An empirical investigation of some economic consequences of FASB statement no. 8

MISSING IMAGE

Material Information

Title:
An empirical investigation of some economic consequences of FASB statement no. 8 foreign currency translation
Uncontrolled:
Foreign currency translation
Physical Description:
viii, 142 leaves : ; 28 cm.
Language:
English
Creator:
Jain, Rohit
Publication Date:

Subjects

Subjects / Keywords:
Statement of financial accounting standards   ( lcsh )
Foreign exchange -- Accounting   ( lcsh )
Financial statements   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1980.
Bibliography:
Includes bibliographical references (leaves 136-141).
Statement of Responsibility:
by Rohit Jain.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000099693
notis - AAL5151
oclc - 07093235
System ID:
AA00003468:00001

Full Text








AN EMPIRICAL INVESTIGATION O0' SOME ECONOMIC CONSEQUENCES
OF FASB STATEMENT NO. 8: FOREIGN CURRENCY TRANSLATION







BY

ROHIT JAIN


A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1980









































Copyright 1980

by

Rohit Jain













ACKNOWLEDGEMENTS


I am indebted to my supervisory committee, Professor A. Rashad

Abdel-Khalik, Professor Dan Smith and Professor Antal Majthay for their

help and guidance throughout' the long process of writing a dissertation.

Professor Abdel-Khalik, in particular, gave valuable advice and encourage-

ment at each step of the research and provided detailed comments that

substantially improved the presentation of this dissertation. I am

grateful to Professor Bipin Ajinkya who, though not formally on my

committee, gave liberally of his time and helped at each stage of the

research process. I am also grateful to Professor Russ Fogler for

reading the draft and participating in my dissertation defense.

I would like to thank the many friends who made my stay in Gainesville

enjoyable, in particular Joseph and Pat Noronha, Paul Mathias and my

various roommates. The other doctoral students in accounting also deserve

my gratitude.

Finally, I would like to thank my wife, Ila, whose encouragement,

support and help made the completion of this dissertation possible.










TABLE OF CONTENTS


ACKNOWLEDGEMENTS ................................................... iii

ABSTRACT........................................................... vi

PAGE
CHAPTER I INTRODUCTION..... ................................... 1

Objectives of Research ............................ 2
Note............................................... 3

CHAPTER II CONCEPTUAL ISSUES IN FOREIGN CURRENCY TRANSLATIONS... 4

Fluctuations in Exchange Rates.................... 4
Accounting & Economic Risk Defined................ 5
Reporting Methods for Foreign Currency
Translations................................... 6
Possible Effects of FASB No. 8..................... 14
Review of Previous Empirical Studies on
FASB No. 8..................................... 21
Notes.............................................. 26

CHAPTER III RESEARCH DESIGN...................................... 27

Hypotheses......................................... 27
Measuring Effects on Security Returns............. 29
Problems in Measuring Effects of FASB No. 8....... 33
Data.............................................. 35
Grouping of Firms.................................. 46
Time Periods for Tests............................. 50
Construction of Relative Risk Equalized Returns
and Average Residuals .......................... 52
Measuring Effects on Financial Structure........... 54
Statistical Tests................................. 56
Interpretation of Expected Results................. 57
Notes.............................................. 60

CHAPTER IV TEST RESULTS AND ANALYSIS............................ 62

Variability of Income ............................. 62
Effect on Security Price Returns.................. 65
Effect on Financial Structure of Firms............. 107
Notes............................................. 116

CHAPTER V SUMMARY AND CONCLUSIONS............................... 117


1











APPENDIX

A QUESTIONNAIRE SENT TO TEST FIRMS.................... 125

B SUMMARY RESULTS OF QUESTIONNAIRE: RESPONSE TO
SELECTED QUESTIONS............................... 130

C LISTING OF THE 340 FIRMS INCLUDED IN THE TEST
SAMPLE......................... ................... 132

REFERENCES ..................... .................................... 136

BIOGRAPHICAL SKETCH.............................................. 142














Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


AN EMPIRICAL INVESTIGATION OF SOME ECONOMIC CONSEQUENCES
OF FASB STATEMENT NO. 8: FOREIGN CURRENCY TRANSLATION

Rohit Jain

August 1980



Chairman: A. Rashad Abdel-Khalik

Major Department: Accounting

In October 1975, the Financial Accounting Standards Board (FASB)

released Statement No. 8, "Accounting for the Translation of Foreign

Currency Transactions and Foreign Currency Financial Statements," the

implementation of which has been widely criticized. Critics in the

popular press claimed that FASB No. 8 causes increased volatility in

reported earnings and does not reflect the economic risk of exchange

rate fluctuations to a firm. Hence investors are supposedly misled,

resulting in lowered security prices and suboptimal decisions by managers

made solely to counteract the effects of FASB No. 8 on earnings.

The research objective of this dissertation was to investigate the

effects of FASB No. 8 on the security price behavior and the financial

structure of multinationals. Descritptive statistics on the translation

methods, reported exchange adjustments, and the variance of reported

earnings were also provided for our sample firms.










The effects of FASB No. 8 on security return distributions were

determined by both multivariate cross-sectional comparisons of relative

risk equalized total returns between multinational (test) and domestic

firms, and by time-series tests on average residual returns. The effect

on financial structure was determined by cross-sectional and time-series

tests on four balance sheet accounts (longterm debt, inventory, fixed

assets, current liability) and two income statement accounts (revenue,

income), each deflated by total assets.

The sample for statistical tests consisted of 281 New York Stock

Exchange (NYSE) multinationals with return data available from 1/1969-

12/1977 on the CRSP tape. Six different grouping schemes were used for

the test firms. Firms were primarily grouped in each period by the

translation method (current/noncurrent, hybrid, monetary/nonmonetary).

Additional variables used for partitioning were "change in exchange ad-

justment" (increase, decrease), magnitude of foreign revenues (large,

small), and "sign of exchange adjustment" (positive, negative) in

grouping schemes 2, 3, and 4 respectively. In grouping scheme 5,

portfolios were formed by the translation method used by firms in the

pre-FASB No. 8 period. Grouping scheme 6 consisted of "Action" firms

and "No Action" firms. Questionnaire data were used to form the group

of "Action" firms in which managements were judged to be taking action

to neutralize the effects of FASB No. 8 on earnings. "No Action"

firms consisted of the remaining MNC. The test period consisted of

1973-1974 and 1975-1976 as the pre and post-FASB No. 8 periods respec-

tively. Only firms that had minimum exchange adjustments of 1% of net

income were included in the tests in any particular year.










In the cross-sectional tests the vector of mean returns was not

significantly different for the test and control (domestic) firms in

any period for any of the grouping schemes. The covariance matrices

of test and control firms were significantly different in all periods

when firms were grouped by the magnitude of foreign revenues. For

other grouping schemes, the differences in covariances were con-

sidered insignificant. In the time-series tests, the means of the

average residuals were not significantly different over time for one

of the groups (scheme 5). For the other groups (schemes 1-4), the

residuals were significantly different in 1976 from 1975 and fairly

significantly different in 1975 from 1974 but there was no difference

between the residuals of 1974 and 1976. No explanation was provided

for the conflicting results. The overall conclusion was that in

general, there had not been any statistically significant effects

of FASB No. 8 on the security returns of multinationals. An

additional conclusion was that an uncontrolled variable, associated

with the magnitude of foreign revenues, was systematically affecting

the variance of returns. The tests on financial structure revealed

that the balance sheet structure of Action and domestic firms was

significantly different in the post-FASB No. 8 period prompting the

conclusion that FASB No. 8 had an effect on the financial structure

of selected firms.


viii










CHAPI 'R I
INTRODU: I'ION



In recent years, the involvement of many U.S. corporations in

foreign operations has expanded substantially. With the advent of

floating foreign currency exchange rates in 1971, reporting the

results of foreign operations in consolidated financial statements

had become an increasingly important task for multinational corpor-

ations (MNC, hereafter).

Pakkala (1975), reported that by the early 1970's MNC were

using an array of methods to "translate"--restate arithmetically

in terms of one base currency--the foreign currency statements of

international subsidiaries into U.S. dollars. To reduce the

diversity of methods, the Financial Accounting Standards Board,

issued in October 1975, Statement of Financial Accounting Standards

No. 8. "Accounting for the Translation of Foreign Currency

Transactions and Foreign Currency Financial Statements" (FASB No. 8,

hereafter).

Corporate leaders, financial analysts and academicians have

criticized FASB No. 8 in the financial press. Critics have claimed

that its effects on earnings are misleading and that it will result

in several adverse economic consequences, including

(1) lowered stock prices

(2) decisions by management made solely to eounteraet the

effect of FASB No. 8 on earnings

(3) adverse impact on the predictability of earnings.










The actual effects of the pronouncement are of concern to management,

investors and to policy making bodies like the Financial Accounting

Standards Board (FASB, hereafter). The FASB has called for research

on the economic consequences of existing and proposed standards.



Objectives of Research

The purpose of this research is an empirical investigation of

some economic consequences of FASB No. 8. Specifically, the

purpose is to study, for firms affected by FASB No. 8,

(1) the association of the announcement or implementation

of FASB No. 8 and security price behavior--the "effect"

on security prices.

(2) the effect on balance sheet structure; and

(3) provide descriptive statistics about the variability

of income and the accounting method used for translation

for the period 1973-1977.

The results of this research should provide information about

some economic consequences that the FASB may find useful in monitoring

the implementation and reviewing the effectiveness of FASB No. 8.

Also, if no economic consequences are found, the results should

settle the controversy surrounding FASB No. 8.

The remaining parts of this dissertation consist of a consideration

in the next chapter of some conceptual issues in foreign currency

translation, of the possible effects of FASB No. 8 on security returns

and on the structure of the balance sheet of MNC. The broad hypotheses










are developed and the research design is presented in Chapter III

and the results are provided and discussed in Chapter IV. The

conclusions and limitations of this study are discussed in Chapter V.








Note

1. See, for example Revzin (1976), Shank (1976) and Snyder (1977).













CHAPTER II
CONCEPTUAL ISSUES IN FOREIGN CURRENCY TRANSLATIONS



Fluctuations in Exchange Rates


Prior to 1971, the exchange rates between various currencies were

fixed under the then prevailing International Monetary System. The fixed

exchange rate system operated under the Bretton Woods treaty reached in

1944. The system called for stable exchange rates that were "pegged" to

par values. Countries could change the par value of their currency if

economic considerations warranted making an adjustment, but normally

countries tried to maintain the par value as long as possible. The U.S.

dollar was very important in that system as it was a reserve currency

and convertible to gold at $35 per ounce. As a result, until the late

sixties, exchange rates changed only rarely, and then foreign currencies

only devalued against the U.S. dollar.

The Deutsch mark appreciated against the dollar for the first time

in 1969. In 1970, the Canadian dollar was allowed to "float" and it also

appreciated against the U.S. dollar. In 1971, economic pressure led to a

devaluation of the U.S. dollar and the end to the Bretton Woods system as

some currencies were permitted to float against each other. In August

1971, the dollar was freed from the gold standard and it was further

devalued in 1973. By 1973, a floating exchange rate system was fully in

existence.

In the floating system, exchange rates are determined by market

forces and may change significantly every day. Giddy and Dufey (1975)








have found that under the current floating system, exchange rate changes

follow a random walk. Hence, short term predictions of rate changes are

not possible. To sum up, in the seventies, the U.S. dollar continued

to be weak and exchange rates fluctuated continuously and unpredictably

against each other.

If relative exchange rates stay fixed, MNC will not have any exchange

gains (losses) and the various translation methods (discussed later) will

give similar results. Under floating exchange rates, large exchange gains

(losses) are possible and mostly depend on the translation method used.



Accounting and Economic Risk Defined


Some definitions are provided next. Accounting risk (exposure) is

defined as the possibility that published financial reports will show

an impact from changes in exchange rates. Those assets and liabilities

that are translated at the rate which is in effect on the date of the

financial statement (the current rate) are said to bear accounting risk,

because currency movements will change the dollar value of those assets

and liabilities.

Economic exposure is defined as any and all effects of the rate

change that will affect the cash flow stream (and the long term survival)

of the MNC. There is no general agreement on the precise definition and

methods of measurement of the economic risk faced by a firm when exchange

rates change.I But it is clear that economic risk may be different from

the accounting risk faced by a firm for a given exchange rate change.

Consider an example adapted from Shank (1976). Suppose a Brazillian

subsidiary of a U.S. corporation is in a capital-intensive, high growth

business. It reports good earnings each year, but it always absorbs more








capital than it generates. It raises the needed capital and buys and sells

in Brazil. It has no direct dealings with the parent and pays no dividends.

Suppose further that Brazil's cruzeiro is devaluing in relation to the

dollar. The debt repayment thus involves no foreign currency translation

risk. Under the accounting methods that translate debt at current rates,

the subsidiary's cruzeiro debt is exposed in an accounting sense. If

the subsidiary were to borrow dollars instead of cruzeiros, there would

be no accounting exposure. There would, however, be economic exposure

because the subsidiary may have to pay different amounts of the cruzeiros

it generated to buy dollars for the repayment of the debt.

The fact that accounting risk is usually different from economic

risk has implications for measuring the economic impact of FASB No. 8.

These problems are discussed later, in the section dealing with the

effects of FASB No. 8 on management.



Reporting Methods for Foreign Currency Translations

The controversy in the literature has been basically about two

questions. First, which accounting method should be used for trans-

lating foreign currency statements, and second, how should the resulting

translation gain or loss be reported? The principal methods suggested

in the literature for the translation of foreign subsidiary statements

are (1) the current/noncurrent method (C), (2) monetary/nonmonetary

method (M), (3) the current rate method (CR) and (4) a modification of

the monetary/nonmonetary method (called the hybrid method (H) here).

Since the choice of the method for translation is not the focus of this

research, the various normative papers dealing with which translation
2
method should be used are not discussed here. A description of the

various methods follows:









Current/noncurrent Method (C). According to this method, all current

assets and current liabilities of foreign subsidiaries are translated at

the current rate (closing spot rate). All other asset, liability, and

equity accounts are translated at the exchange rate in effect when those

accounts were first booked, i.e., the historical rate. In the income

statement, all revenue and expense items are translated at the rates in

effect when those items were accrued (or average rates) except for

revenue and expense items relating to noncurrent assets and liabilities,

which should be translated at the same rates as the corresponding balance

sheet items.

The current/noncurrent method was first accepted by the AIOPA in

1931 and was recommended in the Accounting Research Bulletin No. 43,

Chapter 12. The assumption implicit in this method is that short-term

items are "exposed," whereas fixed assets and long-term debts are not.

Monetary/nonmonetary Method (M) was first suggested by Hepworth (1956).

In 1960, the National Association of Accountants recommended this method

in its report No. 36. Under this method, monetary assets and liabilities

are translated at the current exchange rate. Nonmonetary (physical) assets

are assumed to be free from exchange risk and are translated at their

historical rates. All income statement items are translated at average

exchange rates (usually monthly averages), except for items relating to

nonmonetary assets and liabilities, which are translated at historical

rates.

Hybrid Method (H). One objection raised against the monetary/

nonmonetary method has been with respect to the translation of inventories

at historical rates. Some companies feel that inventories, which are

steadily turned over, are as exposed as other monetary items. This gave

rise to the hybrid method, which is essentially the monetary/nonmonetary

mehtod with inventories translated at current rates as well.









Current Rate Method (CR). All assets, liabilities, revenues and

expenses are translated at current rates; under this method. Only owners'

equity accounts are translated at historical rates, hence only the net

equity of each subsidiary is assumed to be exposed. This method, though

widely used by European firms, was not used by many firms in the U.S.A.

While the monetary/nonmonetary and current/noncurrent methods emphasize the

parent company's viewpoint, the current rate method takes the viewpoint of

the foreign subsidiary. In the current rate method, as all accounts are

translated using one rate (the current rate), the relative magnitudes of

the various accounts in the dollar-statements are the same as in the original,

foreign-currency statements. The relative magnitudes of the various accounts

in the original statements are not maintained in the translated statements,

when other translation methods are used.

Temporal Method. Leonard Lorensen (1972), in Accounting Research Study

No. 12, put forth the temporal method of translation. Under this method,

assets and liabilities are classified by their basic temporality. Items

which are carried at present or future prices are translated at the current

rate. Those assets and liabilities that are carried at their historical

costs are translated at historical exchange rates. The exchange rate

appropriate to translation of inventory depends on its valuation method.

Inventory carried at cost is translated at historical rates, whereas if

inventory is carried at market or net realizable value or replacement cost,

it is translated at current rates. If lower of cost-or-market is used,

translated historical cost should be compared with translated market.

Under current generally accepted accounting rules with historical cost

accounting, the temporal method is essentially the same as the monetary/

nonmonetary method. There are some minor differences between the two








methods--inventory carried at net realizable value or replacement cost and

marketable securities carried at cont--which possibly could be important.

However, for all practical purposes, the two methods are treated as

equivalent and no distinction is made between them in this research.

FASB No. 1. In December 1973, the FASB issued Statement of Financial

Accounting Standards No. 1, "Disclosure of Foreign Currency Translation

Information." FASB No. 1 did not establish any new accounting standards,

but required extensive disclosures of the translation methods used and

disposition of exchange adjustments by affected firms. Broadly, the

statement required disclosure of

1) the accounting method used to translate income statement and

balance sheet accounts and the time of recognition of gain/

loss on forward exchange contracts;

2) the aggregate amount of exchange adjustment

(a) originating in the period, and the part thereof included

in income or deferred;

(b) included in income (regardless of when originated);

3) the aggregate amount of deferred exchange adjustments;

4) the change in value of long-term debt and receivables

translated at historical rates, if they were to be translated

at current rates;

5) the unrecognized gain/loss on unperformed forward exchange

contracts.

FASB No. 1 was effective for firms with fiscal years ended in December

1973. The fact that the information required under FASB No. 1 was inade-

quately disclosed can be seen from the annual survey of 600 companies

conducted by the AICPA (1973). In 1972, only 80 firms disclosed the dollar










value of the translation adjustments. While 164 other firms disclosed

their translation policies, they did not disclose the gain/loss from

translation.

FASB No. 8. On December 31, 1974, the FASB issued an exposure draft,

"Accounting for the Translation of Foreign Currency Transactions and

Foreign Currency Financial Statements." This exposure draft was finalized

as the FASB Statement No. 8 in October 1975. FASB No. 8 requires a uniform

method of translating foreign currency transactions and financial statements.

The rules apply to financial statements of foreign affiliates of U.S. com-

panies which are either consolidated in U.S. statements or are accounted

for by the equity method. The major provisions of FASB No. 8 are

(a) Foreign currency transactions are to be recorded at the rate

in effect on the transaction date.

(b) Financial statements are to be translated using the "temporal"

method.

(c) With the exception of gains (losses) on certain forward exchange

contracts, exchange gains and losses are to be reflected in

operations for the period in which the exchange rate changes.

(d) The aggregate exchange gain (loss) included in income for the

period is to be disclosed.

(e) The statement is effective for fiscal years beginning on or

after January 1, 1976.

In this research, the translation method required under FASB No. 8

(temporal method) is assumed to be equivalent to the monetary/nonmonetary

method. The three main translation methods used by U.S. companies are

compared below.










Comparison of Main Translation Methods. The exchange rates used to

translate the various accounts in the monetary/nonmonetary, current/

noncurrent and hybrid translation methods are shown in Table 1.


Table 1

Comparison of Exchange Rates Used


in Three Translation Methods


Account


Inventory

Other Current Assets

Long Term Assets

Current Liabilities

Long Term Debt (LTD)

Apart from common
items, exposed are:
aCR = Current Rate


Current/
Noncurrent
CRa

CR

HR

CR

HR


Inventory


Monetary/
Nonmonetary
HRb

CR

HR

CR

CR


LTD


Hybrid


CR

CR

HR

CR

CR

Inventory
& LTD


bHR = Historical Rate


From Table 1, it can be-seen that inventory and long-term debt are the

only two accounts that are affected by the different translation methods.

Apart from the common items that are exposed under all three translation

methods (other current assets and current liabilities), the accounts that

are exposed under each method are also shown.

Pakkala (1975) gives the weight of the various accounts in a typical

subsidiary balance sheet. It is shown in Table 2, with a slight change.

Based on these account weights, the accounting exposed position and the

corresponding effect of the devaluation or appreciation of the U.S. dollar

on translation adjustment can be calculated for each of the three translation

methods. Under the current/noncurrent method, the firm has an exposed net


V









asset position, whereas it has an exposed net liability position under the

monetary/nonmonetary method. For a given exchange rate change, the

translation adjustment is correspondingly opposite under the two methods.


Table 2

A Representative Balance Sheet


Assets Equities

Other Current Assets 28% Current Liabilities 25%
Inventory 24% Long Term Debt 20%
Other 5%
Net Plant 40%
Other Fixed Assets 8% Owners Equity 50%
Total Assets 100% Total Equities 100%

Source: Pakkala (1975)


Under the hybrid method, the exposed position is in between that of the other

two accounting methods. The implications of the three translation methods

for the hypothetical firm are summarized in Table 3.


Comparison


Table 3

of Accounting Exposure in Various Translation Methods


Effect of $
Accounting Method Exposed Position Appreciation Devaluation

Current/
Noncurrent 27% (Net Assets) Loss Gain

Hybrid 2% (Net Assets) Small Loss Small Gain
Monetary/
Nonmonetary 22% (Net Liabilities) Gain Loss



The impact of a change to FASB No. 8--defined as the difference in

reported earnings under the new method and the earnings based on the pre-

viously used translation method--depends on the previously used translation

method, the structure of the balance sheet and the change in exchange rates









that took place. The effect of a change to the monetary/nonmonetary

method on the earnings of the hypothetical firm shown in Table 2 is

shown in Table 4.



Table 4

Effect of Change to M/MN Method on Earnings of Hypothetical Firm



Effect on Earnings
Due to $
Previous Accounting Method Appreciation Devaluation

Current/Noncurrent Large Increase Large Decrease

Hybrid Increase Decrease

Monetary/Nonmonetary No Change No Change


The firms previously using the current/noncurrent method would be most

affected, whereas the firms using the monetary/nonmonetary method would not

be affected at all by the adoption of FASB No. 8.

Treatment of Exchange Adjustments. The second controversy concerning

the choice of the reporting method deals with the treatment of the unrealized

exchange adjustment (gain/loss) due to the translation of foreign currency

financial statements. The treatment of realized exchange adjustments, which

arises when payments or receipts are made in foreign currency, is not

controversial because realized exchange adjustments are immediately included

in income. Exchange adjustments due to translation can be immediately

recognized in income or deferred and included in income over time. The

principal alternatives are (1) to include the exchange gain/losses in income

in each period (2) to defer both the gains and losses in a balance sheet

reserve account and (3) to include only the losses (in excess of previously










deferred gains) in income, but defer gains. Prior to FASB No. 8, firms

used one of the above three methods or a combination of all the three

methods.

FASB No. 8 requires that firms include the unrealized exchange

adjustment in income each period and that reserve accounts are no longer

permitted. The reaction of the financial community to this requirement

of the FASB is discussed in the next section.



Possible Effects of FASB No. 8

Adoption of FASB No. 8 may have effects on the behavior of security

price returns and on the structure of the balance sheets of MNC for

reasons that may be characterized as investor behavior and management

behavior. These two reasons are discussed in detail below.

Investor Behavior. Gonedes (1978) describes investor behavior in

the following way:

The distributions of future values of firms' securities
define the gambles available in the capital market.
These distributions are, therefore, among the ultimate
objects of analysis when agents are making optimal
consumption-investment decisions and, thus, when firm's
securities are "priced out" in the capital market. (p. 29)

The question then is, did FASB No. 8 have any effect on market agents'

assessment of distributions of returns of MNC1 One advantage claimed for

FASB No. 8 is consistency in accounting for foreign currency translations.

Prior to FASB No. 8, firms were using three different methods or some

modifications of them. This could have caused difficulties in the com-

parison of financial statements of firms in a meaningful way. Aside from

this advantage, FASB No. 8 has been basically criticized on two points.








First, FASB No. 8 does not require providing information to measure the

economic exposure of MNC. Second, the inclusion of unrealized translation

adjustments in income is likely to increase the variability of income.

These points are discussed below.

Shank (1976) and Radebaugh (1974), among others, point out that some

reported translation gains/losses will never be realized, especially for

independent autonomous subsidiaries with significant local financing.

Hence, the exchange adjustment reported in conformity with FASB No. 8 may

not reflect the economic exposure of the firm. Even though the FASB

clearly said in the discussion to FASB No. 8 (para. 96-111) that they are

not trying to assess the economic effect of rate changes, critics feel

that the reported gain/losses may be construed by the market as economic,

rather than accounting effects. In a letter to the FASB, for example,

Rexford Bruno, Senior Vice President of UAL, Inc., said:

We believe there is an inequitable, misleading, unrealistic
and unintended situation arising from application of
FASB No. 8. ...we believe that the application of
Statement No. 8 in Western's situation does not reflect
economic reality when, in fact, there is limited economic
exposure to Western. (May 31, 1978)

Hoyt (1972) and Prindl (1976) are among those who claim that the actual

economic effects of a rate change are difficult to determine. Foreign

currency translation is a complicated area and even expert analysts may

disagree as to what constitutes an economic change to a firm. The reason

is that economic effects are conditioned on all facets of a company's

operations including decisions such as the extent to which the selling price

of goods can be increased. The actual tax effects of a currency change are

difficult to determine due to the differences in tax laws in various

countries. Also, information about economic effects may not be available

to the market from other published sources.









With the advent of floating exchange rates, large random changes in

the translation adjustment may occur and their inclusion in income may

cause an increase in variability of earnings, especially for quarterly

earnings. Some managers and authors felt that this variability in

reported earnings will hurt firms in the stock market. "Executives have

been grumbling that this new procedure increases the volatility of their

companies' reported profits, confuses investors and, consequently, puts

a damper on stock prices" (Snyder, 1977, p. 85). To sum up, the impli-

cation is that a company's reported profits are likely to be distorted

by the implementation of FASB No. 8, and that distortion is likely to

hinder the ability of investors to correctly assess the distributions of

future values of firms' securities. A consideration of modern finance

theory leads to a different view, however.

The semi-strong form of the efficient market hypothesis states that

all relevant, publicly available information is quickly and unbiasedly

incorporated in the market prices of securities. Fama (1970) reviews

many studies which conclude that capital markets are informationally

efficient. Kaplan (1978) and Gonedes and Dopuch (1974) extensively

review research that measures the effect of changes in accounting tech-

niques on stock prices. The evidence shows that accounting changes are

not associated with changes in stock prices, unless the accounting

change has real cash flow implications. The accounting changes required

by FASB No. 8 do not in themselves have any such consequences; so, the

stock prices of multinationals should not be affected--at least not due

to any distortion or increase in the volatility of reported earnings.

Kaplan also reports on studies that have found that as long as adequate










disclosure is made, it does not matter if the disclosure is made in

footnotes, in the income statement or the balance sheet. Consequently,

the disclosure of the translation adjustment as part of income instead

of being part of a reserve account should not be expected to have any

effect on the stock price behavior of MNC. Had FASB No. 8 required the

disclosure of previously publicly unavailable information, it would have

been expected to have an effect on the security price behavior of firms.

Whereas FASB No. 1 may have required the disclosure of important new

information useful for valuing the securities of MNC, it is assumed in

this research that FASB No. 8 did not call for providing any significant

information, beyond what had been required by FASB No. 1, that would be

useful to market agents in analyzing security prices.

One possible effect of the disclousre of translation adjustments

as part of income, required under FASB No. 8, could be the lowering of

credit ratings of MNC by agencies that use the stability of earnings

as a factor in ratings. Some firms may also run into difficulties with

restrictive debt covenants that might require refinancing under un-

favorable terms. If such is the case, an efficient capital market may

react adversely to the affected firm's securities. However, sufficient

data are lacking to identify and test these effects on a specific set

of firms.

Management Behavior. Implementation of FASB No. 8 may indirectly

impact on the security price behavior and the structure of the balance

sheets of MNC, through the behavior of management. How investors

actually react to different accounting methods and disclosures may be

quite different from management's perception of how investors behave.










If managers believe that investors ire fooled by accounting changes or by

the inclusion of exchange adjustments in income, that is, the capital

market is not efficient in processing information, they may take action

to minimize the accounting risk due to FASB No. 8. There is some evidence

that management indeed believe the capital market to be inefficient.

Senior Vice President of UAL, Inc., Rexford Bruno, in a letter to the

FASB (May 1978):

From another standpoint, our management thinking regarding
foreign investment has changed significantly as a result
of Statement No. 8. No longer do we evaluate an investment
based only on its economic merit. We must now evaluate
the potential for financial statement fluctuations caused
by unrealized foreign currency translation gains and
losses and what impact they may have on users of financial
statements. It is our opinion that the inclusion of
unrealized gains and losses in earnings is definitely
misleading to all but perhaps the most sophisticated user.


Citibank, in its Monthly Economic Letter (May, 1978): "Companies are

reported to be borrowing or lending in various currencies with the objective,

at least in part, of reducing their exposure under FASB #8" (p. 8).

Managements may also be motivated to minimize the impact of FASB No. 8

on reported earnings due to its effect on managements' reward structure

(bonuses, stock options, etc.). In a questionnaire used in this study

(discussed later) 18 out of 100 firms reported that managers' rewards are
3
linked to the accounting effect of FASB No. 8. The results of the

questionnaire show that in 1975, 1976, and 1977 the number of firms that

had taken action to cover accounting exposure was 11, 16, and 24 respectively.

Hedging Methods. To counter the effects of exchange rate fluctuations,

firms can use a number of techniques. These techniques can be divided










into two main categories. The firrt category consists of various operating

and financing decisions. The second category consists of the use of forward

exchange contracts. The operating and financing techniques include local

currency borrowing and the management of leads and lags in trade settlement.

As the name implies, leads and lags refer to a firm's ability to expedite

or delay payments, receipts, transfer of dividends, or transfer of other

funds across countries. Firms with subsidiaries in a number of countries

may also try to set transfer prices to maximize total company dollar profits.

Of course, the choice of the country in which foreign direct investment is

to be made, or from which goods are bought or sold is a major long range

decision.

The second method available to firms is to hedge their positions by

entering into forward exchange contracts to buy or sell foreign currency

at previously contracted rates. The market for foreign currency is similar

to the market for commodities. In the forward exchange market, companies

can, for most major currencies, buy or sell foreign currency for future

delivery. The general principle is as follows:

(a) To hedge net exposed foreign liability position or pay foreign

currency in future--purchase foreign currency forward contracts

now, and

(b) To hedge net exposed foreign asset position or receive foreign

currency in future--sell foreign currency forward contracts

now.

Net exposed foreign asset/liability position in an accounting sense is

defined as net assets (assets--liabilities) translated at current exchange

rates.









Firms have usually hedged foreign currency in order to protect the

economic value of the firm. The amount of hedging would generally depend

upon the extent of the firm's ability or desire to bear economic risk in

the short term. Nowadays, for firms concerned about the reporting effects

of FASB No. 8 on market agents, hedging the accounting exposure risks can

neutralize the consequent gain/loss to be reported--thus actually causing

a change in the firm's economic value by attempting to neutralize a re-

porting artifact.

This hedging of accounting risk causes an economic impact on a firm

that is expected to affect the security price behavior of the firm.

Whereas the security returns of firms hedging accounting risk are expected

to be negatively affected in the long term (5-10 year period), the

directional effects on security prices cannot be specified on a year-to-year

basis. The reason is that a forward exchange contract made to cover

accounting exposure may result in short run real profits or losses to the

firm, depending on the realized currency movements of the contract. Firm

specific data about forward exchange contracts are not publicly available.

Effects on Balance Sheet Structure. Management behavior to counteract

the effects of FASB No. 8, to the extent that it took place, may also

affect the balance sheet structure of the corresponding firms. This is

possible if managements change their operating financing decisions in

response to FASB No. 8. In a study sponsored by the FASB, Evans, Folk, and

Jilling (1978) assessed the impact of FASB No. 8 on the foreign exchange

risk management (FERM, hereafter) practices of MNC. They sent question-

naires to 430 MNC of which 156 completed the questionnaires.









Some of their findings are:

(1) Firms that changed accounting method are more aggressive in

FERM and are still adapting to FASB No. 8.

(2) FASB No. 8 had an impact on firms' FERM practices.

(3) The major adjustments in FERM methods include changing foreign

currency borrowing levels, changes in subsidiary divident

payment and use of forward exchange contracts.

(4) FASB No. 8 had an impact on firms' investment policies.

Due to lack of specific data, we cannot specify any expected directional

effects of management behavior on firms' balance sheet structure.

Review of Previous Empirical Studies on FASB No. 8

Rodriguez (1977) examined the impact of FASB No. 8 on reported

earnings of firms in 1974 and 1975. Basically, she surveyed the annual

reports of 70 multinational corporations. Rodriguez found that in only

23 firms did FASB No. 8 have a "material" impact on reported earnings

measured as the difference in reported earnings under FASB No. 8 and the

earnings based on the previously used translation method. Ten of the 23

firms had an impact on earnings by less than 5%. The author concluded

that the adoption of FASB No. 8 did not appear to have significantly

changed the earnings reported by most MNC. In a review article,

Griffin (1979) pointed out that Rodriguez's study is based on earnings of

companies that had adopted FASB No. 8 voluntarily prior to the effective

date. It is therefore possible that the firms surveyed are likely to be

the least affected by the switch to FASB No. 8.

In a study sponsored by the FASB, Dukes (1978) investigated the effects

of FASB No. 8 on the security return behavior of MNC. On the basis of three









test periods and five methods of analysis, Dukes concluded that the

"issuance and implementation of FASB Statement No. 8 does not appear to

have had significant detectable effects on the security returns of

multinational firms" (p. 3). The approach and methodology used by Dukes

are in many ways similar to those used here. However, many additional

tests were performed here.

Dukes selected 479 New York Stock Exchange firms that had security

return data available from January 1965 through December 1976, as his

information sample. He classified the firms into six groups according

to the monetary/nonmonetary, current/noncurrent, and hybrid accounting

methods and by whether the exchange adjustment was included in income

or deferred by firms. He obtained the accounting method used by firms

from Disclosure Journal (1975), which has data from the 1974 annual reports.

One possible error in this classification method, recognized by Dukes,

is that firms in years prior to 1974 could have been using different

translation methods. Dukes did not change the composition of portfolios

in 1975 or 1976, but about half the firms switched to the monetary/non-

monetary method in 1976. So, his portfolios do not consist of firms using

the same accounting method throughout in 1975 and 1976.

The three test periods studied by Dukes were (1) January 1968-

December 1969 (2) January 1970-December 1974 (3) January 1975-December 1976.

As FASB No. 1 became effective on December 1973 and presumably provided

new disclosures starting in 1973, the years 1968-1972 may not be appropriate

control periods.

For tests on portfolios, Dukes first compared the means of returns of

MNC firms with those of domestic firms and found no significant difference










in each of the three test periods. Second, he used the market portfolio

as a control and found no significant difference between MNC and the

market portfolio in period 3. However, he did find a significant dif-

ference in periods 1 and 2. Also, when he compared MNC previously

following FASB No. 8 (his portfolio No. 3) with other MNC he found a

significant difference in test period 1. As he found that return distri-

butions of test firms were different from those of control firms in the

pre-FASB No. 8 period, for two testing methods, conclusions about the

test results of the post-FASB No. 8 period may be less than accurate.

At the individual security level, Dukes conducted two tests. First,

he used the Chow test to see if the regression model's parameters had

changed for the market model. He used the years 1972-1975 as the pre-FASB

No. 8 period and 1976 as the post-FASB No. 8 period, and found no

statistical difference in the model structure. As about half the firms

switched to FASB No. 8 in 1975 and as Dukes himself included 1975 in the

post-FASB No. 8 period in his first three methods of analysis, the Chow

test results must be interpreted with caution. Second, Dukes compared

the variability of the return residuals of MNC for 1975 & 1976. Again,

he found no increase in the variability of the return residuals. The

criticism made in the case of the Chow test holds here also. As about

one half of the MNC switched to FASB No. 8 in 1975 and the other one

half in 1976, any increase in variability of return residuals due to

FASB No. 8 would not be detectable by Duke's procedure. So, his con-

clusion that "both individual security level tests suggest that the number

of firms whose security returns were measurably affected by Statement

No. 8 is small, if indeed they can be identified at all" may be premature.


1









Finally, Dukes compared the returns of the MNC having a high

magnitude of foreign operations with those of MNC having a low magnitude

of foreign operations. He used (a) foreign sales as a percentage of

total sales (b) foreign assets as a percentage of total assets and

(c) exchange adjustment as a percentage of operating income as alternative

measures of magnitude of foreign operations. His test period was January

1975 to December 1977 in each case. The information on the magnitude

of foreign operations was obtained from 1976 annual reports, and due to

inadequate sample size he left out firms using the monetary/nonmonetary

method (his groups 3 and 4) from the statistical tests. He found that

foreign assets was the only measure of magnitude that produced significant

results at the 10% level. While foreign sales and foreign assets are

appropriate measures of the magnitude of foreign operations, the exchange

adjustment does not measure the magnitude of foreign operations because

the exchange adjustment reported by a firm is a function of the translation

method used, the balance sheet structure of the company, hedging behavior

of the company and movement in foreign exchange rates. It is interesting

to note that in his analyses of the magnitude of foreign operations, both

for exchange adjustments (his Table 13, page 58) and for foreign revenues

(his Exhibit C5, page 95), Dukes found unusual security behavior in 1977.

The cumulative return differences increased dramatically in 1977. No

good explanation for this difference has thus far been provided.

Fredrikson and Mogus (1978) examined the security return behavior of

MNC to determine if the returns of firms that were required to change their

translation methods were statistically different from the returns of firms

that were already using a translation method that approximated the FASB No. 8

method. They selected 400 firms that were listed on the PDE compustat tape









and whose translation method prior to FASB No. 8 was known; it is not

clear exactly which year's annual reports were used to determine the

translation methods. The 400 firms were classified into four groups by

translation method (monetary/nonmonetary and current/noncurrent) and

treatment of the exchange adjustment (recognize currently and deferral).

The firms using the monetary/nonmonetary method with current recognition

of the exchange adjustments were treated as the control group, against

which the other three groups were compared.

The test period was the 24 month period November 1975 through

October 1976. The authors compared the average residual returns of

the three test groups against the control group. Using the F test and

Dunett's t, the authors found no significant difference between the

security returns of the test and control groups.

The authors also partitioned and compared firms by size of foreign

revenue (more than 30% vs less than 20%), firm size (more than $1000 M vs

less than $300 M), number ofcountries (more than 12 vs less than 5) and

industry (4 groups). In each case, they found no significant effect of

FASB No. 8 on security prices.

Makin (1978) examined the impact of FASB No. 8 on security returns

of MNC. Makin used three groups of firms, classified into domestic, MNC,

and MNC "sensitive" to FASB No. 8, and five test periods. Using the

effect on market model parameters to infer the impact of FASB No. 8, he

found that the cost of capital of the "sensitive" group was slightly

adversely affected. A commentator has noted that there are some method-

ological weaknesses in this study. Some firms included in the "sensitive"


1










group were using the monetary/nonmonetary method with immediate recognition

of exchange adjustment prior to FASB No. 8, and consequently should have

been unaffected by FASB No. 8.







Notes


1. See, for instance, Giddy (1977), Prindl (1976) and Shank (1976)
for various measurements of economic exposure.

2. Some papers dealing with the choice of the translation method are:
Fredrikson (1973), Gillick (1974), Hayes (1972), Lorensen (1972),
Pakkala (1975), Parkinson (1972), Radebaugh (1974), Seidler (1972),
and Shwyader (1972).

3. The results of the questionnaire are given in Appendix B.














CHAPTER III
RESEARCH DESIGN


In the previous chapter, the possible effects of FASB No. 8 on

security return behavior and on the balance sheet structure of MNC were

discussed. The broad hypothesis about expected FASB No. 8 effects on

security price distributions and balance sheet structure are developed

next. Then, the theory and measurement methods of any effects on

security returns are presented in this chapter.


Hypotheses

The possible hypotheses that follow from the discussion in Chapter

II, formulated in the null form, are:

(1) FASB No. 8 had no impact on the security return distributions

of MNC as market agents are not misled by the inclusion of

unrealized translation adjustments in earnings and the

resultant increased volatility of earnings. (The market is

informationallly efficient)

(2) Capital markets are efficient and FASB No. 8 had no impact on

the security return distributions of MNC because there was no

release to the capital market of new information useful in

assessing the return distributions of MNC.

(3) Capital markets are efficient and FASB No. 8 had no impact

on the security return distributions of MNC only because there

were no actions to neutralize the effects of FASB No. 8 on

reported earnings (accounting risk).








(4) Capital markets are efficient and FASB No. 8 had no impact

on the security return distributions of MNC because there

were no effects of FASB No. 8 on firms' financing activities

through restrictive debt covenants and adverse credit ratings.

(5) FASB No. 8 had no impact on the structure of the balance sheet

of MNC, because there were no management actions to neutralize

accounting risk.

For reasons discussed in the previous chapter, the null forms of hypotheses

(1) and (2) are assumed to be correct in this research. However, it should

be noted that in the testing of hypotheses (3), (4) and (5), the tests

are really joint evaluations of those hypotheses and hypotheses (1) and (2).

Based on the discussion in Chapter II, it is reasonable to expect

the rejection of hypotheses (3) and (4). If hypothesis (3) is rejected,

the alternative hypothesis is that FASB No. 8 had an impact on the security

return distributions of MNC because management took actions to neutralize

accounting risk. This impact on the security returns of MNC may be positive

or negative as there is no a priori basis to expect a directional impact.

If hypothesis (4) is rejected, the alternative hypothesis is that

FASB No. 8 had a negative impact on the security return distributions of

MNC. However, in this research, no attempt is made to test hypothesis (4).

Therefore, if a negative impact on the security returns of MNC is observed

in empirical testing, it will be impossible to differentiate between

hypothesis (3) and hypothesis (4). But as hypothesis (4) is expected to

affect relatively few firms, any observed impact may be assumed to be due

to the rejection of hypothesis (3).

Some a priori support also exists for the rejection of hypothesis (5).

Again, no expected directional impact can be specified for the alternative

hypothesis.












Measuring Effects on Security Returns

In measuring the effects of accounting changes on security price

behavior, the assumption made is that the assessment of the probability

distribution of returns is the object of interest to market agents. The

semi-strong form of the informational efficiency of the cpaital market

is a maintained hypothesis. This hypothesis states that all publicly

available information (relevant to valuing firms) is rapidly and

unbiasedly incorporated in security prices.

Given a capital market in equilibrium, a new disclosure is defined

to have information content if the distribution of returns of the

corresponding security is affected. One version of the Sharp-Lintner

two-parameter capital asset pricing model relates expected returns and

relative risks in equilibrium as


E(Rit) = Rft + (E(Rmt) Rft) Bit

where Rit = Rate of return on security i for their period t

Rmt = Rate of return on the market portfolio

Rft = Risk-free rate of return

Bit = Cov (Rit, Rmt)/Var(Rmt) = relative risk

E = the expectation operator, and tildes denote random variables.

This model states that the equilibrium expected return on an asset is

a linear function of its relative risk, Bi and that two assets with the

same Bi must have the same expected return. This property is exploited

in both cross-sectional tests and time-series tests to determine the

information content of FASB No. 8.









Cross-sectional Tests on Security Returns. The approach used here

was introduced by Gonedes (1975). His design, based on multivariate

comparisons of vectors of mean returns, has been used by Harrison (1977),

Ro (1978) and Ajinkya (1978) among others. According to this approach

the actual returns of a number of relative risk equalized portfolios

are simultaneously compared cross-sectionally. The main points of the

method are as follows:

Let Oit be a variable conveying new information pertinent to valuing

firm i at time t. Then the information content of Ot can be determined

by comparing the distribution function of returns, Rt conditional on two

realizations (t 1, t2) of Ot. Let G stand for the multivariate distri-

bution function of returns Rt. If the assets have equal systematic risk,

B*, obtaining G(Rt|Otl) # G(Rtlet2) J G(Rt) implies Ot has information

content. Assuming the distributions to be multivariate normal, the

comparison reduces to testing for differences in the mean vectors and the

variance-covariance matrices of the two distributions. The appropriate

test for differences of mean vectors is the Hotelling's T2 test, and the

test for differences in covariance matrices is Box's x2 test. The

construction of the time series of relative risk equalized returns is

explained in a later section.

In this study, three different variables are independently considered

to see if there is any interaction between them and the accounting method

for translation. The first variable is 'change in exchange adjustment' (CEA),

defined as:

CEA = (Exchange Adjustment)t (Exchange Adjustment)t_1 where t=year.

It is assumed that the knowledge of the exchange adjustment is somehow









useful to investors in setting security prices and it is assumed that

investors try to predict the amount of gain (loss) due to the exchange

adjustment component. It is possible that our assumption about forecasts

of CEA by investors is unwarranted, in which case this variable will not

have any effect in the tests. We further assume that the exchange adjust-

ments follow a random walk and that investors use a naive prediction

model, i.e.

E (Exchange adjustment)t = (Exchange adjustment)t_1


where E = Expectation operator

t = Year.

Therefore CEA is a measure of the forecast error for exchange adjustments,

which may have information content. In this study, only the sign of the

CEA is used for simplicity. So, CEA has only two levels, "increase" and

"decrease," depending on whether the reported exchange adjustment increases

or decreases in period t, for a particular firm.

The second variable considered is the sign of the exchange adjustment

(SEA) reported in any period t, and this has two levels (positive or

negative). In previous research, the errors in earnings forecasts have

been shown to have information content (Ball & Brown, 1968). After the

implementation of FASB No. 8, the earnings of MNC include the exchange

adjustment. As the magnitude and sign of the exchange adjustment are

difficult to forecast, and usually unknown to market agents before financial

reports are released, the disclosure of the exchange adjustment (which

is a component of net income) may be assumed to modify the earnings

forecast. So, if investors take the exchange adjustment into account in

setting security prices, there may be differential reaction to firms









having positive and negative exchange adjustments as investors will

be able to use this information to modify their forecast of earnings

without the exchange adjustment.

Interaction between translation methods and the size of foreign

revenues (FR) of a firm is also hypothesized, based on previous research

by Robbins & Staubaugh (1972) finding a definite relation between the

magnitude of foreign operations and management's attitude toward risk

and choice of hedging methods. So, different magnitudes of FR are

used as a surrogate for different management attitudes toward accounting

risk, and an effort is made to isolate management's reaction through

subgrouping by size of FR. Robbins and Staubaugh classified firms with

foreign revenues greater than $500 million as large, firms with foreign

sales between $100 million and $500 million as medium, and firms with

foreign revenues less than $100 million as small. In this study, firms

with foreign sales greater than $200 million are classified as large,

while other firms are considered small. The latter classification scheme

was chosen to obtain approximately equal numbers of firms in the large

and small categories.

Time Series Tests on Security Returns. In previous studies using

the Gonedes methodology, the test groups are assumed to be invariant

over time. In this study, we would like to test the information content

of various groups of firms before and after the change in accounting

method. The cross-sectional tests discussed above cannot be used unless

some modification is made to remove the effect of changing market

conditions, measured by Rmt. An alternative approach, introduced by










Ajinkya (1978), is used instead. The method is essentially a multivariate

analog of the univariate cumulative average residuals (CAR) procedures

used widely (Fama, Fisher, Jensen & Roll (1969); Abdel-Khalik & McKeown (1978).

One commonly used form of the asset pricing model, based on ex post

data, is called the "market model":

Rit = ai + BiRmt + uit

where Bi = estimated relative risk parameter

ai = estimated intercept

Rit = the observed return on security i in period t

Rmt = the observed return on a market index in period t

uit = the unexpected component of total return Rit

For a portfolio of N securities, the average residual return is given

by:
N
Ut= N- uit, where t = 1, 2,...,T, the test period.
i=1

Subject to the assumptions of the market model, this Ut metric is free of

market influences and can be used for comparison of test (MNC) firms
2
before and after the accounting change. As in the cross-sectional tests,

the distribution functions of the average residuals are compared simul-

taneously, for the various groups of interest. The method of obtaining

the average residuals is explained in a later section.




Problems in Measuring Effects of FASB No. 8

In the procedures described above to measure the effects of FASB

No. 8 on security return behavior, it is implicitly assumed that no other

information relevant to valuing firms is being simultaneously released

to the market, or that the additional information is being randomized









among the firms in a particular test portfolio. These are two factors

that may cause problems. The first is the economic effect of currency

fluctuations on MNC and the second is the effect of the annual earnings

announcements.

Effect of Currency Fluctuations. A devaluation (revaluation) of

a foreign currency has important implications for the profitability

of a firm with a subsidiary in that country. Under the regime of

floating exchange rates that started in 1971, these changes in exchange

rates have become continuous, even though the magnitude of the change

in any period may have become much smaller than the sharp changes that

sometimes occurred in the days of the fixed parity system.

The economic effect of changes in exchange rates on a particular

firm depends on a number of factors, including changes in prices of

inputs and outputs, taxes, interest rates, financial structure of the

firm and government controls on foreign currency transfers. In

evaluating the effects of FASB No. 8, it would be ideal if the real

effects of floating exchange rates on a firm in each period could

be isolated. The problem is both the lack of an operational model

that adequately links changes in exchange rates to the values of firms,

and detailed firm-specific data.

Consequently, the assumption made in our analysis (and implicitly

in other studies on the impact of FASB No. 8) is that changes in exchange

rates are randomized for the portfolio of MNC. This assumption may

not be unreasonable, given that each firm usually deals in many currencies

and that each portfolio consists of many firms.










Effect of Annual Earnings Announcements. Many studies have docu-

mented the information content of the annual earnings numbers. Kaplan

(1978) reviews the findings of the major studies, and concludes that

the studies provide fairly convincing evidence that the earnings numbers

have information content. Gonedes (1978), in testing the joint effects

of annual dividend and extraordinary items with annual income signals

concluded that the dividend and extraordinary items had no information

content beyond what is included in the earnings numbers. The implication

for this study is that the effect of earnings announcements may bias the

results about the effects of FASB No. 8. In this research, it is assumed

that the effect of earnings announcements will be randomized for each

portfolio of test and control firms.



Data

The initial objective was to identify a sample of MNC that would

be affected by FASB No. 8. At this stage, one restriction on the sample

was that data for the firm were continuously available on the CRSP tape

from January, 1969, through December, 1977. Second, for convenience

only New York Stock Exchange (NYSE) firms were selected. Disclosure

Journal (1974) was used to identify firms under the categories of

"current/noncurrent method," "monetary currency translation method," etc.

This list was further augmented by a list of firms that were included in

Accounting Trends and Techniques and had foreign currency translation

disclosures. Firms included in a Peat, Marwick, Mitchell & Co. survey (1977)

and those listed in a survey by Vaupel and Curhan (1968) were also

included. This selection process resulted in 427 firms. A questionnaire,

described below, was mailed to these 427 firms. A copy of the questionnaire










and cover letter is included in Appendix A. The principal data requested

in the questionnaire for each of the years 1970-1977 were:

(1) the exchange adjustment for each year, disaggregated by

gains (losses) due to translation, transaction and

forward exchange contracts.

(2) the accounting method used for translation of foreign

currency statements.

(3) the magnitude of the foreign revenues of the firm.

(4) whether the firm hedges against the accounting risk asso-

ciated with currency translations, and the methods used

for such hedging.

One hundred firms responded to the questionnaire, with most firms

providing all the data requested. Summary responses to some of the

questions not directly used in this research are given in appendix B.

Only one mailing was sent, without any follow-up letters. As in all

questionnaires, there may be-some response-rate and self-selection bias.

While the response rate was not poor (24%) for a survey of this type, a

sample size of 100 firms was considered inadequate for this study. To

further increase the sample size, annual reports kept at the University

of Florida library were used to manually collect data on firms on which

adequate questionnaire responses were unavailable. The data collected

for each of the years 1973-1977 from the annual reports were:

(1) the accounting method used for translation;

(2) the magnitude of the exchange adjustments;

(3) the magnitude of the foreign revenues;

(4) whether the firm took action to neutralize accounting risk.










As most firms did not provide an unambiguous breakdown of realized

(transaction) and unrealized (translation) exchange adjustments, only

total exchange adjustments were used in the analysis. A total of 340

firms (inclusive of the 100 firms that responded to the questionnaire)

with some or all of the required data, were selected at this stage. A

listing of these 340 firms is provided in appendix C.

For some firms, annual reports for particular years were unavailable.

Also, sometimes, the magnitude of the exchange adjustment was mentioned

to be "immaterial" or not mentioned at all. In both circumstances,

the exchange adjustment was coded as zero. That is, a zero exchange

adjustment could imply a missing report, or an "immaterial" exchange

adjustment. Of the 340 firms for which data were available the exchange

adjustment was recorded as zero for 55 firms, in each of the five years

of interest (1973-1977). Due to the filter used (discussed later) on

the magnitude of exchange adjustments, these 55 firms were excluded from

all of the statistical analysis except for descriptive statistics on the

data. An additional 4 firms had a January or February fiscal year end

and were removed (for reasons explained in the section on time periods

on tests), leaving a final sample of 281 test (MNC) firms.

Forty-four of the 281 multinational firms selected (15.5%) did not

disclose their foreign revenues in any of the five years. For these

firms, the foreign revenues were estimated by this researcher on a

case-by-case basis. If the foreign revenue was available for at least

one year, that revenue number was used for the remaining years. So, while

the data on foreign revenues are softer than the data on exchange adjust-

ment and accounting method, their importance is limited because foreign











revenues are used only to classify firms into "large" and "small"

foreign revenue groups.

A few firms mentioned in the questionnaire or in the annual reports

that they took particular hedging actions to neutralize the accounting

exchange adjustments. These data were used to identify firms that were

taking "action" against accounting exposure.

Control (Domestic) Firms. The pool of control firms consisted of

all the firms on the CRSP tape that met the data requirement, and

were not included in the pool of test firms. From this control pool,

281 firms were selected to match the industry classifications of the

test firms. An attempt was made to pairwise match the selected domestic

firms to the 281 test firms on the basis of the two-digit SEC industry

code. Due to the limited number of firms in the control pool, the matching

has not been satisfactory. A note of caution about the control firms is

appropriate. By the selection process, it was attempted to limit control

firms only to domestic firms. However, no specific checks were made

to ensure that the firms included in the domestic group did not have

significant foreign direct investment, as these checks would not be

practical. Additionally, the selection process resulted in selecting

control firms that were smaller in size that the test firms.

Descriptive Statistics. Some descriptive statistics of the test

and control firms are shown in Tables 5 through 9. Table 5 provides the

frequency distribution of the accoutning method used by the test firms

for the years 1973-1977. The change to the monetary/nonmonetary method

seems to be equally divided between 1975 (108 firms) and 1976 (104

firms).3









The frequency distribution of the magnitude of foreign revenues

is reported in Table 6. These numbers should be viewed with the fact

in mind that the foreign revenues for 15.5% of the firms were

extrapolated by the researcher. In Table 7, the frequency distribution

of exchange adjustments as a percentage of net income is provided for

the years 1973-1977. Note that the numbers given are percentages of

net income before extraordinary items, as reported on the COMPUSTAT tape.

Insofar as the exchange adjustment is already included in the income

number, the percentage exchange adjustment reported is biased upward

for losses and biased downward for gains. For cases where the net income

is negative, the biases are reversed. The number of firms reporting

large exchange adjustments increased every year, with a substantial jump

in the 1974 fiscal year. For example, the number of firms with exchange

adjustments of 3% or greater are 95, 115, 115, 136, and 111 for the

years 1973-1977 respectively. Some possible explanations are: (1) the

exchange rates could have become more volatile in that period, under the

floating exchange rate system; (2) increased disclosure of exchange

adjustments by firms; and (3) changes in accounting method by firms may

affect the magnitude of the exchange adjustment reported in 1975 and 1976.

The industry numbers of the test and control firms are reported

in Table 8A. The matching is obviously poor, but both groups are well

diversified over industries. The test firms do have a high concentration

in classification 28 (Chemicals & Allied Products) and in classification

35 (Engines, Machinery, Turbines) which may introduce some bias. Also,

control firms include some utilities (classification No. 48 and 49)

not present in test firms.









The distribution of the month of the fiscal year end is shown

in Table 8B. Seventy-four percent of the test firms and 69 percent of

the control firms have December end fiscal years. As mentioned earlier,

firms with fiscal periods ending January and February were removed from
5
the test group.

In Table 9, comparative statistics about the income and revenue

of the test and control firms are provided for each of the years

1973-1977. On the average, the test firms are four times as large as

the domestic firms.









TABLE 5

Frequency Distribution by Accounlting Method: Test (MNC) Firms



Accounting Method Frequency

1973 1974 1975 1976 1977


C 140 135 64 15 7

H 118 124 81 25 13

M 67 70 178 282 304

Other 8 8 6 1 0

not available 7 3 11 17 16


total 340 340 340 340 340




TABLE 6

Frequency Distribution by Size of Foreign Revenue: Test (MNC) Firms



Size (million $) Frequency

1973 1974 1975 1976 1977


500 and above 66 78 81 85 81

400-499 18 13 17 17 23

300-399 18 26 27 21 20

200-299 45 39 39 42 42

100-199 83 80 78 76 78

less than 100 110 104 98 99 96


total 340 340 340 340 340












TABLE 7

A. Frequency Distribution of Exchange Adjustment as a
Percentage of Net Income: Test Firms


Exchange 1973 1974 1975 1976 1977
Adjustment (%) Fa CFb F CF F CF F CF F CF

-21 and less 2 2 8 8 11 11 11 11 9 9

-11 to -20.9 8 10 17 25 14 25 19 30 18 27

-7 to -10.9 4 14 19 44 12 37 29 59 20 47

-3 to -6.9 18 32 36 80 38 75 43 102 43 90

-1 to -2.9 17 49 35 115 39 114 30 132 39 129

-.9 to +.9 183 232 159 274 142 256 145 277 176 305

1 to 2.9 45 277 31 305 23 279 24 301 14 319

3 to 6.9 36 313 20 325 31 310 16 317 10 329

7 to 10.9 14 327 2 327 14 324 5 322 5 334

11 to 20.9 7 334 7 334 7 331 9 331 3 337

21 and above 6 340 6 340 9 340 9 340 3 340


bCF = Cumulative Frequency


aF = Frequency











TABLI. 8

A. Firms by Industry Number


Test Firms

Fb CFc


Control Firms

IN F CF


2
9
12
13
37
38
41
44
46
49
61
62
112
119
124
136
148
161
203
220
240
253
258
260
262
263
264
275
276
277
278
281


12
13
23
24
26
47
53
67
75
76
78
88
96
108
114
120
128
148
167
177
194
205
222
234
239
244
247
248
250
263
265
266
267
275
276
277
278
279
280
281


aIN = 2-Digit Industry Number
F = Frequency

CCF = Cumulative Frequency




44





TABLE 8 (continued)

B. Firms by Fiscal Year Ending: Test and
Matched Domestic Firms



Fiscal Year Test Control

1 -- 6

2 1 4

3 5 11

4 1 5

5 1 5

6 20 13

7 6 6

8 5 10

9 12 17

10 16 4

11 6 6

12 208 194


TOTAL









TABLE 9


A. Statistics on Net Income of Firms
(all figures in millions of dollars)


TE S I

S.D.a

228

238

195

257

282


Min.b

-1

-157

-207

-78

-81


Max.C

2443

3142

2503

2903

3339


Mean

22

29

25

32

32


CON

S.D.

74

108

80

93

111


B. Statistics on Revenue of Firms
(all figures in millions of dollars)


TEST
S.D.

3171

3995

4226

4946

5684


Min.

29

35

32

40

47


Max.

35798

42062

44365

48631

54961


Mean

427

550

560

625

687


CON
S.D.

969

1482

1376

1565

1918


= Standard Deviation

= Minimum Value

= Maximum Value


Note: No. of firms
Sample =
No. of firms
Sample =


in Test
340
in Control
281


Year

1973

1974

1975

1976

1977


Mean

92

99

90

113

123


TR OL

Min.

-119

-38

-86

-10

-448


Max.

1071

1627

1152

1326

1535


Year

1973

1974

1975

1976

1977


Mean

1585

1990

2086

2334

2558


TROL

Min.

9

10

10

13

12


Max.

11176

19795

17516

20388

26052


aS.D.

Min.

CMax.


-----


--------













Grouping of Firms

Firms are classified in various ways to test for information

content of the accounting change. A particular grouping is chosen

to enhance the possibilities of rejecting the null hypothesis of no

information content, if in fact that hypothesis is false. Six

different grouping schemes were employed in the testing procedure,

and firms were primarily classified by the accounting method used for

translation. All firms were screened by a filter on the materiality

of the exchange adjustment to increase the power of the independent

variable. In each period, firms were retained in the test group only

if the magnitude of the exchange adjustment was equal to or greater than

some designated percentage of net income. The successive percentage

filter rules were 1%, 3% and 4%, which were arbitrarily chosen as a

balance between obtaining an adequate sample size while excluding firms

with insignificant exchange adjustments.

The first grouping scheme is based on the three accounting methods

for translation used by test firms. These methods are:

(a) current/noncurrent method (C)

(b) all accounts at current rate except fixed assets. This

method is essentially the same as the monetary/nonmonetary

method with inventory at current rates and is called the

hybrid method (H)

(c) monetary/nonmonetary method (M)










For each of the years 1973-1977, the 281 test firms that passed the

above-noted filter (and the corresponding control firms) were put in one

of the portfolios Pc, Ph or Pm, if the accounting method used was C, H

or M respectively. Note that a specific firm would be in different

portfolios in different years if the firm changed its accounting method

for translation.

In the second grouping scheme, six portfolios were formed on the

basis of a cross-tabulation of the three accounting methods and the

two levels (increase/decrease) of change in exchange adjustment (CEA)

each year.

The third and fourth grouping schemes were similar to the second.

In the third scheme, the variables are accounting method and magnitude

of foreign revenues (large/small). In scheme four, the variables are

accounting method and sign of exchange adjustment (SEA) which has two

levels (positive/negative).

In scheme five, three portfolios were formed. The first portfolio

consisted of firms that used the M accounting method in 1973-1974 and

essentially continued doing so after FASB No. 8. The second and third

portfolio consisted of firms that were using the H and C accounting

methods respectively in 1973-1974 and had to switch to M after FASB

No. 8. The criteria for including firms in this scheme were:

(a) Each firm must have used the same translation method in

both 1973 and 1974 (i.e., M or H or C).

(b) Each firm must have used method M in 1976 (i.e. all firms

included switched to method M in 1975 or 1976).










(c) As the aim was to keep the same firms in each portfolio in

all the years, the 1% filter was not applied to firms in

each year. But of the sample of 281 firms, 79 firms were

excluded from all four years as they had "zero" exchange

adjustments in three of the four years.

In contrast to grouping schemes 1 through 4, a particular firm stays

in the same portfolio for all four years in scheme 5. Grouping scheme 5

was used solely for time series tests. In that design, the test firms

served as their own control, as is explained later.

The sixth grouping scheme was based on test firms in which the

management were judged (based on the best available evidence) to have

hedged against the accounting risk of translation. The data were

obtained from the firms responding to the questionnaire and from

disclosures in annual reports, though incomplete. These firms were

denoted as "Action" firms, while "No Action" firms were the remainder

of the test firms. Note that the No Action pool of firms could include

firms in which managements hedged against accounting risk but we are

unaware of it. The Action firms selected in this manner were matched

to the No Action firms by accounting method in each year. The Action

firms were not screened on the basis of any filter on exchange adjustments.

As before, a matched group of domestic firms was also selected.

The grouping schemes 1-6 are summarized in Table 10.


TABLE 10

Summary of GRouping Schemes


Grouping Scheme 1: Accounting Method (for Years 1973-1977)
Portfolio Accounting Method
Pc C
Ph H
Pm M




49





TABLE 10
(conti llned)


Grouping Scheme 2: Accounting Method and CEA (for Years 1974-1977)


Portfolio
Pci
Pcd
Phi
Phd
Pmi
Pmd


Accounting Method
C
C
H
H
M
M


CEA
increase
decrease
increase
decrease
increase
decrease


Grouping Scheme 3: Accounting Method and Magnitude of Foreign Revenues
(for Years 1973-1977)


Accounting Method
C
C
H
H
M
M


Foreign Revenues
large
small
large
small
large
small


Grouping Scheme 4:

Portfolio
Pc+
Pc-
Ph+
Ph-
Pm+-
Pm-


Grouping Scheme 5:


Accounting Method and SEA (for years 1973-1977)


Accounting Method
C
C
H
H
M
M


Firms Classified by Accounting Method Used in Pre-


FASB No. 8 Period


Accounting Method
Used in 1973 & 1974


Accounting Method
Used in 1976
M
M
M


Grouping Scheme 6: Action/No Action Firms by Accounting Method


Accounting Method
Used in 1973 & 1974
C


Portfolio


Action Firms


No Action Firms
NAc
NAh
NAm


Portfolio
Pcl
Pcs
Phl
Phs
Pml
Pms


SEA
positive
negative
positive
negative
positive
negative


Portfolio


Pcm
Phm
Pmm











Time Periods for Tests

In studies devoted to testing the information content of accounting
7 8
changes or accounting signals, the time period chosen for testing of

any effects is of some importance. Usually, the date of the release of

annual financial information is taken as the critical date. If the

accounting data have any information content, the effect is expected to

be most pronounced at the time annual reports with the changed method

are first released. For operational convenience, the date of release

of annual reports is taken as a surrogate for the date the financial

information was first released to the market. On the average, practically

all annual reports are released within three months of the fiscal year

end.

In this research, for the tests described earlier, twelve month

holding periods are used for tests on total returns and cumulative

average residuals. In an effort to increase sample size, firms with

different fiscal year ends were included, as long as their annual

reports were released some time during the holding period. As most

firms in the sample have December fiscal year ends, the holding period

used is April through March of the following year. Firms with fiscal

year ends of January and February are excluded, as their annual reports
9
may have been released before April. An example will clarify the

process. For 1973 reports, the sample includes firms with fiscal year

ending March 1973 through December 1973. The corresponding holding

period to test the effect of 1973 reports on security returns is

April 1973 through March 1974. The assumption inherent in the above









process is that by the end of March 1973, the capital market had reached

equilibrium with respect to 1972 disclosures for all firms in our

sample. So, any effect of 1973 disclosures should be captured in the

returns of the 12-month period, April 1973 through March 1974.

Now, the appropriate period for measuring the market effects of

FASB No. 8 is considered. The various critical events and corresponding

dates relevant to this research are given in Table 11A.



TABLE 11
Critical Events for FASB No. 8 and Test Periods

A. Relevant Critical Events

Dec. '73 Dec '74 Oct. '75 Jan '76 March '77 Dec. '77

FASB No. 1 FASB No. 8 FASB No. 8 FASB No. 8 first final data
adopted exposure adopted effective required on CRSP
and draft disclosure tape
effective available
in annual
report

B. Various Test Periods

Test
Period April '73 April '74 April '75 April '76 April '77 Dec. '77

1 <

2

3

4

5




A consideration of the critical events does not lead to an unambiguous

determination of the period when effects of the switch to FASB No. 8 should

be most pronounced. The date of the exposure draft, the date of adoption









of FASB No. 8, the effective date or the date when a firm actually

changed to the new method are all candidates for expecting an effect

of the new standard. A complicating factor is that in FASB No. 8 we

have a non-discretionary accounting change, and all MNC were required

to switch to the temporal method (equivalent to the monetary/nonmonetary

method) by the fiscal year beginning January 1976.10 However, early

adoption was encouraged, and many firms switched in 1975 and early 1976.

For the firms that adopted FASB No. 8 early, there could be some

signaling effects11 and investors may treat the early adopters different-

ially from the firms that adopted FASB No. 8 only at the effective date.

In this study, comparisons will be made in various test periods,

to determine in which period, if any, FASB No. 8 had an effect. The

five test periods initially selected are shown in Table 11B.

Some of the considerations and expectations for the test periods

are discussed next. As mentioned in Chapter II, FASB No. 1 required

firms to make significant disclosures aobut foreign currency trans-

lations. For many firms, the market first received public information

about currency translations after FASB No. 1 was adopted. The years

1973 and 1974 are considered the post-FASB No. 1 pre-FASB No. 8 period.12

In the 1975-1977 period, presumably no new information was provided, but

a nondiscretionary accounting change took place. The years 1975-1977

are considered the post-FASB No. 8 period.


Construction of Relative Risk Equalized Returns and Average Residuals.

Relative Risk Equalized Returns. From 60 to 48 months of monthly

common stock return data (with dividends) from the CRSP tape were used









to estimate the relative risk of each firm by ordinary least squares

regression. The value weighted market index was used. The market

model was re-estimated for each of the periods 4/68-3/73, 4/69-3/74,

4/70-3/75, 4/71-3/76, 4/72-3/77, to minimize any bias due to the

instability of betas. For each of these five periods, firms in each

portfolio were weighted to form a portfolio of beta (relative risk)

equal to one. The procedure is as follows: firms are ranked in each
A
portfolio in descending order of B, and divided into two groups. Let

the top half of the firms have an average beta of 1 and the bottom half

of the firms have an average beta of $2. The two groups are then

linearly weighted to give a combined portfolio beta of one. The weights

for each group, x and (l-x) are calculated from the equation:


xl1 + (I-x) 2 = 1.

If both i1 and A2 have values that are larger or smaller than 1, then x

is no longer in the interval (0,1), implying short selling. In this

procedure, groups in which the weights were larger than 1.75 were

thrown out to prevent spurious weighted returns.13 The monthly returns

of each firm in the top and bottom half of a portfolio are then weighted

by x and (l-x) respectively, to form the time-series of the weighted

returns for each portfolio for the 12-month test period. The five

corresponding test periods are 4/73-3/74, 4/74-3/75, 4/75-3/76, 4/76-3/77,

4/77-12/77. The weighted returns of the test and control firms for

each portfolio are then difference to provide the sample for the statistical

tests. The above procedure was followed for the test and domestic firms

in the portfolios for grouping schemes 1, 2, 3, 4 and 6.








Cumulative Average Residuals. The parameters of the market model

were estimated for all test firms for each of the same five estimation

periods as in the case of weighted returns above. Then, the estimated

parameters were used to calculate the residuals for each of the cor-

responding 12-month test periods, as above. Finally, the residuals

were accumulated across firms in each portfolio to provide the average

residuals. The average residuals were calculated for test portfolios

for grouping schemes 1 through 5.


Measuring Effects on Financial Structure

Changes in the financial structure of firms were tested by examining

the structure of the balance sheet and income statement before and after

the change in accounting method. The test firms were grouped according

to grouping scheme 5. Annual data for the years 1973-1977 were-obtained

from the COMPUSTAT tape for the following accounts: (1) long-term assets;

(2) inventory; (3) current liabilities: (4) long-term debt; (5) net

income before extraordinary items; and (6) total revenue. To permit

comparisons across firms and time periods, ratios were formed by de-

flating the above six accounts by total assets. We are unaware of any theory

that would provide guidance on choosing the structural variables that may

be affected by FASB No. 8. Intuitively, and from comments in the finan-

cial press, one can infer that managements reacting to FASB No. 8 may

change the direct investment and borrowing patterns in foreign countries.

Various techniques of hedging may affect the total income of a firm and

the foreign revenues generated by a firm. Changes in translation methods

may also affect the valuation of foreign inventories and thus the con-

solidated inventory account. This reasoning was the basis for choosing

these six variables.








Changes in the financial structure were determined by cross-section-

ally comparing the means of the avcrige ratios of the three portfolios,

Pcm, Phm, Pmm in the before-FASB No. 8 period (1973-74) and the after-

FASB No. 8 period (1976-77). For each period, the four balance sheet

ratios were simultaneously compared for the three portfolios. Similarly,

the two income statement ratios were also simultaneously compared for the

three portfolios in the before and after periods. The comparisons made

are illustrated in Table 12A. The subscripts B and A refer to the before

and after periods, respectively.



TABLE 12
Comparisons Made to Test Changes in Financial Structure

A. Cross-Sectional Tests (Balance Sheet and Income Statement)

"Before" period (1973-1974) "After" Period (1976-1977)

Pcm/B vs. Phm/B Pcm/A vs. Phm/A
Pcm/B vs. Pmm/B Pcm/A vs. Pmm/A

B. Time Series Tests (Balance Sheet and Income Statement)

Pcm/B vs. Pcm/A
Phm/B vs. Phm/A
Pmm/B vs. Pmm/A


Note: letters c, h, m refer to the three accounting methods.



Each of the three portfolios were also compared before and after the

accounting change, as shown in Table 12B.

If reports in the financial press about management reaction to

FASB No. 8 are correct, and these actions lead to structural changes,

then we should find a change in the structure of the Pcm and Phm firms.










Statistical Tests

Hotelling's T2 statistic is used to test for differences in the

means of two vectors. Differences in the covariance matrices of two

2
distributions are determined by Box's 2 test for homogeneity of co-

variance matrices. The two tests are briefly described below.


Hotelling's T2 Test

Let p be the number of portfolios.

Let T be a (pxl) column vector of average monthly total returns

for the test firms (or average monthly residuals for test firms in the

'before' period).

Let C be a (pxl) column vector of average monthly total returns

for the control firms (or average monthly residuals for control firms in

the 'after' period).

The difference vector D = T C is first formed. The T2 statistic

is given by:1

T2 = N(D-po)' Sd-1 (D-po)

where N = number of monthly observations

Vo = vector of zeros

Sd = variance-covariance matrix for the differences.

If the null hypothesis holds, the quantity F = T2 .(N-p) has

p(N-l)
the F distribution with (p, N-p) degrees of freedom. The decision rule

is to reject the null hyopthesis of no difference between vector means

if F is greater than Fa,p,N-p).










Box's 2 Test5

The statistic used in Box's test is

-2a Ln (X)
2
which is approximately distributed as X with p (p+l) (k-l)/2 degrees

of freedom, where

k = number of groups or covariance matrices being compared.

p = number of portfolios.
k 2
1 1 2 + 3 1
A = ( -) ( P P
i=l N N 6(p+l) (k-l)

th
N. = number of observations in the i group

N = EN. = total number of observations
1

NPN/2 =k (det Si)Ni/



i= i


det S. = determinant of (within) ith group covariance matrix (of

Rit or Uit), i=l,...,k

det S = determinant of the pooled covariance matrix.

Large values of -2Ln(X) lead to rejecting the null hypothesis of

no difference in the covariance matrices.


Interpretation of Expected Results

In this study, the tests for effects on the security return behavior

of MNC--comparison of the return distributions of "test" and "control"

firms--are conducted over two periods. The years 1973 and 1974, which

precede the issuance of FASB No. 8, are considered the pre-FASB No. 8

period. The years 1975 through 1977 are considered the post-FASB No. 8

period.











The pre-FASB No. 8 period is utilized as a control or pre-test

period in this study. During this period, FASB No. 1 was in effect,

and all MNC were assumed to be making the required disclosures about

the translation methods used and the amount and disposition of exchange

adjustments. Hence, the security return distributions of all MNC--

irrespective of the translation method used--are not expected to be

significantly different from that of control firms of similar relative

risk.

During the post-FASB No. 8 period, all MNC are required to follow

a uniform method of accounting for foreign currency translations.

Following previous research on the effects of accounting changes, reviewed

by Kaplan (1978), no effect on the security return behavior of MNC due to

the switch to the method recommended by FASB No. 8 should be expected.

However, as discussed in Chapter II, if managements expend real resources

in neutralizing the accounting exposure due to FASB No. 8, then the

return distributions of MNC are expected to be significantly different

from that of control firms of similar relative risk. In this discussion,

the implicit assumption is that managements had no motivation to expend

real resources in hedging accounting exposure before the implementation

of FASB No. 8, as they were free to choose the translation method that

best suited them.

In the actual test results, the security return distributions of

MNC could be either significantly different or not different from the

return distributions of control firms, in both the pre- and post-FASB

No. 8 periods. The four possible outcomes are characterized in Table 13.








TABLE 13

Possible Outcomes of the Comparisoni~ of the Return Distributions of Test
and Control Firms

Post-FASB No. 8 period outcome


not
different
Pre-FASB No. 8
Period Outcome
different


not different different


1 2


3 4


Note: "Different" ("not different") refers to the distributions of test
firms being significantly different (not different) from control firms.

The various outcomes are interpreted as follows:

Cell 1: The different accounting methods have no effect on the

security returns of MNC in both the pre- and post-FASB

No. 8 periods, implying that hypothesis 1 through 4

cannot be rejected.

Cell 2: No effect in the pre-FASB No. 8 period provides evidence

that hypothesis 1 cannot be rejected. Then, a significant

difference in the post-FASB No. 8 period implies that

hypothesis 3 and/or hypothesis 4 should be rejected.

Cell 3: There is no good explanation for this result. One

possibility is that an uncontrolled systematic bias

exists between test and control firms. (This can occur

as it was not possible to randomly assign firms to

treatments.)

Cell 4: This outcome is difficult to interpret. One explanation

is that in the pre-test period there is some systematic

factor affecting economic values of securities in the










sample, that is correlated with the accounting method for

translation. Then, interpretation of the post-FASB No. 8

result becomes ambiguous.



Notes

1. This asset pricing model is based on several widely cited assump-
tions. See, for instance, Fama (1970). Black (1972) has formu-
lated a two-factor version of the asset pricing model which is:
E(Rit) = E(Rzt) + (E(Rt)-E(Rzt)) Bit
where~Rzt = Rate of return on a minimum variance portfolio having
Cov (Rzt, Rmt) = 0. Other terms are as defined previously.

2. The assumptions of the market model have been widely discussed. See
for instance Fama (1976). Some of the assumptions are:
(1) E (uit) = 0
(2) Var (uit) = constant
(3) Cov (uit, uis) = 0, t # s
(4) Cov (Rmt, uit) = 0
(5) Cov (uit, ujt) = 0
(6) ai, Bi are constant over time.

3. In the 1977 fiscal year, all firms should have switched to the
monetary/nonmonetary method under FASB No. 8. So, a check was
made to see if any errors had been made. It was found that some
firms had given wrong data in their questionnaire responses. A
check of the annual reports of these firms revealed that four
firms were continuing to use the hybrid method but claimed that
inventories were small. The other firms had switched to the
FASB No. 8 method in 1976.

4. The income used is data item 18 on the tape.

5. One February firm inadvertently remained in the test group, and the
error was discovered after most of the tests were completed. But
one firm out of 281 firms should not cause any problems.

6. See Gonedes (1975, p. 231) for a discussion of this point.

7. See Ro (1978) for an example in leases.


8. See, for example, Gonedes (1978).









9. In many papers, e.g. Gonedes (1975), Ro (1978), March fiscal year
firms are also excluded. But -, the reports of these firms can be
released in April at the earliest, they are kept in our sample.
Anyway, our sample has only a few March fiscal year firms.

10. December 1975, for 53-week fiscal year firms.

11. See Gonedes (1978) for a discussion of signalling in an accounting
context.

12. The period 1970-1972 was initially considered for inclusion in the
testing but was dropped due to lack of appropriate data. The major
problem was that disclosures about currency translations was uneven
and provided on a discretionary basis in that period.

13. Large-scale factors cause error. E.g., for weights 15, -.5 and returns
0.2,.4, weighted returns are .5 x 2 .5 x .4 = -0.1. But equal
weights of 5, -5 give returns of 5 x .2 5 x .4 = -1.0.

14. See Morrison (1967) for details.


15. See Box (1949), Giri (1977) for details.












CHAPTER IV
TEST RESULTS AND ANALYSIS



Variability of Income

In this section, the variability of net income reported by firms

over time is investigated. No attempt is made to ascribe any causal

linkage to FASB No. 8. An attempt is made to simply determine the

variability in income during the post-FASB No. 8 period.

Ideally, the variance of income for each firm before and after

FASB No. 8 should be used to determine if adoption of FASB No. 8 is

associated with an increase in the variability of income. But only a

few data points before and after the accounting change are available

for each firm, so the variance cannot be calculated. The mean absolute

percentage change in income is used as a surrogate for the variability

in income each year. The values of the absolute percentage change in

income for test and control firms are given in Table 14A. From the

values of the coefficient of variation, it seems that the volatility of

income for test firms increased in 1975, stayed volatile in 1976

and then came down to levels comparable to the volatility of income

of the control firms in 1977. The average values of the absolute

exchange adjustment as a percentage of net income are shown in

Table 14B. The mean exchange adjustment seems to have increased in 1974

and then again in 1976, before decreasing slightly in 1977. However,

the results are difficult to interpret due to the presence of some










TABLE 14

A. Mean Absolute Percentage Change


in Incomea b


TEST
c d e
Year N Mean S.D.C Max C.V.

1977 331 39.3 72.7 900.0 1.84

1976 337 63.2 147.4 2100.0 2.33

1975 337 45.2 104.8 1266.6 2.31

1974 338 50.8 105.0 1408.3 2.06




CONTROL

Year N Mean S.D. Max C.V.

1977 260 63.8 108.6 966.6 1.69

1976 260 81.0 132.8 1100.0 1.62

1975 261 62.2 92.6 650.0 1.48

1974 261 68.18 155.1 2100.0 2.27


percentage Income Change is


bMinimum Value = 0


given by:ABS((NIi-NIi_l)/(NIi_-))
i= 1974, 75,.,77


where i = year
in each case


cS.D. = Standard Deviation

dMax. = Maximum Value


eC.V. = Coefficient of Variation = S.D.
'lean





64




TABLE 14 (Continued)

B. Mean Absolute Exchange Adjustment as a Percentage of Net Income
of Test Firms

Standard
c a b
Year N Mean (%) Deviation Min. Max. (%)

1973 285 3.90 8.85 0 84.7

1974 285 5.82 16.9 0 244.0

1975 285 5.95 9.96 0 64.7

1976 285 9.61 42.3 0 555.0

1977 285 7.78 53.8 0 898.0

aMin. = Minimum Value bimum Value mum Value N = No. of Firms






C. Mean Absolute Exchange Adjustment as a Percentage of Net Income
With Outliers Removed


Year N Mean without Outliers* No. of Outliers

1973 278 2.71 7

1974 276 3.71 9

1975 272 4.16 13

1976 269 4.42 16

1977 278 3.71 7


Outliers defined as firms having exchange adjustment > 25%










outliers, so the mean exchange adjustment was recalculated without the

outliers, and the results are shown in Table 14C. With the outliers

removed, the exchange adjustment can be seen to have increased in 1974

and then again in 1975 before decreasing slightly in 1977. As FASB No. 8

was first required in 1976, the increases in exchange adjustments are

probably due to currency fluctuations and not due to FASB No. 8.



Effect on Security Price Returns

The results of the various cross-sectional and time series tests

on security returns are presented in this section. When a number of

tests are conducted to evaluate a particular research problem, the

results of all the tests should be considered together to draw a con-

clusion. This analysis of all the test results, by each test period,

is done after all the individual results are presented.

Validation of Control. Before the test and control firms are

compared under various treatment conditions one should be confident that

the control is a valid one. The validity of the control was assessed

by comparing the mean returns of the total group of control firms with

the mean turns of the total group of test firms, after adjusting for

relative risk differences. If the control is appropriate there should be

no differences in the two mean returns.

The results of the comparisons of the relative risk equalized

returns of all 281 test and control firms for each of the five test

periods is shown in Table 15. In each of the test periods, MNC had lower

returns than the domestic firms. As expected, there was no significant











TABLE 15

Total Test Firms Vs Total


Control Firms


of Comparison of Risk


(Test-Control)
S Means of Return
N Differences X100


1 (1973 &
1974)

2 (1974)

3 (1975)

4 (1976)

5 (1976 &
1977)


-.1639


-.3322

-.1301

-.749

-.8511


Equalized Returns

Std. Deviation
of Return
Differences X10


0.281


0.355

0.168

0.147

0.133


aNumber of firms

b
p = 1% (20 DF)


in each portfolio = 281


p = 10% (11 DF)


dN = No. of months


B. Average of Unweighted Betas of Test and Control Group

AVERAGE BETA


Test Group


Control Group


4/68-3/73 1.147 1.136

4/69-3/74 1.247 1.26

4/70-3/75 1.17 1.18

4/71-3/76 1.16 1.15

4/72-3/77 1.16 1.09


A. Results


Test
Period


t value


-0.28


-0.32

-0.26

-1.76c

-2.92b


Period










difference between the returns of tbl two portfolios, using a two-tailed

t test, in the first four periods. However, in the 1976-1977 period,

the mean returns were significantly different at the 1% level, implying

that some change took place in the period 4/77-12/77. The lower part of

Table 15 shows the unweighted average betas for the two portfolios for

the five periods. The average betas are similar for the two portfolios

and appear to be stable over time. Of course, the betas may have changed

in the latter part of 1977 for some reason. Dukes (1978) also found

some unusual activity in 1977, in his tests on the effect of the magnitude

of foreign operations (p. 63 and p. 96). It is unlikely that FASB No. 8

was responsible for the unexpected results in test period 5. Since only

nine months of data were used for 1977, there could be some quarterly

reaction that possibly averages out over the fiscal year. Further

investigation of this surprising result is left for future research.

However, the above finding does have important implications for inter-

pretation of other tests on security prices in the 4/76-12/77 period.

The significant differences found in the 4/76-12/77 period for other

classification schemes cannot now be taken to imply effects of FASB No. 8.

As such, the many test results that were significant in period 5 are

probably contaminated, and so are ignored. Only the test results for

test periods 1-4 are presented here.

Test Results. The detailed test results of the cross-sectional

and time-series tests on security returns conducted for grouping

schemes 1-6 are presented in Tables 16-23. The results of all the

tests conducted on security price returns are summarized in Table 24.









TABLE 16

Grouping Scheme 1: liy Accounting Method
Filter = 1%


A. Number of


Firms in Each Portfolio in Each Year


Portfolio '73 '74 '75 '76 Firms

Pc 62 71 29 9

Ph 44 57 45 12 Test

Pm 35 43 109 163



Pc 62 70 29 8

Ph 47 57 46 11 Control

Pm 35 43 108 161


aDefined in Chap. III.

Note: FILTER = ABS (Exch. Adjustment
Note: FILTER Net Income
Net Income








TABLE 16 (continued)

B. Results of Tests on Weighted Return Differences


Test Period: 1 (1973 & 1974) N = 24

Pc Ph Pm
Mean of Return Differences (Test-Control) X 100 -.64 -.09 -.49
Standard Deviation of Return Differences X 10 .24 .38 .44
t Values -1.31 -.12 -.54

2
T = 2.4, F = .737
X2 (6 df) = 11.9 (p = 6.2%)**
Test Period: 2 (1974) N = 12

Pc Ph Pm
Mean of Return Differences (Test-Control) X 100 -.85 -.26 -.44
Standard Deviation of Return Differences X 10 .25 .44 .57
t Values -.72 -.20 -.26
ST2 = 2.4, F = .65

X2 (6 df) = 3.94 (p = 68%)
Test Period: 3 (1975) N = 12

Pc Ph Pm
Mean of Return Differences (Test-Control) X 100 1.18 -.40 0
Standard Deviation of Return Differences X 10 .19 .27 .19
t Values 2.15** -.51 0
T = 6.06, F = 1.65
X2 (6 df) = 10.5 (p = 10.2%)
Test Period: 4 (1976) N = 12

Pc Ph Pm
Mean of Return Differences (Test-Control) X 100 -2.92 -.37 -.87
Standard Deviation of Return Differences X 10 .36 .40 .16
t Values -2.74* -.31 -1.88**
2
T = 8.6, F = 2.3
X2 (6 df) = 11.44 (p = 7.5%)**

Two-Tailed Test .05(*) .10(**)
t(lldf) = 2.2 1.79

t(23df) = 2.06 1.71


.05(*) .10(**)
F(3,21) = 3.07 2.36
F(3,9) = 3.86 2.81

Note: Chi-square statistic applies to Box test.










TABLE 16 continuede)

C. Results of Tests on Monthly Residuals


Test Period: 1973 & 1974 vs 1975 & 1976 N = 24
Pc Ph Pm
Means of Residual Differences (Before-After) X 100 .24 .25 .39

Standard Deviations of Residual Differences X 10 .43 .33 .36

t Values .27 .37 .52
2
T = 0.34, F = .10
Test Period: 1974 vs 1975 N = 12
Pc Ph Pm
Means of Residual Differences (Before-After) X 100 -1.15 .28 .28

Standard Deviations of Residual Differences X 10 .49 .27 .40

t Values -.81 .35 .24
2
T = 13.38, F = 3.65**
Test Period: 1974 vs 1976 N = 12
Pc Ph Pm
Means of Residual Differences (Before-After) X 100 .59 -.31 .82

Standard Deviations of Residual Differences X 10 .51 .32 .39

t Values .39 -.34 .72
2
T2 = 1.663, F = 0.45
Test Period : 1975 vs 1976 N = 12
Pc Ph Pm
Means of Residual Differences (Before-After) X 100 1.75 -.60 .53

Standard Deviations of Residual Differences X 10 .36 .27 .13

t Values 1.65 -.75 1.4
T2 = 17.1, F = 4.66*
T = 17.1, F = 4.66*


05(*)

F(3,9) = 3.86
F(3,21) = 3.07


.10(**)

2.81
2.36





A.

Tests on Grouping Scheme 1: By Accounting Method. The number of

firms for each portfolio for each ye;ir is shown in Table 16A. The

sample size for the test and control firms is slightly different for

some years due to a few return observations missing on the CRSP tape

for the respective periods. Results of the cross-sectional tests on

risk equalized monthly returns are provided in Table 16B. The T

statistic, the corresponding F value, and the X statistic is shown

for the four periods over which comparisons are made. For each period,

N is the number of monthly observations used. The critical values of F

are also provided for the 5% and 10% confidence levels. The F value is

not significant for the first four periods; however, the t test shows

that the firms using the C method had mean returns larger than the control

in 1975 and mean returns significantly smaller than the control in 1976.

Firms using the M method also had mean returns smaller than the control

in 1976, imploying that individual portfolios may have been affected by
2
the implementation of FASB No. 8. The X statistic is significant at the

10% level in periods one and-four, weakly imploying differences between

the return covariance matrices of the test and control firms in those

periods.

Results of time-series tests on average residuals are presented in

Table 16C. For the test firms, mean residuals for a given test period

(before FASB No. 8 period) are compared with the mean residuals for

another test period of interest (after FASB No. 8 period). The T2

statistic and F value is shown for the four comparisons. The com-

parisons between 1974 and 1975 are significant at the 10% confidence

level, while the comparisons between 1975 and 1976 are significant at

the 5% level. Interestingly, the comparison of the average residuals

of 1974 versus the residuals of 1976 shows no significant differences

for the three portfolios. Apparently, in 1975 there was some change in











TABLE 17

Grouping Scheme 1: By Accounting Method
Filter = 4%

A. Number of Firms in Each Portfolio in Each Year


Portfolio '73 '74 '75 '76 Firms

Pc 29 33 15 6

Ph 15 28 30 8 Test

Pm 23 23 65 108



Pc 29 33 15 5

Ph 17 29 31 8 Control

Pm 24 24 65 108


aDefined in Chap. III.

Note: FILTER = ABS (Exch. Adjustment
Note: FILTER Net Income
Net Income









TABLE 17 (continued)


B. Results of Tests on Weighted Return


Differences


Pc Ph Pm
Test Period: 1 (1973 & 1974) N = 24
Means of Return Differences (Test-Control) X 100 -1.17 0.26 -.25
Standard Deviation of Return Differences X 10 .25 .42 .60
t Values -2.22* 0.3 -.20

T. = 7.02, F 2.14
2
X (6 df) = 9.72 (p = 13.6%)
Test Period: 2 (1974) N = 12

Means of Return Differences (Test-Control) X 100 -1.49 -.41 -.22
Standard Deviation of Return Differences X 10 .25 .44 .82
t Values -2.0** -.32 -.09

T2 = 4.56, F = 1.24
2
X (6 df) = 4.45 (p 61%)
Test Period: 3 (1975) N = 12

Means of Return Differences (Test-Control) X 100 1.85 -.14 -.02
Standard Deviation of Return Differences X 10 .40 .31 .24
t Values 1.57 -.15 -.03
2
T = 2.49, F = 0.68

X2 (6 df) = 11.51 (p = 7.37%)**
Test Period: 4 (1976) N = 12

Means of Return Differences (Test-Control) X 100 -3.54 -.64 -1.21
Standard Deviation of Return Differences X 10 .53 .52 .20
t Values -2.27* -.42 -2.03**


2
S= (6 df) = 4.78 (p 26**57.1%)
X2 (6 df) = 4.78 (p = 57.1%)


.05(*)


F(3,21) =

F(3,9) =


3.07

3.86


.10(**)

2.36


2.81


Note: Chi-square statistic applies to Box test.


T











TABLE 17 (continued)

C. Results of Tests on Monthly Residuals


Pc Ph Pm
Test Period: 1973 & 1974 vs. 1975 & 1976 N = 24

Means of Residual Differences (Before-After) X 100 -.14 .12 .42

Standard Deviation of Residual Differences X 10 .59 .50 .42

t Values -.11 .12 .49
2
T = .897, F = .273
Test Period: 1974 vs. 1975 N = 12

Means of Residual Differences (Before-After) X 100 -1.32 -.08 .14

Standard Deviation of Residual Differences X 10 .56 .47 .51

t Values -.81 -.06 .09
2
T = 6.42, F = 1.75
Test Period: 1974 vs. 1976 N = 12

Means of Residual Differences (Before-After) X 100 .95 -.59 .84

Standard Deviation of Residual Differences X 10 .72 .57 .41

t Values .45 -.35 .70
2
T2 = 2.89, F = .789
Test Period: 1975 vs. 1976 N = 12

Means of Residual Differences (Before-After) X 100 2.27 -.51 .70

Standard Deviation of Residual Differences X 10 .41 .31 .18

t Values 1.92* -.58 1.29
T2 = 14.69, F = 4.0*
T = 14.69, F = 4.0*


F(3,21) =

F(3,9) =


.05(*)
3.07

3.86


.10(**)
2.36

2.81









the security return behavior of MNC which reversed itself in 1976,

leading to similar returns in 1974 and 1976.

As each portfolio (Pc, Ph and Pm) consists of different firms

over time, results of Table 16C provide evidence on portfolios of firms

using the same accounting method over time. The effect of a change in

accounting method over time on a specific set of firms is tested on

classification scheme 5, in Table 22.

To determine if the filter on the materiality of exchange adjust-

ments is adequate, all the tests on grouping scheme 1 are repeated with

a filter of 4%. Now, the test firms only include MNC that have exchange

adjustments that are 4% of income or greater. The number of firms in

each portfolio, for each year, are presented in Table 17A. Results

of the cross-sectional tests on returns and time-series tests on

residuals are presented in Table 17B and 17C respectively. From the tests

on returns, the F statistic is found to be weakly significant (10% level)

in period 4 only. As in Table 16B, the mean returns for the individual


portfolios are predominantly smaller for the test

control firms in Table 17B. The t test indicates

C method have smaller returns than the control in

As far as the covariance matrix is concerned, the

significant (7.3%) in period three only.

From Table 17C, it appears that the tests on

significant than in Table 16C. The residuals are

only in the comparisons of 1975 versus 1976, when

on exchange adjustment. Comments on the adequacy


firms vis-a-vis the

that firms using the

1973, 1974 and 1976.
2
X statistic is weakly



residuals are less

significantly different

a 4% filter is used

and effects of the


filter on exchange adjustment are deferred until more evidence is

presented with other filter magnitudes.















A.


Portfolioa


TABLE 18

Grouping Scheme 2: Accounting Method and CEAb
Filter = 1%

Number of Firms in Each Portfolio in Each Year


Firms


1*

8

3

9

60

103


Pci

Pcd

Phi

Phd

Pmi

Pmd


1*

7

4

7

61

100


Returns not usuable for comparisons.

aDefined in Chap. III.


CEA = Change in Exchange Adjustment.


Note: FILTER = ABS (Exch. Adjustment
Net Income


Test


Control











B. Results of


TABLE 18

Tests on


(continued)

Weihted Return Differences


Test Period: 2 (1974)


T2 = 14.4, F
T =14.4, F


1.31;

Pci


N = 12

X2(21 df) =

Pcd Phi


Means of Return
Differences
(Test-Control) X 100

Standard Deviation
of Return Differences
X 10
t Values


.02 -1.54


.33

-1.64


.53 -.19 -.99 -.35


.45

-.15


.81

-.42


.66

-.19


Test Period: 3 (1975)


T2 = 13.05,
T = 13.05,


F = 1.186;


N = 12

2 df)
x (21 df) =


17.67, (p = 66%)


Pci Pcd Phi Phd Pmi Pmd


Means of Return
Differences
(Test-Control) X 100

Standard Deviation
of Return Differences
X 10
t Values


2.32


.41

1.93**


.49 -.40 -.04


.62

-.23


.31

-.04


.50 -.54


.16

1.03


.25

-.73


Test Period: 4 (1976)


T2 = 20.56,
T = 20.56,


F = 2.61;

Pci


N = 12

X2(15 df) = 23.
Pcd Phi


Means of Return
Differences
(Test-Control) X 100

Standard Deviation
of Return Differences
X 10
t Values


-2.66 -1.20


--- .40

--- -2.28*


.74

-.56


1.11 -0.12 -1.37


.39 .18 .19

.98 -.23 -2.4*


.05(*)

F(6,6) = 4.28

F(5,7) = 3.97

Note: Chi-square statistic applies to Box test.


27.77, (p

Phd


= 14.65%)

Pmi


7, (p =
Phd


7.03%)**
Pmi


.10(**)

3.05

2.88










TABLE 18 (continued)


C. Results of


Test Period: 1974 vs. 1975


Tests on Monthly Residuals


N = 12


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


Phd


-.59 -1.41


.21 -1.04


t Value


-.44 -.84


.28 .21 -.78


T = 53.63, F = 4.87*


Test Period: 1974 vs. 1976


N = 12


Pci Pcd Phi Phd Pmi Pmd


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


t Value


.41 -.99 -.03 -1.06 1.58


.24 -.87 -.03


-.84 1.19


2
T = 6.45, F = 0.8212


Test Period: 1975 vs. 1976


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


t Value


Phd


- 1.82 -1.33


Pmd


-.24 -.02


.35 .60 .26


--- 1.76 -.76 -.31 -.04 2.45*


T2 = 48.73, F = 6.2*


F(6,6)
F(5,7)


.05(*)
4.28
3.97


.10(**)
3.05
2.88


Pmd


N = 12


_ ___


Pmd









Tests on Grouping Scheme 2: Accounting Method and Change in

Exchange Adjustments. In the second classification scheme, firms are

grouped into six portfolios by the three accounting methods and the

two levels of the variable, Change in Exchange Adjustment (CEA).

Table 18A presents the number of firms in each portfolio in each year.

As 1973 was the first year of our data base on firms, the CEA variable

could not be calculated for 1973. The first available CEA value was

for 1974. Therefore, for grouping scheme 2, no tests were conducted for

the first test period (1973 and 1974). In Table 18A, the cells with an

asterisk in them indicate the portfolios that were not included in the

statistical tests due to inadequate sample size, for the corresponding test

periods. Results of the tests on weighted return differences are

presented in Table 18B, along with the T2, F and X statistics for each

period. The F value is not significant for any of the four test periods,

implying that the variable CEA does not improve the partitioning of

firms beyond translation methods. More perplexing is the result that

the X2 statistic is significant at the 7% level in period 4 (1976).

The results of tests on residuals are shown in Table 18C. As in

the case of grouping scheme 1, the F statistic is significant for the

comparisons of 1974 vs. 1975 and 1975 vs. 1976. Again, there is no

significant difference between the residuals of 1974 and 1976.

All the tests on grouping scheme 2 were repeated with a filter of

3% on the exchange adjustment of firms. The number of firms in each

portfolio, the results of the cross-sectional tests and the time-series

tests are shown in Tables 19A, 19B and 19C respectively. The results









TABLE 19

Grouping Scheme 2: Ac hinting g Method and CEAb
Filter = 3%

A. Number of Firms in Each Portfolio in Each Year


Portfolioa


Pci

Pcd

Phi

Phd

Pmi

Pmd


Pci

Pcd

Phi

Phd

Pmi

Pmd


1*

5

2

7

37

82


Returns not usable in comparisons.

aDefined in Chap. III.


bCEA = Change in Exchange Adjustment.

Note: FILTER = ABS (Exch. Adjustment
Note: FILTER Net Income


Firms


Test


Control









TABLE 19 (continued)

B. Results of Tests on Weighted Return Differences


Test Period: 2 (1974)


T2 = 9.77,


F = 0.88;


N = 12

2 (21 df) = 37.28, (p


Pcd


Phd


Means of Return
Differences
(Test-Control) X 100

Standard Deviation of
Return Differences X 10


t Values


1.03 -1.68


1.0


.35 -2.3*


2.5 -.77 -1.45 -.46


1.34 .37

.64 -.71


-.53 -.17


Test Period: 3 (1975)


2 = 18.3
T = 18.3,


F = 1.66;


N = 12

2 (21 df) = 22.34, (p = 37.9%)
X (21 df) = 22.34, (p = 37.9%)


Pcd


Means of Return
Differences
(Test-Control) X 100

Standard Deviation of
Return Differences X 10


t Values


.59 -.24


.81 -.70


-.18 1.29


.31

-.76


Test Period: 4 (1976)


2 = 13.7,
T = 13.7,


F = 1.75;


N = 12

2 (15 df) = 20.5, (p = 16.6
X (15 df) = 20.5, (p = 16.6%)


Pcd


Means of Return
Differences
(Test-Control) X 100

Standard Deviation of
Return Differences X 10


-3.62 -3.36


.54 1.18


----2.32* -.98


.05(*)

F(6,6) = 4.28

F(5,7) = 3.97


.97 -.95 -2.17**


.10(**)

3.05

2.88


Note: Chi-square statistic applies to Box test.


= 1.56%)*
Pmi


t Values


I


1.11 -.53 -1.51










TABLE 19 (continued)


C. Results of


Tests on Monthly Residuals


Test Period: 1974 vs. 1975


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


t Values


N = 12
Pcd PI


1.71 -1.96


Phd


.23 -1.07


.57 .46


1.33 -1.05


.17 -.61


2
T = 39.14, F = 3.558**


Test Period: 1974 vs. 1976


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


N = 12
1 Phi


.35 -1.29


Pmd


-.01 -1.29 1.43


.16 -.88 -.01 -.96


T = 7.78, F = 0.99


Test Period: 1975 vs. 1976


Means of Residual
Differences
(Before-After) X 100

Standard Deviation
of Residual
Differences X 10


t Values


Pci


2.32 -1.35 -.24 -.22 1.08


--- 1.82** -.74


-.34 -.30 3.12*


T2 = 47.2, F = 6.01
T = 47.2, F = 6.01*


.05(*)

F(6,6) = 4.28
F(5,7) = 3.97


Pmd


t Values


N = 12


.10(**)

3.05
2.88










are basically similar to the results in the 1% filter case. However,
2
as shown in Table 19B, now the X statistic for the test on covariance

of return differences is not significant in period 4, but is significant

at the 1% level in period 2. The above result is unexpected, and possibly

a chance outcome.

Table 19C shows that in the 3% filter case, the residuals in 1975

are still significantly different from the residuals in 1976 at the 5%

level. But for the 1974 vs. the 1975 comparison, the F statistic is

now significant at the 10% level instead of the previous 5% level found

for the 1% filter case.

At this stage some comments can be made about the effect of varying

the magnitude of the filter on the materiality of exchange adjustments.

These comments are based on the results of the test on grouping schemes 1

and 2, and not on any formal tests. It seems that a 1% filter provides

results generally comparable to the 3% and 4% filters. The effect of

increasing the magnitude of the filter on the testing for significant

differences in the means of the return distribution vectors was very

weak. Only group 1 was affected in test period 4. However, the indi-

vidual portfolio t values for firms using the C method were affected

by changes in the magnitude of the filter in test periods 1-3. The

effect on the test statistic (X2) used in determining differences in

the variances of return distributions was mixed, and one cannot claim

any advantage of a larger filter. The effect of a larger filter on the

F statistic used in comparing the means of the residual return distri-

butions was to reduce significant differences in the 1974 vs. 1975









comparisons. To sum up, using a 1% filter seems appropriate as it

permitted a larger sample size without substantially affecting the

outcome of the test results.

Tests on Grouping Scheme 3: Accounting Method and Magnitude of

Foreign Revenues. In the third classification scheme, six portfolios

are formed by grouping firms by the three accounting methods and the

two levels of the magnitude of foreign revenues (FR). The filter

used is 1% only. The number of firms in each portfolio for each year

is shown in Table 20A. The tests performed are identical to the tests

on grouping schemes 1 and 2. The results of the cross-sectional

tests are shown in Table 20B and the results of the time-series tests

are provided in Table 20C. In Table 20B, the F value is insignificant

in each of the four test periods, implying that the means of the return

distributions of the test and control firms are not different in either

the pre- or post-FASB No. 8 period. The results of Box's X2 test tell
2
another story. The X value is significant at the 5% level in all

periods except 1974. In 1974, the X2 value is significant at the 6.8%
2
level. The X test results suggest that MNC grouped by accounting

method and magnitude of foreign revenues have return covariances different

from that of domestic firms. We can speculate about the cause of the

above results. The fact that the return distributions are different in

each of the four periods leads to the suspicion that the implementation

of FASB No. 8 is not a factor. Considering the results of grouping

scheme 1 and the fact that most firms have shifted from portolios Pcl

and Pcs to Pml and Pms3 with each passing year, it seems that the

particular accoutning method used may not be an adequate explanation.













Grouping Scheme 3:


A. Number of


TABLE 20

Accounting Method and Foreign Revenues
Filter = 1%

Firms in Each Portfolio in Each Year


'76

7*

2*

5

7

70

93


Portfolioa

Pcl

Pcs

Phl

Phs

Pml

Pms


Pcl

Pcs

Phl

Phs

Pml

Pms


7*

1*

5

6

71

90


Returns not usable in comparisons for this portfolio.


aDefined in Chap. III.

Note: FILTER = ABS (Exch. Adjustment
Net Income


Firms


Test


Control


--







TABLE 20 (Continued)
B. Results of Tests on Weighted Return Differences


Test Period: 1 (1973 & 1975)


T2 = 8.72,
T = 8.72,


F = 1.13;


N = 24
X2 (21 df) =


33.1, (p = 4.4%)*


Means of Return Diffs.
(Test-Control) X 100
Standard Deviation of
Return Diffs. X 10
t Values


-1.28


.27
-2.31*


.08 -.18 -.23 -.34 -.41


.11 -.22 -.26 -.39 -.31


Test Period: 2 (1974) N = 12

T2 = 10.84, F = 0.98; X2 (21 df) = 31.3, (p = 6.8%)**

Pcl Pcs Phl Phs Pml Pms
Means of Return Diffs.
(Test-Control) X 100 -1.56 -.00 -1.21 .07 -.56 -.03
Standard Deviation of
Return Diffs. X 10 .22 .45 .41 .55 .50 .79
t Values -2.41* -.00 -1.01 .19 -.36 -.01

Test Period: 3 (1975) N = 12
T2 = 8.59, F = 0.78; x2 (21 df) = 36.29, (p = 2.0%)*

Pcl Pcs Phl Phs Pml Pms
Means of Return Diffs.
(Test-Control) X 100 .90 1.50 .04 -.50 -.15 -.01
Standard Deviation of
Return Diffs. X 10 .21 .43 .75 .34 .21 .20
t Values .14 1.19 .02 -.11 -.25 -.02

Test Period: 4 (1976) N = 12

T2 = 4.21, F = 0.76; X2 (10 df) = 37.78, (p = 0.01%)*

Pcl Pcs Phl Phs Pml Pms
Means of Return Diffs.
(Test-Control) X 100 --- --- -.45 -.87 -.81 -.89
Standard Deviation of
Return Diffs. X 10 --- --- .62 .56 .17 .18
t Values --- --- -.25 -.53 -1.63 -1.64


Note: Chi-Square statistic


.05(*)
F(6,18) = 2.66
F(6,6) = 4.28
F(4,8) = 3.84
applies to Box test.


.10(**)
2.13
3.05
2.81










TABLE 20 (rtntinued)

C. Results of Tests on Monthly Residuals


Test Period: 1973 & 1974 vs. 1975 & 1976


Means of Residual Diffs.
(Before-After) X 100

Standard Deviations of
Residual Diffs. X 10

t Values
2
T = 1.47, F = 0.32
Test Period: 1974 vs. 1975


Means of Residual Diffs.
(Before-After) X 100

Standard Deviations of
Residual Diffs. X 10


t Values


N = 24


Pcl Pcs Phl Phs Pml Pms


.14 .61


.45


.23


.34 .53


.13 1.12


N = 12

Pcl Pcs Phl Phs Ptil Pms


-1.18 -.82 -.50


-.55 -.58 -.31 1.22


.17 .39


.63 .19

.97 .68


2
T2 = 15.7, F = 1.43
Test Period: 1974 vs. 1976 N = 12

Pcl Pcs Phl Phs Pml Pms
Means of Residual Diffs.
(Before-After) X 100 -- --- -.11 -.56 .49 1.09

Standard Deviations of
Residual Diffs. X 10 -- --- .43 .30 .56 .22

t Values -- -- -.08 -.63 .30 1.67
2
T2 = 4.23, F = 0.77
Test Period: 1975 vs. 1976 N = 12

Pcl Pcs Phl Phs Pml Pms
Means of Residual Diffs.
(Before-After) X 100 --- --- .39 -1.21 .32 .70

Standard Deviations of
Residual Diffs. X 10 --- --- .49 .24 .26 .11

t Values --- --- .27 -1.68 .41 2.04*
T2 = 44.9 F = 8.16
T = 44.9, F 8.16*


F(4,20)
F(6,6)
F(4,8)


.05(*)
= 2.87
= 4.28
= 3.84


.10(**)
2.25
3.05
2.81










Grouping by the magnitude of foreign revenues seems to be the critical

factor. Possibly, the environmental factor of floating exchange

rates in the 1973-1977 period affects firms with differential foreign

revenues differently. One problem may be the noise in the foreign

revenue data. Interpretation of the above results is not straightforward,

and some better explanation may be available.

Table 20C shows that in the time series comparison of residuals, the

F value is significant at the 1% level in the 1975 versus 1976 comparison,

implying that the returns grouped according to scheme 3 were significantly

different in the 1976 report period, as compared to the 1975 report

period returns. Comparisons in other periods did not show significant

results.

Tests on Grouping Scheme 4: Accounting Method and Sign of the Exchange

Adjustment. As shown in Chapter III, Table 10, six portfolios are

formed by grouping firms according to the three accounting methods and

the sign of the aggregate exchange adjustment reported each year.

Table 21A gives the number of firms in each portfolio in each year.

The results of the cross-sectional tests on risk equalized monthly

returns are provided in Table 21B. The table shows the four periods

over which comparisons were made. For each period, N is the number of

monthly observations used. As can be seen from Table 21B, none of

T2 values are significant for scheme 4. This implies that we cannot

reject the hypothesis of no difference between test and domestic firms

in any of the four periods. The univariate t values and the means of

the return differences do not exhibit any clear pattern. Though the

returns of the MNC are lower than the domestic firm returns for most of










TABLE 21

Grouping Scheme 4: Accounting Method and Sign of Exchange


A. Number

Portfolio

Pc+

Pc-

Ph+

Ph-

Pm+

Pm-


Pc+

Pc-

Ph+

Ph-

Pm+

Pm-


Adjustment (SEA)
Filter = 1%

of Firms in Each Portfolio


for Each Year


'73 '74 '75 '76 Firms


1*

8

3

9

52

107


1*

7

4

7

55

106


Test


Control


*


Returns not usable in comparisons.

aDefined in Chap. III.
Exch. Adjustment
Note: FILTER = ABS Net Income








TABLE 21 (continued)

B. Results of Tests on Weighted Return Differences


Test Period: 1 (1973 & 1974) N = 24
T2 = 9.16, F = 1.194; X2 (21 df) = 25.9, (p = 20%)
Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Return Diffs.
(Test-Control) X 100 -.43 -.65 -.59 .49 -1.19 .55
Standard Deviation of
Return Diffs. X 10 .32 .25 .39 .42 .53 .57
t Values -.65 -1.26 -.75 .57 -1.09 .47

Test Period: 2 (1974) N = 12
T2 = 15.25, F = 1.38; X (21 df) = 18.0, (p = 64.3%)

Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Return Diffs.
(Test-Control) X 100 -.26 -1.45 -.78 -.03 -1.17 -.70
Standard Deviation of
Return Diffs. X 10 .39 .18 .43 .52 .70 .68
t Values -.22 -2.73* -.62 -.02 -.58 -.50

Test Period: 3 (1975) N = 12
T2= 30.49, F = 2.77; 2 (21 df) = 28.5, (p = 12.5%)
Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Return Diffs.
(Test-Control) X 100 1.24 1.20 -1.12 .17 .85 -.79
Standard Deviation of
Return Diffs. X 10 .39 .28 .61 .30 .12 .24
t Values 1.09 1.49 -.63 .19 2.37* -1.11

Test Period: 4 (1976) N = 12


T2 = 19.1,
T = 19.1,


F = 2.44;


Means of Return Diffs.
(Test-Control) X 100
Standard Deviation of
Return Diffs. X 10
t Values


F(6,18)
F(6,6)
F(5,7)


X (15 df) = 24.3**, (p = 5.99%)
Pc+ Pc- Ph+ Ph- Pm+


--- -2.66 -.83


--- -2.28* -.40


.05(*)
2.66
4.28
3.97


-.04 -.20


.31 .18
-.05 -.38


-1.12


.10(**)
2.13
3.05
2.88


Note: Chi-square statistic applies to Box test.


- -







TABLE 21 (continued)

C. Results of Tests on Monthly Residuals


Test Period: 1973 & 1974 vs.


1975 & 1976


Pc+ Pc-


Means of Residual Diffs
(Before-After) X 100.
Standard Deviation of
Residual Diffs. X 10
t Values
2
T = 2.39, F = .395


--- .30 -.04


.81 -.26


.47


-.06


.94 -.30


Test Period: 1974 vs. 1975 N = 12

Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Residual Diffs.
(Before-After) X 100 -.78 -1.36 .52 .15 -.64 .73
Standard Deviation of
Residual Diffs. X 10 .60 .55 .39 .35 .49 .45
t Values -.44 -.84 .45 .15 -.45 .55
T2 = 4.09, F = 3.72**
Test Period: 1974 vs. 1976 N = 12

Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Residual Diffs. --- .67 -1.30 .13 -.16 1.24
(Before-After) X 100
Standard Deviation of
Residual Diffs. X 10 --- .61 .29 .37 .45 .47
t Values -- .37 -1.50 .12 -.12 .89

2
T2 = 9.76, F = 1.24
Test Period: 1975 vs. 1976 N = 12

Pc+ Pc- Ph+ Ph- Pm+ Pm-
Means of Residual Diffs.
(Before-After) X 100 --- .02 -.01 -1.5 .48 .50
Standard Deviation of
Residual Diffs. X 10 --- .38 .47 .30 .25 .08
t Values --- .01 -.01 -1.7 .67 2.08**
T2 = 37.9, F = 4.82*
T = 37.9, F = 4.82*


F(6,6)
F(5,19)
F(5,7)


.05(*)
4.28
2.74
3.97


.10(**)
3.05
2.18
2.88


N = 24


Ph+


Pm+











the portfolio-periods, no inference can be drawn. The results of
2
Box's X tests on the covariance matrices are also shown in Table 13B.
2
None of the X values are significant at the 5% level. However, the
2
X for 1976 is significant at the 5.99% level, implying that the return

distributions of the MNC may be different from the return distributions of

the matched domestic firms, for the 1976 period.

The results of the time series tests on grouping scheme 4 are

shown in Table 21C. The F statistic is significant at the 10%

confidence level for the 1974 vs 1975 comparison and significant at

the 5% level for the 1975 vs 1976 comparison. This implies that the

hypothesis of no difference in returns over time can be rejected for

the 1975-1976 period.

Tests on Grouping Scheme 5: By Accounting Method Used in Pre-FASB

No. 8 Period. Results of the tests on return residuals (Uit) on

grouping scheme 5 are shown in Table 22B. In this table, residuals

of test firms are compared in the before and after FASB No. 8 periods to

see if there is any effect of the accounting change. Table 10,

Chapter III shows that portfolios Pcm and Phm consist of those firms

that were using the C and H accounting methods, respectively, and

switched to the M method after FASB No. 8. Portfolio Pcm consists

of firms that used the M method before and after FASB No. 8. Also,

it is worth emphasizing that, as opposed to grouping schemes 1 through

4, in scheme 5 each portfolio consists of the same firms in each

period. The number of firms in each portfolio is shown in Table 22A.

Apart from comparing the residuals for the aggregate period of 1973 and

1974 vs 1975 and 1976, comparisons are also made between 1974 vs 1976

and 1975 vs 1976.