRXQW EDODQFH DQG H[HUW GRZQZDUG SUHVVXUH RQ VHFRQGDU\ PDUNHW SULFHV 7KH UDWH RI LQIODWLRQ LV XVHG DV D SUR[\ IRU RYHUDOO HFRQRPLF FRQGLWLRQV :KHQ LQIODWLRQ ULVHV &RORPELDQ PRQHWDU\ DXWKRULWLHV JHQHUDOO\ DWWHPSW WR WLJKWHQ FUHGLW FRQGLWLRQV )XOOHUWRQ )DLQERLP DQG $JXGHOR f 3DUDPHWHU HVWLPDWHV IRU HDFK YDULDEOH KDYH WKH H[SHFWHG DULWKPHWLF VLJQV EXW WKH WVWDWLVWLF IRU WKH LQIODWLRQ WHUP LV QRW VLJQLILFDQW DW WKH SHUFHQW OHYHO ,Q WKH FDVH RI (FXDGRU (TXDWLRQ LQ 7DEOH RQO\ WZR SUHGHWHUPLQHG YDULDEOHV DUH LQFOXGHG LQ WKH WKUHHVWDJHG OHDVW VTXDUHV HTXDWLRQ 7KH ILUVW LV D RQHSHULRG ODJ RI WKH VHFRQGDU\ PDUNHW GHEW SULFH VHULHV 7KH VHFRQG LV WKH PRQWK /,%25 UDWH 7KHVH VHULHV DUH LQFOXGHG IRU WKH VDPH UHDVRQV DV WKH\ ZHUH XVHG LQ WKH &RORPELDQ HTXDWLRQ %RWK FRHIILFLHQWV KDYH WKH H[SHFWHG VLJQV DQG DUH VWDWLVWLFDOO\ VLJQLILFDQW 7KH DEVHQFH RI RWKHU EDODQFH RI SD\PHQW LQGLFDWRU VHULHV VXFK DV H[SRUW SULFHV PD\ VHHP VXUSULVLQJ %HFDXVH (FXDGRU VLPLODU WR &RORPELD KDV D UHODWLYHO\ GLYHUVH FRPPRGLW\ H[SRUW EDVNHW QR VLQJOH SULFH VHULHV ZLOO VXIILFH )XOOHUWRQ D Ef 7KLV LV QRW WKH FDVH IRU 9HQH]XHOD ZKHUH SHWUROHXP SURGXFWV DFFRXQW IRU PRUH WKDQ SHUFHQW RI WRWDO PHUFKDQGLVH H[SRUWV )XOOHUWRQ Ff PAGE 78 7DEOH 7KUHH6WDJH /HDVW 6TXDUHV 5HJUHVVLRQ 5HVXOWV &RORPELD 3W r3W r/,%250 f f f r&3,W f 5 4 f 6(5 4f )VWDW 3URE)VWDWf (FXDGRU 3W r3W f f r/,%250 f 5 4 f 6(5 )VWDW 4f 3URE)VWDWf 9HQH]XHOD 3W OOr3W f f r/,%250 f r,5W f r2,/W f 5 6(5 )VWDW 4f 4f 3URE)VWDWf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f :RUOG RLO SULFHV SURYLGH D PHDVXUH RI JOREDO GHPDQG IRU WKH QDWLRQnV SULQFLSDO H[SRUW $OO RI WKH UHJUHVVRU FRHIILFLHQWV KDYH WKH H[SHFWHG VLJQV 7KH WVWDWLVWLF IRU LQWHUQDWLRQDO UHVHUYHV KRZHYHU LV QRW VLJQLILFDQW DW WKH SHUFHQW OHYHO 7KH 6/6 WHFKQLTXH XWLOL]HG WR HVWLPDWH WKH SDUDPHWHUV UHSRUWHG LQ 7DEOH DOORZV IRU SRWHQWLDO FURVV HTXDWLRQ FRQWHPSRUDQHRXV HUURU FRUUHODWLRQ D UHDVRQDEOH DVVXPSWLRQ IRU WKH VHFRQGDU\ PDUNHW IRU VRYHUHLJQ GHEW SDSHU 7R H[DPLQH ZKHWKHU WKLV DVVXPSWLRQ LV QHFHVVDU\ UHVLGXDOV IURP WKH ILUVW VWDJH RUGLQDU\ OHDVW VTXDUHV 2/6f UHJUHVVLRQ DUH WHVWHG IRU FRQWHPSRUDQHRXV FRUUHODWLRQ XVLQJ D VWDQGDUG WWHVW PHWKRGRORJ\ VHH 2VWOH DQG 0HQVLQJ RU 6QHGHFRU f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f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f b OYOf ODJf ODJf PAGE 83 7R IXUWKHU HQKDQFH WKH UHDOLVP RI WKH H[WUDSRODWLRQ VFHQDULRV DFWXDO IRUHFDVW GDWD DYDLODEOH WR WKH ILQDQFLDO PDUNHWV GXULQJ HDFK RI WKH VXESHULRGV DUH LQFRUSRUDWHG LQ WKH VLPXODWLRQV %\ UHO\LQJ RQ IRUHFDVW GDWD ZKLFK DUH XQFRQGLWLRQDO XSRQ DQ\ LQIRUPDWLRQ QRW DYDLODEOH SULRU WR WKH VWDUW RI DQ\ VLPXODWLRQ SHULRG .OHLQnV f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f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f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f 7KH ODWWHU FRQWLQXH WR HQMR\ D FHQWUDO UROH LQ EXVLQHVV DQG JRYHUQPHQW SODQQLQJ H[HUFLVHV WKURXJKRXW WKH ZRUOG *LYHQ WKH VKRUWWHUP XQFHUWDLQWLHV SUHVHQW LQ /DWLQ $PHULFD JXDUWHUO\ PRGHOLQJ DQG IRUHFDVWLQJ ZLOO OLNHO\ EH ZHOFRPHG ZLWK HQWKXVLDVP $W WKLV MXQFWXUH PXFK ZRUN UHPDLQV WR EH GRQH EXW LQLWLDO HIIRUWV VXFK DV WKRVH SUHVHQWHG DERYH SRLQW WR QXPHURXV SRWHQWLDO VXFFHVVHV WR EH JDLQHG IURP IXWXUH UHVHDUFK HIIRUWV PAGE 90 5()(5(1&(6 $QD\LRWRV DQG GH 3LQLV 7KH 6HFRQGDU\ 0DUNHW DQG WKH ,QWHUQDWLRQDO 'HEW 3UREOHP :RUOG 'HYHORSPHQW %DQFR GHO 3DFILFR D (FRQRPLF 5HSRUW 6HSWHPEHU2FWREHUf %DQFR GHO 3DFILFR E (FRQRPLF 5HSRUW 1RYHPEHU'HFHPEHUf %DQFR GHO 3DFILFR (FRQRPLF 5HSRUW 0DUFK $SULOf %DUJHU + DQG /5 .OHLQ $ 4XDUWHUO\ 0RGHO IRU WKH 8QLWHG 6WDWHV (FRQRP\ -RXUQDO RI WKH $PHULFDQ 6WDWLVWLFDO $VVRFLDWLRQ %DWWHQ '6 DQG 07 %HORQJLD 0RQHWDU\ 3ROLF\ ([FKDQJH 5DWHV DQG 86 $JULFXOWXUDO ([SRUWV $PHULFDQ -RXUQDO RI $JULFXOWXUDO (FRQRPLFV %HOWUÂ£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Â£PLFD GH ODV 6HULHV GH 7LHPSR (O &DVR GHO ,QGLFH 1DFLRQDO GH 3UHFLRV DO &RQVXPLGRU(PSOHDGRV 5HYLVWD GHO %DQFR GH OD 5HSEOLFD &KDWILHOG & 7KH $QDO\VLV RI 7LPH 6HULHV 1HZ PAGE 92 )ULVFLD $% &RPPHQW RQ )XOOHUWRQ $WODQWLF (FRQRPLF -RXUQDO 0DUFKf )XOOHUWRQ 70 -U D $ &RPSRVLWH $SSURDFK WR )RUHFDVWLQJ 6WDWH *RYHUQPHQW 5HYHQXHV ,QWHUQDWLRQDO -RXUQDO RI )RUHFDVWLQJ )XOOHUWRQ 70 -U E ([FKDQJH 5DWH ,QGH[HV IRU (FXDGRU DQG 9HQH]XHOD /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS -XQHf )XOOHUWRQ 70 -U F (FXDGRU /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS 6HSWHPEHUf )XOOHUWRQ 70 -U D (FXDGRU /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS -XQHf )XOOHUWRQ 70 -U E &RORPELD /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS 0DUFKf )XOOHUWRQ 70 -U F 9HQH]XHOD /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS 6HSWHPEHUf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f )XOOHUWRQ 70 -U 5 +LUWK DQG 0 6PLWK ,QIODWLRQDU\ '\QDPLFV DQG WKH $QJHOO-RKQVRQ 3URSRVDOV $WODQWLF (FRQRPLF -RXUQDO 0DUFKf )XOOHUWRQ 70 -U DQG $ .DSXU 3UHGLFFLQ GH 0XOWLSOLFDGRUHV 0RQHWDULRV HQ &RORPELD (FXDGRU \ 9HQH]XHOD /HFWXUDV GH (FRQRPD *DMGHF]ND 3 DQG 05 6WRQH 7KH 6HFRQGDU\ 0DUNHW IRU 'HYHORSLQJ &RXQWU\ /RDQV )LQDQFH DQG 'HYHORSPHQW 'HFHPEHUf *DUORZ '& (FXDGRU /DWLQ $PHULFD (FRQRPLF 2XWORRN 7KH :()$ *URXS 6HSWHPEHUf *HQQRWWH +.KDUDV DQG 66DGHT $ 9DOXDWLRQ 0RGHO IRU 'HYHORSLQJ&RXQWU\ 'HEW ZLWK (QGRJHQRXV 5HVFKHGXOLQJ :RUOG %DQN (FRQRPLF 5HYLHZ *UDQJHU &:)RUHFDVWLQJ 6WRFN 0DUNHW 3ULFHV ,QWHUQDWLRQDO -RXUQDO RI )RUHFDVWLQJ +DNNLR & DQG 0 5XVK &RLQWHJUDWLRQ +RZ 6KRUW LV WKH /RQJ5XQ" -RXUQDO RI ,QWHUQDWLRQDO 0RQH\ DQG )LQDQFH +DOOPDQ -5' 3RUWHU DQG '+ 6PDOO ,V WKH 3ULFH /HYHO 7LHG WR WKH 0 0RQHWDU\ $JJUHJDWH LQ WKH /RQJ 5XQ" $PHULFDQ (FRQRPLF 5HYLHZ +DPLOWRQ -' 7LPH 6HULHV $QDO\VLV 3ULQFHWRQ 13ULQFHWRQ 8QLYHUVLW\ 3UHVV +DQVRQ -$ ,QIODWLRQ DQG ,PSRUWHG ,QSXW 3ULFHV LQ VRPH ,QIODWLRQDU\ /DWLQ $PHULFDQ (FRQRPLHV -RXUQDO RI 'HYHORSPHQW (FRQRPLFV PAGE 94 +DUEHUJHU $ 7KH '\QDPLFV RI ,QIODWLRQ LQ &KLOH LQ 0HDVXUHPHQW LQ (FRQRPLFV 6WXGLHV LQ 0DWKHPDWLFDO (FRQRPLFV DQG (FRQRPHWULFV LQ 0HPRU\ RI PAGE 95 0RWOH\ % ,QIODWLRQ DQG *URZWK )5%6) :HHNO\ /HWWHU 'HFHPEHU f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f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t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a!n $VVLVWDQW 3URIHVVRU RI (FRQRPLFV FHUWLI\ WKDW KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ X 7HUU\ / 0F&R\ 3URIHVVRU RI /DWLQ $PHULFDQ 6WXGLHV 7KLV GLVVHUWDWLRQ ZDV VXEPLWWHG WR WKH *UDGXDWH )DFXOW\ RI WKH 'HSDUWPHQW RI (FRQRPLFV LQ WKH &ROOHJH RI %XVLQHVV $GPLQLVWUDWLRQ DQG WR WKH *UDGXDWH 6FKRRO DQG ZDV DFFHSWHG DV SDUWLDO IXOILOOPHQW RI WKH UHTXLUHPHQWV IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ $XJXVW .DUHQ $ +ROEURRN 'HDQ *UDGXDWH 6FKRRO xml version 1.0 encoding UTF-8 REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd INGEST IEID EJFCDMDO8_C5CGBK INGEST_TIME 2014-06-17T15:24:45Z PACKAGE AA00022260_00001 AGREEMENT_INFO ACCOUNT UF PROJECT UFDC FILES |