Evaluation of a Multimedia Extension Program in Honduras
By Michael J. Martin and Timothy G. Taylor*
*The authors are Graduate Research Assistant and Professor, respectively at the University of Florida, McCarty Hall, Department of Food and Resource Economics, Gainesville, FL 32602.
Technology adoption is a critically important issue in developing countries where rising populations depend directly on dwindling resource bases. New technologies have rapidly become available to small-scale agricultural producers in Central America whose production systems, rooted in ancient Mayan traditions, provide daily sustenance. Slash and burn methods of field preparation developed reflexively over generations, as farmers sought to maximize efficiency within the physical and institutional parameters imposed on them. Abundant land resources of Mayan civilizations permitted farmers to abandon fields long enough to become naturally rejuvenated. However, unlike their ancestors, modem-day Central American farmers are not afforded this luxury. Colonial institutions and population growth have tightened land constraints and forced farmers to cultivate the same land year after year.
Reduction in soil fertility has provided impetus for farmers in Central America and other developing areas to seek alternative production methods. Expanded markets and complementary technologies have provided further impetus. The capacity of farmers to respond to such pressures often corresponds to their ability to purchase complementary human capital, either in themselves or through hired labor. Small farmers, however, usually lack the funds to purchase necessary assistance for technology adoption.
Education and training have long been considered indispensable to technology diffusion and the correction of attendant economic disequilibria.'
Lack of skill and knowledge lie at the root of production inefficiencies which have restricted production systems in developing countries from reaching their technical frontiers.'
Local research stations often demonstrate that new and feasible technologies enhance production efficiency, but they are often not adopted by agricultural producers. Early studies showed that locally developed technologies were economically important and that human capital enhanced farmer efficiency., Human capital developed by formal schooling and extension training has also been shown to have a positive influence on both technology adoption and production efficiency.! Nutrition and religion are other forms of human capital that are considered to have an economic impact.s
Several studies have examined the influence of education and training on productivity. Jamison and Lau (1982) showed a strong link between education and productivity, but a weaker link between extension services (as measured by site visits) and productivity.' However, Shapiro and Miller (1977) found that a farmer's exposure to extension and other non-formal education have a positive effect on output.7 Kalirajan and Shand (1985) compared the impact of formal versus non-formal education on Indian rice farmers. Their results indicate that farmers' non-formal education, as measured by "understanding", had a significant impact on yield, but the effect of education was insignificant. "Understanding" was determined by testing how well farmers could identify the "best practice" as defined by local agricultural experts. Interviews indicated that understanding was
achieved from mass media, neighbors, landlords and only slightly from extension agents. Most farmers, especially those with low educational levels, relied on mass media as their primary source of technological information.!
The purpose of this article is to examine the impact of a multimedia extension program which promoted a variety of technologies that varied across crops and regions. Several previous studies have examined how extension influences the adoption of specific technologies? but none to date have analyzed the effectiveness of different types of communication methods used by extension services to diffuse a wide assortment of technologies. In addition, rather than measure extension impact by the usual procedure of enumerating site visits, this study allows the farmer to identify the source of learning instruction for the technologies upon which he relies. Farmers were scored as either adopters or non-adopters of technology through a survey instrument that asked series of openended questions regarding their agricultural practices. No producer was asked directly whether he adopted a recommended technology, but all were asked where, or from whom, they learned the technologies they employed.
As well as investigating the relative effectiveness of different extension methods, this study examines demographic characteristics related to human capital, land quality and institutional structure. Demographic information is included to enhance the explanatory power of the model and to provide insights to policy makers regarding which groups may be effectively targeted.
Thus, two types of exogenous forces that contribute to farmers' decisions
to adopt new technologies are deciphered. The first are demographic characteristics that affect producer amenity to new methods, the second is the direct training obtained from both formal and casual sources. The following section describes extension challenges as they pertain to Honduras, the country in which the study was conducted. The methodology used to analyze the influences of dichotomous choices is discussed in the third section. Empirical results are then presented, followed by conclusions and implications.
HUMAN CAPITAL AND TECHNOLOGY DIFFUSION IN HONDURAS Technology diffusion is hampered by low levels of human capital in Central America in general, and Honduras in particular. Functional literacy levels in rural Honduras stand at about fifty percent.o Institutional factors also influence the accessibility and appropriateness of technology. Land tenure arrangements of small farms in Central America are tenuous, increasing the risk associated with new technologies. Credit is available on a sporadic basis, reducing producers' capacity to sustain newly adopted technologies. Farm and household size vary in the region as well, playing important roles in the selection of technologies.
Technology transfer in rural Honduras is confronted with some formidable obstacles. Assistance programs to introduce new technologies are often criticized for not taking into account a common farmer's opportunity set, production constraints and, perhaps most important of all, mental preparedness to respond
to assistance. If outside technology is introduced to farmers that lack the skills to interpret and screen relevant information, then the "spill in" of the technology will not occur. Compounding the human capital and institutional problems is the geographical dispersion of Honduran farmers. Many live far from main and sometimes even remote transportation arteries. Providing training and visit extension services (T&V) is usually prohibitively expensive. One logical solution has been mass media. Radio and printed materials offer ways to reach widely dispersed farmers at a much lower cost. However, most farmers in the area are illiterate. And though many farmers own a radio, most have found generic agricultural radio programs unresponsive to their specific needs.
Problems such as these form the basis of a multi-media approach to transfer technology that attempted to respond to particular production needs in the face of funding constraints on extension. The program, located in the Comayagua region of Honduras, ostensibly tailored assistance programs to fit the geographical, cultural and capital-base conditions of small farmers.12 Vehicles of instruction included T&V, elementary pamphlets with self-explanatory illustrations, and a regularly broadcast radio program that covered common agricultural problems. The objective of the program was to design and diffuse key technologies amenable to farmers who operate with diverse production sets. All recommended technologies were designed to respond to the farmers' individual needs.
EMPIRICAL MODEL AND DATA
Ultimately, the effectiveness of an extension program can be gauged by the increased efficiency or profit of farmers operating in the target area. A necessary intermediate step in that improvement is the actual adoption of the technology, the simple yes or no decision of the farmer. Evaluating adoption rates requires far less data than evaluating economic responses and may be conducted at early program stages to monitor the relative effectiveness of promotion techniques.
A common procedure in technology adoption studies is to examine dichotomous adoption decisions as a function of factors related to the farm production system or the farmer's background. Most studies enumerate site visits to measure the impact of extension services." A few have attempted to directly identify the source of the farmer's knowledge regarding a technology." This study analyzes adoption rates as a function of primary learning sources as identified by the farmer. It also incorporates firm, human capital and policy factors in evaluating farmers' propensity to adopt a range of technologies that differ across crops and regional boundaries.
Farmers are divided into two groups, commercial, who produce tomatoes and rice, and subsistence who produce maize and beans. Regressions were run separately for commercial and subsistence farmers because they constitute two distinct target groups for extension services and consequently merit separate examination of the factors that influence adoption.
Analyzing the decision to adopt or not adopt a given technology in a
regression equation requires the specification of a binary dependent variable. Probit or logit models" have been used extensively to investigate factors that influence technology adoption. Assuming the basic linear form of these models is given by:
Wi = 3'xi + e
where = 1 if adopt
= 0 otherwise
It follows that
Prob(Wti = 1) = Prob (ei > -1'x)
= 1 F(-3'x)
where F is the cumulative distribution function for e.
The program was designed to assist farmers who produce standard crops in diverse regions that require different technologies. Farmers operating in high altitude zones on poor soil received different recommendations regarding seed spacing, fertilization practices and weeding techniques, for example, than their counterparts producing the same crops in other areas. Evaluating each technology individually was not statistically feasible because there were not enough degrees of freedom for any one technology. Thus all technologies are examined in each regression as functions of common variables which influence adoption. Careful selection of critical technologies allows for the assumption to be made that all technologies are equally weighted in importance and accessibility."' All technologies considered in this paper were determined through fieldwork to be
key technologies by extension agents, researchers and farmers.
Farm size, human capital, labor availability and land tenure are considered important characteristic variables that explain adoption rates." The communication program in the Comayagua region was designed to promote technologies that complement local farmers' cultural and technological environments. The communication program thus attempted to tap into farmers' propensity to adopt improved technologies.
The following logit model, which applies to both commercial and subsistence farmers, includes two sets of variables which influence farmers' decisions to adopt new technologies. The first set (variables associated with parameters fl, to fl9) relates to the propensity to adopt technologies, and the second set (variables on parameters #j10 to 1316) relates to technology promotion: ADOPT = o + flRESIDE + 02AGE + MEMBERS +
SCHOOLING + flsLITERACY +fl6SOLD + fl7REFORM
+ fl8HILL + flPLAIN + 13oRADIO + 0 EXT/GOV + 812FAMILY + 013PHAMPLET + 914SALESMAN +
15FRIEND + 316EXT/PVO.
ADOPT = 1 if farmer adopted the technology, 0 otherwise.
RESIDE = Number of years producer has lived in area.
MEMBERS = Number of household members.
SCHOOLING = Number of years of schooling.
LITERACY = 1 if producer is literate, 0 otherwise.
ILLITERACY" = 1 if producer is illiterate, 0 otherwise.
SOLD = Proportion of output sold.
REFORM = Manzanas19 of agrarian reform lands used in
PRIVATE = Manzanas of private, non-agrarian reform lands used
HILL = Amount of substandard hill lands used in production.
PLAIN = Amount of preferred level lands used in production.
RADIO = 1 if producer learned the technology from radio,
EXT/GOV = 1 if producer learned the technology from government
extension agents, 0 otherwise. FAMILY = 1 if producer learned the technology from a family
member, 0 otherwise.
PAMPHLET = 1 if producer learned the technology from a pamphlet, O otherwise.
SALES = 1 if producer learned the technology from a
salesperson, 0 otherwise.
FRIEND = 1 if producer learned the technology from a friend,
EXT/PVO = 1 if producer learned the technology from a PVO
extension agent, 0 otherwise. CUSTOM1 = 1 if producer uses technology as matter of custom,
Birkhaeuser, et al."2 (1991) offer a virtual litany of common characteristics that may bias results of extension impact studies. Endogeneity may introduce bias because successful farmers are frequently sought out by extension agents. Similarly, extension efforts directed at productive or fertile areas may show unjustifiably high adoption and productivity rates unless the type of land is accounted for in the model. Indirect or secondary information flows present another bias in that many farmers may learn from friends or sales people. Generally, the presence of secondary information flows causes extension impacts to be understated. Farmers who have had no direct contact with extension agents often learn from farmers who have. Institutional surroundings also influence adoption rates because farmers in isolated areas, out of communication with other farmers depend directly on site visits, whereas farmers who inhabit more densely populated areas or belong to cooperatives could learn from their neighbors or colleagues. Site visits would thus have a multiplier effect.
The model and the survey technique attempted to account for such estimation biases. In order to avoid spurious correlations between T&V and adoption, farmers were asked to directly identify their source of instruction. Thus, farmers who received a relatively high number of site visits but learned new technology from a friend or a salesman could be so accounted. Human capital factors regarding schooling, literacy and age also provide some indication as to whether or not "more productive" or "well-known" farmers adopt technologies and thus may be sought out by extension agents.21
Inferior (HILL) land may inhibit farmers from adopting new technologies (although the extension system was ostensibly designed to assist such marginalized producers). Similarly, farmers who operate on agrarian reform lands (REFORM) may be more amenable to new technologies because they are operating in a new environment and because they are more involved in cooperative associations where new technologies are taught and demonstrated. One shortcoming of this approach is that partial influences of learning sources are disregarded since primary learning sources were scored as the sole source of instruction. Radio and pamphlets, for example, may supplement the identified source of instruction, but remain unrecognized by the model. On the other hand, this method allows adopters of new technologies to identify the most effective instruction method.22
The data used for this analysis were gathered in the Comayagua region of Honduras in early 1990. Variable distributions and frequencies (for dichotomous variables) are displayed in Table I. All of the surveys followed the same format for inquiring into the adoption of technology. Farmers were asked a series of open ended questions about the technological methods they employed, and were then asked to identify the source from which the technology was learned. The intent of the survey structure was to provide a consistent manner of evaluating technology adoption for different crops and geographical zones covered by the communications program.
The samples of farmers were randomly drawn from a larger randomly drawn sample used in ongoing research to update and adjust extension programs.3 Four crops, corn, beans, rice and tomatoes were investigated in the Comayagua region. The former two are traditional subsistence crops and the latter are grown primarily for commercial markets. Each interviewee was queried about three technologies for the preceding two production seasons of one crop. Technologies were selected based on their availability and importance to improving yield. Fertilization practices were examined regarding all four crops; planting practices were examined for corn, rice and beans; weed control was examined for corn and rice; insect control was examined for tomatoes and beans; and soil preparation was examined for tomatoes. The number of farmers surveyed for each crop varied proportionately with the total number of farmers in the area who produce that crop. Data were gathered in January and February of 1990. Only full adoption of a technology was considered a successful adoption.'
The model was estimated under the assumption that the cumulative distribution function was logistic. Thus, a logit model was used in the analysis. Estimation was accomplished using maximum likelihood in SPSS (Statistical Package for the Social Sciences).
Summary statistics (TABLE II) indicate that the model provides an
acceptable "fit" to the data. Correctly classified estimates (based on a 50-50 classification scheme) amounted to 90.7 percent for subsistence crops and 70 percent for commercial crops. The model Chi-Square, which tests the overall significance of the model, was significant at the one percent level for both regressions.
Coefficient estimates are shown in Table III. Seven of the subsistence crop coefficients and five of the commercial crop coefficients are significant at the ten percent level. The coefficients indicate that the propensity to adopt new technologies is different between commercial and substance farmers, but that a common pattern exists for learning source influences.
All of the statistically significant learning sources coefficients are positive. This indicates that the survey instrument succeeded in distinguishing between producers who employed traditional methods and those who adopted advanced methods without asking the question directly. Survey questions were open-ended; farmers were asked to describe planting techniques, fertilization practices, etc. Respondents who improperly adopted a technology were scored as non-adopters by survey takers prior to identifying the learning source. Farmers who thought they were adopting a new technology promoted by the extension program could not be scored as an adopter unless the method they used satisfied all the recommendations of the technology in question. Correct practices varied by both
crop and by region.
The most salient result of the estimates on learning source is that personal contact is important in promoting new technologies. The coefficients on government extension agents (EXT/GOV) are positive for both subsistence and commercial producers and statistically significant. Salespeople also positively influenced adoption for both commercial and subsistence producers; the coefficients on SALES are significant at the ten percent level. FRIEND showed a positive and significant influence on adoption rates for commercial producers only.
It would be interesting to decipher the impact of FAMILY between parents and siblings (in the context of Central American agriculture, fathers and brothers). Technologies passed from father to son are almost by definition traditional technologies, but brothers may pass new information on to one another
- particularly older to younger brothers. This could account for the insignificance of the estimate on FAMILY. It might also serve to explain the different outcomes of the constant terms, which represent illiterate farmers who operate on private lands and use traditional technology.26 Tomato and rice technologies were developed relatively more recently then maize and bean technologies and are less likely to be considered customary. Customary practices for the traditional subsistence crops of maize and beans may actually represent recommended technologies in some zones, especially for young producers.
PAMPHLET and RADIO were cited as primary learning sources by few
producers and do not bear strong results in terms of adoption. However, their marginal returns may be much higher for extension services in light of the large discrepancy between costs of mass media programs and personal site visits.
Propensity for Adoption
Coefficients which measure the propensity for adoption show no immediate similarity between commercial and subsistence crops. However, results do conform to the Boserup hypothesis that population pressures induce technological innovation.' The only such factor which showed a significant influence on commercial farmers is the proportion of output sold. The most curious results for subsistence farmers are the negative and statistically significant estimates of both types of land use, HILL and PLAIN. Subsistence farmers with relatively extensive land holdings are less inclined to incorporate technologies that are designed to improve land productivity.
Subsistence farmers who operate on agrarian reform lands also demonstrated a positive inclination to utilize new technologies. The coefficient on REFORM was positive and significant at the five percent level. Farmers who operate on agrarian reform lands are obliged to belong to cooperatives which provide credit, inputs and technical assistance. Moreover, farmers in cooperatives are more likely to see technologies tested on collective production plots. This has important implications for cooperatives which, in the agrarian reform sector, are often considered to be net social losers due to administrative
corruption and financial mismanagement.
Oddly, a subsistence farmer's capacity to read and write bears a negative influence on adoption. The same estimate for commercial farmers is positive, but less significant. The opposite signs could be attributed to the fact that maize and bean production is an aside for some farmers in the subsistence crops sample. With their main source of income in other endeavors, they have not considered the marginal product obtained from increasing basic grain production adequate to cover the cost of learning or implementing the necessary technologies.
SUMMARY AND CONCLUSIONS
Technology diffusion presents major challenges to international development. Efficiency enhancing techniques will have a greater impact the more they are targeted to receptive groups via effective communication methods. This paper analyzed the effectiveness of a multi-media extension program targeted at Honduran farmers who have low human capital investments and whose traditional farming methods are being threatened by radical changes in their resource bases.
Dichotomous, yes/no decisions were examined in a binary logistic regression. The survey through which the data were gathered asked farmers to identify their source of learning to account for bias problems which enumerate site visits. Institutional and human capital variables were included to diminish other forms of bias and to identify which groups of farmers are amenable to
adopting new technologies.
An entire extension program for a region in Honduras was analyzed to determine what types of farmers are responsive to new technologies and what types of training promote the adoption of those technologies. Producers of commercial crops demonstrated an increasing inclination to adopt new technologies as the they increased the market proportion of their output. Subsistence farmers, on the other hand, demonstrated a negative correlation between land extensiveness and technology adoption. Both results conform to Boserup's. Cooperative association, and perhaps more importantly the communication and credit amenities that accompany it, also had a positive influence on adoption.
The results indicated that T&V extension methods have a multiplier effect through personal contact of experts and friends, and are very effective in motivating technology adoption in contrast to impersonal multi media techniques. Nonetheless, mass communication media cannot be discounted and in fact, given their low cost of making contact per farmer, present a viable alternative. The relative pay-off between personal and mass communication offers potential for further research.
1. Schultz, T.W., "The Value of Ability to Deal with Disequilibria," Journal of Economic Literature, 13 (1975): 827-46. Nelson, R.R and E.S. Phelps, "Investment in Humans, Technological Diffusion and Economic Growth," American Economic Review, 56 (1966): 69-75.
2. Birkhaeuser, Dean Robert E. Evenson and Gershon Feder, "The Economic Impact of Agricultural Extension: A Review," Economic Development and Cultural Change, 39 (1991):607-50.
3.Grilliches, Zvi, "The Sources of Measured Productivity Growth: United States Agriculture, 1940-60." Journal of Political Economy, 71 (1963): 331-46. Schultz, T.W., 1954, The Economic Organization ofAgriculture, McGraw Hill.
4. For review, see Jamison, Dean T. and L.J.Lau, Farmer Education and Farmer Efficiency, (Baltimore: The Johns Hopkins University Press; 1982), Birkhaeuser, Evenson and Feder, 1991 Op Cit.; and Feder, Gershon, Richard E. Just and David Zilberman, "Adoption of Agricultural Innovation in Developing Countries: A Survey", Economic Development and Cultural Change, 33 (1985):255-98.
5. Bliss, Christopher and Nicholas Stem, "Productivity, Wages and Nutrition," Journal ofDevelopment Economics, 5 (1978): 331-97, examine Leibenstein's full wage hypothesis; The link between childhood malnutrition and schooling is
examined by Moock, Peter R. and Joanne Leslie, "Childhood Malnutrition and Schooling in the Terai Region of Nepal," Journal ofDevelopment Economics, 20 (1986): 33-52; Selowsky, Marcelo and Lance Taylor, "The Economics of Malnourished Children: An Example of Disinvestment in Human Capital," Economic Development and Cultural Change, 22 (1973): 17-30; and Jamison, Dean T. and M.E. Lockheed, "Participation in Schooling: Determinants and learning Outcomes in Nepal," Economic Development and Cultural Change, 35 (1987): 279-306.
6.Jamison and Lau (1982), Op. Cit. conducted their own study on data sets from Thailand, Korea and Malaysia. Their findings also lend support to Schultz's hypothesis that education is important in a changing environment.
7.Shapiro, K.H. and J. MUller, "Sources of Technical Efficiency: The Roles of Modernization and Innovation," Economic Development and Cultural Change, 25(1977): 293-310.
8.Kalirajan, K.P. and R.T. Shand, "Types of Education and Agricultural Productivity: A Quantitative Analysis of Tamil Nadu Rice Farming", The Journal of Development Studies (1985) 232-43.
9. See Birkhaeuser, Evenson and Feder Op. Cit for review, also Jamison and
Lau, Op. Cit.; Feder, Gershon and R. Slade, "The Acquisition of Information and the Adoption of Technology," American Journal ofAgricultural Economics, 66(1984): 312-20.
10.World Fact Book 1988, CIA.
11.Stringer, Randy, 1989 "Honduras: Toward Conflict and Agrarian Reform" in Searching for Agrarian Reform in Latin America, Ed. William C. Thiesenhusen, Boston: Unwin Hyman Inc.
12.Ministry of Natural Resources, Government of Honduras, Tegucigalpa.
13.See Birkhaeuser, Evenson and Feder, 1991 Op. Cit.
14.Kalirajan and Shand, Op. Cit.; J.K.Harper, et al., "Factors influencing the Adoption of Insect Management Technology", American Journal ofAgricultural Economics, 72 (1990): 997-1005.
15.A detailed description is provided by Maddala, G.S., Limited Dependent and Qualitative Variables in Econometrics, Englewood, N.J. Cambridge University Press 1983. See also Amemiya, Takeshi, 1981, "Qualitative Response Models: A Survey", Journal of Economic Literature, 19: 1483 1536.
16.Accessibility was never a problem because it was a criterion in the development of recommended technologies. Some potentially useful technologies were discarded because they could not be reliably imported.
17.Feder et al. 1985, Op. Cit.
18.Deleted from regression to avoid a singular matrix.
19.A manzana equals roughly .706 hectares. 20.1991, Op Cit.
21.This does not wholly account for the endogeneity problem that results from extension agents seeking out farmers who are likely to adopt new technologies. However, endogeneity is more vexatious in studies that consider productivity or technology adoption as a function of site visits than in the present study that examines adoption rates as a function of learning sources because identifying learning sources puts the efficacy of extension under greater scrutiny. More productive farmers may be more exposed to T&V, but ineffectual site visits are discounted.
22.Price instability, the subsidizing of certain technologies and inappropriate
technologies also typically pose problems in dichotomous adoption studies. None characterize the conditions of the region from which the data in this study were drawn.
23.The larger sample was drawn from a census of all farmers living in the area. The survey used for this study was administered on a smaller subsample of farmers and focused more specifically on technology adoption than surveys used for ongoing research.
24.Dichotomous choice models have been criticized for their inability to account for partial adoption (Feder et. al., 1985, Op. Cit.), but partial adoption holds no assurance of positively influencing yield. Increasing the number of seeds planted in each hole, for example, may actually be detrimental if seed and row spacings are not expanded. Partial adoption of complementary components may weaken a given technology.
25.The logistic function is f(91) = e"/(1 + e), where f(%) varies from 0 to 1 as 9t varies from -oo to + co. A common alternative distribution for dichotomous choice models is the cumulative normal distribution used in probit models. Logit and probit models render similar results in small samples. 26.Those characteristics were registered in the variables deleted from the model and thus are embedded in the constant term. See Johnston, J. Ecnometric
Methods, McGraw Hill, New York, New York, 1884. 27.Boserup's hypothesis holds that intensification occurs as population pressure forces farmers to extract higher yields from their fixed land supply. See Boserup, Ester, 1965, The Conditions ofAgricultural Growth: The Economics ofAgrarian Change Under Population Pressures, New York: Aldine Publishing Company.
TABLE I VARIABLE DISTRIBUTION SUBSISTENCE CROPS: COMMERCIAL CROPS: CORN & BEANS TOMATOES & RICE
VARIABLE Percent* Average" Percent Average
RESIDE 31.2 33.4
AGE 44.7 50.2
MEMBERS 6.4 6.3
SCHOOLING 2.8 2.1
LITERACY 66.0 59.3
SOLD 0.3 0.84
REFORM 0.24 0.26
PRIVATE 1.1 0.80
HILL 0.42 0.0
PLAIN 0.9 1.10
RADIO 1.1 0.4
EXT/GOV 12.3 18.2
FAMILY 10.3 2.8
PAMPHLET 0.0 1.6
SALES 1.1 2.77
FRIEND 13.0 26.5
EXT/PVO 3.8 0.8
CUSTOM 60.4 43.5
OBSERVATIONS 172 253
*Percentage figures represent the proportion of occurrences in the sample for binary variables. .Standard deviations are in parentheses.
TABLE II SUMMARY STATISTICS
SUBSISTENCE CROPS: COMMERCIAL CROPS:
MAIZE & BEANS TOMATOES & RICE
Chi-Square Significance Chi-Square Significance
-2 Log Likelihood 88.053 1.000 272.229 0.0577
Model Chi-Square 47.276 0.000 33.581 0.0039
Improvement 47.276 0.000 33.581 0.0039
Goodness of Fit 143.549 0.7538 234.554 0.5327
Correctly Classified 90.7% 69.96%
Senstivity 52.17% 14.86%
Specificity 96.64% 92.74%
TABLE III: MAXIMUM LIKELIHOOD ESTIMATES OF TECHNOLOGY ADOPTION VARIABLE SUBSISTENCE CROPS: MAIZE & BEANS COMMERCIAL CROPS: TOMATOES & RICE
BETA S.E. WALD SIG BETA S.E. WALD SIG
RESIDE 0.0338 0.0218 2.3933 0.1219 0.0004 0.0091 0.0021 0.9638
AGE -0.0960" 0.041 5.4905 0.0191 -0.0016 0.0171 0.0087 0.9255
MEMBERS 0.0790 0.0926 0.7275 0.3937 0.0900 0.0616 2.1367 0.1438
SCHOOLING -0.0568 0.1557 0.1329 0.7154 -0.1206 0.1112 1.1766 0.278
LITERACY -1.4729* 0.8912 2.7315 0.0984 0.7726 0.5313 2.1149 0.1459
SOLD 0.0581 0.6376 0.0083 0.9273 1.7087* 0.9378 3.32 0.0684
REFORM 1.279*" 0.6273 4.1574 0.0415 0.2955 0.2903 1.0365 0.3086
HILL -1.8457* 0.9711 3.6125 0.0573
PLAIN -1.4139*' 0.7008 4.0711 0.0436 0.1363 0.2232 0.3728 0.5415
RADIO -6.3014 62.6672 0.0101 0.9199 7.3872 22.2419 0.1103 0.7398
EXT/GOV 2.1516*' 0.7691 7.8271 0.0051 1.7383' 0.4888 12.6496 0.0004
PAMPHLET 1.2918 1.1767 1.2053 0.2723
SALES 3.2543* 1.7094 3.6246 0.0569 1.4723* 0.8449 3.0367 0.0814
FREIND -0.4132 1.141 0.1312 0.7172 0.8900r 0.3804 5.4741 0.0193
EXT/PVO 1.4977 1.4501 1.0668 0.3017 -4.3742 15.7348 0.0773 0.781
FAMILY -6.8030 26.5146 0.0658 0.7975 0.5311 0.8974 0.3502 0.554
CONSTANT 2.4446 1.9328 1.5997 0.206 -4.0472' 1.1901 11.565 0.0007
*One, two, and three asterisks indicate coefficients that are statistically different from zero at the 0.1, 0.05 and 0.01 levels of probability, respectively.