Déterminants des services de vulgarisation agricole: Le cas d'Haïti

Diego Arias Juan José Leguía Abdoulaye Sy May 24, 2013 Determinants of Agricultural Extension Services: The Case of Haiti LCSSD Occasional Paper Series on Food Prices Latin America and the Caribbean Region Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized El Desarrollo Agropecuario y Rural en Paraguay. Diagnóstico y opciones de política 3 Cover photos courtesy of Mrs Barbara Coello. The work has been partly financed by the Trust Fund for Environmentally and Socially Sustainable Development (TFESSD) Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 1 Diego Ari A s Ju A n José Leguí A Ab D ou LA ye s y Wor LD bA nk, LC s A r May 24, 2013 LC ss D Foo D P AP ers s eries Determinants of a gricultural e xtension s ervices: t he c ase of h ait i LAT in AM eri CA A n D TH e CA ribbe A n region Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 2 This paper extracts relevant lessons from historical data on factors influencing the receipt of extension services in Haiti, taking stock of the use of agricultural extension services prior to the 2010 earthquake. The goal is to influ- ence future policies and development projects involving the provision of extension services as well as the type of extension services offered. This paper uses data from the 2010 Agricultural Cen- sus and examines the characteristics of farmers in Haiti receiving extension services by gender, education, agricultural training, farm size, and type of crop. Through in-depth study of each variable and a review of trends in the receipt of agricultural extension services, the study analyzes the equilibrium between the demand for and supply of extension services to particular farmer groups. u sing a fixed effects probit model to isolate the marginal effect of each characteristic on the likelihood of receiv- ing extension services, and controlling for various factors, the study draws the following nine key conclusions: 1. The proportion of households receiving agricultural extension services in Haiti is non-negligible. 2. Location is an important determinant of the recipients of agricultural extension services. executive summar Y 3. There are no statistical differences between men and women in terms of receipt of extension services; how- ever, the impact of agricultural training and farm size change when the head of household is a woman. 4. Education level has a positive, yet small, effect on re- ceiving extension services. 5. Prior agricultural training is a major determinant of the recipients of extension services. 6. Rehabilitation of the Ecoles Moyennes Agricoles (EMAs) for vocational and farmer field education on a nationwide scale would increase the demand for extension services, especially among small farmers. 7. Farmers with larger farms receive more agricultural extension services. 8. Coffee producers make more use of extension services than other farmers. 9. Promoting a hybrid system of extension may be more efficient than supporting only public or NGO-provided extension services. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 3 t able of c ontents i . o verview of Agricultural e xtension s ervices in Haiti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 b ackground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 i nstitutional s tructure of Agricultural e xtension s ervices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 ii . Data and s ummary s tatistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 iii . Analysis of Potential Determinants of Agricultural e xtension . . . . . . . . . . . . . . . . . . . . . . . . . 9 g ender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 e ducation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Agricultural Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Farm s ize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Type of Crop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 i V. Conclusions and r ecommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 r eferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Annexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 4 i . b ackgroun D The Haitian population is among the poorest in the world, with over 78 percent living on less than us $2 a day and over 50 percent living on less than us $1 a day. i n rural areas, 88 percent of individuals live below the poverty line and basic services are practically nonexistent. The devas- tating January 12, 2010 earthquake was a major setback to the economy and aggravated an already precari- ous social situation. r elaunching agricultural production is among the Haitian g overnment’s top priorities of the country’s reconstruction program. The transfer of knowl- edge, technologies, and practices through agricultural extension services is a critical building block to raising ag- ricultural productivity and production in an environment dominated by very small farmers. This paper takes stock of the different uses of extension services in Haiti during the 2008−2010 period and aims to provide some historical lessons as a tool for investing most effectively in agricul- tural extension services in a post-earthquake era. The concept of extension services has changed over time. While technological transfer is still important, more emphasis is being placed on expanding the skills and knowledge of farmers (i.e., human capital development), enhancing rural livelihoods, achieving food security, and creating more efficient farmer-based organizations ( s wanson, 2008). Christoplos et al. (2010) defines extension as “all the different activities that provide the information and advisory services that are needed and demanded by farmers and other actors in agrifood systems and rural development.” i t also includes, for instance, “facilitation, brokering and coaching of different actors to improve market access.” According to Christoplos et al. (2012), agricultural extension services can be classified primarily into three areas: • Technology and information sharing • Advice related to farm, organizational, and business management • Facilitation and brokerage in rural development value chains. The most recent Agricultural Census in Haiti, conducted by the Ministry of Agriculture, n atural r esources, and r ural Development (MA rn D r ) during the 2008–2010 pe- riod, classified extension services in the following nine categories: (i) advisory services related to seed/crop se- lection, (ii) arboriculture techniques, (iii) soil preparation and conditioning, (iv) livestock, (v) aviculture, (vi) api- culture, (vii) aquaculture, (viii) post-harvest techniques, and (ix) commercialization. u sing the aforementioned classification, categories (i) to (viii) transferred informa- tion and knowledge to farmers and provided them with guidance on farm management skills, while category (ix) may have given farmers business management skills and facilitated their linkage to value chains and mar- kets. i nstitutional s tructure of a gricultural e xtension s ervices The MA rn D r is responsible for the provision of exten- sion services (through the organic law of s eptem- ber 30, 1987), and is divided into several decentralized structures: 10 Departmental Agriculture Directorates o verview of a gricultural e xtension s ervices in h aiti Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 5 ( Direction Départementale d’Agriculture, DDA ), four sub-Departmental Directorates, and several Agriculture b ureaus ( Bureaux Agricoles ) located in 30 municipalities (among 135 in the country). i n addition, about 15 re- search and training centers are located throughout the country and are directly linked to central services (mainly r &D) in the MA rn D r . These institutions contribute to the provision of various services for plant production, animal husbandry, and natural resource management. While the MA rn D r and its sub-branches fund the provision of vari- ous services for plant production, animal husbandry, and natural resource management and steer and control the regulation of the agricultural sector, the provision of ser- vices and the implementation of investments are gener- ally handled by ngo s, producer organizations, or private entities. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 6 ii . e xtension service coverage in Latin America and the Caribbean varies widely across countries. The oe CD (2011) points out that in Mexico, 3 percent to 10 percent of agricultural units are provided with technical as- sistance, whereas in Chile, the i nstitute of Agricultural Development delivered technical assistance and credit programs to 42 percent of small farmers in 2006. i n n ica- ragua, a country with poverty levels comparable to those of Haiti, the n icaraguan i nstitute of Agricultural Technol- ogy ( in TA) serves about 20 percent of all farm families, according to the 2001 Agricultural Census. o wing to the limited availability of data, this study con- siders three out of 10 departments in Haiti: s outh e ast, Center, and s outh. i n these departments, 13.9 percent of household heads reported having received at least one of the nine aforementioned extension services. g raph 1 shows the relative importance of each type of extension service out of the total delivered in Haiti. The services most frequently delivered are those related to the first stages of the value chain (production), namely choice of seeds and varieties and agricultural techniques and practices, which account for over 50 percent of all services delivered. e xtension services for livestock (cattle and poultry) account for another 42 percent of services received while post-harvest services (storage, processing, and marketing) account for only 6 percent of services delivered. Data an D s ummar Y s tatistics g raph 1: tY pe an D composition of extension services receive D b Y farmers Crop Election (14%) Conditioning, Storage, and Transformation (3%) Arboriculture Techniques (17%) Field Techniques (20%) Livestock (22%) Aviculture (20%) Apiculture (1%) Aquaculture (0%) s ource: Agricultural Census 2008–2010. Authors’ calculations. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 7 Tables 2.1 and 2.2 display information on both the number of households that received extension services and the number of households that reported that they needed extension services in the s outh e ast and Center de- partments. For instance, in the s outh e ast department, 13.51 percent of heads of household received extension services, while only 11.54 percent reported they needed them. i n the Center department, 14.16 percent of house- holds received extension services, while 11.13 percent reported they needed extension. i t appears that in these departments, demand for extension services is fully met. Therefore, this analysis addresses both how the determi- nants of the receipt of extension services proposed in this paper interact not only with the supply (i.e., why these farmers have less or more access to extension services), but also with the demand (i.e., why these farmers think they do not need extension services). i n fact, everyone who needs extension services in these departments seems to have access to them. Furthermore, reporting that the services are needed does not ensure a marginal private benefit of extension (i.e., the demand) since there are transaction costs involved in requesting and partici- pating in the service. Hence, the demand for extension services may be even lower than that reflected in the census. b y contrast, data collected from the s outh department and displayed in Table 2.3 tell a different story. i n that department, 98.79 percent of household heads reported that they needed extension services, while only 13.79 per- cent received at least one service. While many explana- tions can be entertained, a mechanical explanation should not be discarded. The census in Haiti was carried t able 2.1: Deman D for e xtension s ervices – s outh e ast Need Not Need Total N % N % N % Received 9,969 11.54 1,705 1.97 11,674 13.51 Not received 0 0.00 74,735 86.49 74,735 86.49 Total 9,969 11.54 76,440 88.46 86,409 100.00 s ource: Agricultural Census 2008−2010. Authors’ calculations. t able 2.2: Deman D for e xtension s ervices – c enter Need Not Need Total N % N % N % Received 13,764 11.06 3,859 3.10 17,623 14.16 Not received 91 0.07 106,757 85.77 106,848 85.84 Total 13,855 11.13 110,616 88.87 124,471 100.00 s ource: Agricultural Census 2008−2010. Authors’ calculations. t able 2.3: Deman D for e xtension s ervices – s outh Need Not Need Total N % N % N % Received 12,490 13.71 73 0.08 12,563 13.79 Not received 77,528 85.08 1,031 1.13 78,559 86.21 Total 90,018 98.79 1,104 1.21 91,122 100.00 s ource: Agricultural Census 2008−2010. Authors’ calculations. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 8 out over a period of three years (2008−2010), which means that some households were surveyed after the earthquake of 2010. i f some places were systematically surveyed after the earthquake (for example, the s outh department), the tremendous shock caused by the disas- ter could explain these differences. However, this unex- plained difference in demand for extension services in the s outh does not alter the econometric results of this paper as our dependent variable is receipt of and not demand for extension services. Map 1 shows the percentage of household heads by com- mune that have received some extension services. Com- munes are classified into three distinct groups according to the terciles of the distribution in which they fall—less than 5.64 percent, between 5.65 percent and 14.37 per- cent, and 14.38 percent and over. i n Map 1 we observe that there are pockets of low and high receipt of extension services. These pockets could be influenced by factors such as irrigation, geography, past interventions, political configurations, or distance to the closest DDA. m ap 1: a gricultural extension services at the commune leve l s ource: Agricultural Census 2008–2010. Authors’ preparation. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 9 iii . The previous section highlighted overall trends in agri- cultural extension services in Haiti, concluding that there are vast differences across communes. However, there might also be differences within communes. i ndeed, by exploiting the variations within them, we are able to study the relationship between farmer-level character- istics and the likelihood of receiving extension services. We are specifically interested in assessing the correlation between extension services and the following farmer- specific variables: gender of head of household, educa- tion level, agricultural training, farm size, and type of crop produced. To better isolate the importance of each of these variables in predicting which farmers are more likely to receive extension services, we take into account the effect of all unobserved commune-specific variables that may be affecting both the variables under study and the receipt of extension services, particularly the distance to the nearest DDA, geography, irrigation, and political structures. i n order to do this, we introduce “Commune Fixed e ffects” into our probit model. The purpose of this ex- ercise is not to find the causal effects, but the conditional correlations between the variables under examination and the likelihood of receiving extension services. We de- fine the following econometric specification: Pr( Y ij = 1) = G ( β ’ X ij + δ j + e ij ) The equation above describes a fixed effects probit mod- el, where Y ij = 1 if the household head receives at least one type of extension service and is 0 otherwise; G is the normal cumulative density function; β is a row vector with all the coefficients of the variables under study; X ij is a col- umn vector with all the farmer characteristics under analysis such as gender, education (a dummy variable for each level), agricultural training (a dummy variable for each level), farm size (a dummy variable for each size range), crop type (a dummy variable for each type of crop considered), and interactions of each of these variables with gender. The reason we include gender interactions is to assess the effect of each of these vari- ables conditioned on the gender of the household head. Finally, δ j is the commune-specific fixed effect term, and e ij is the idiosyncratic error term. We run the regression using data pooled from all the departments under study and also for each department separately ( s outh e ast, Cen- ter, and s outh). We used clustered standard errors at the district level ( section communale ). Table 3.1 shows the results of the regression. The coeffi- cients for the commune dummies are not presented in the tables; however in all cases, they are jointly significant at the 0.05 level. Therefore, as discussed previously, loca- tion is quite important in determining the level of exten- sion services, and it is necessary to further investigate commune-specific variables causing these pockets of low reception of extension services. For instance, as already mentioned, it may be that the distribution of DDAs is un- equal across communes. e ven if the majority of extension services are provided by ngo s or private entities, distance to the nearest DDA may still have an effect if ngo s and development projects are located near DDAs or Bureaux Agricoles Communales ( b ACs). This may be the case for two reasons: (i) When targeting beneficiaries, ngo s may follow the advice of DDAs, which may tend to favor people located nearby, and (ii) DDAs may implement de- velopment projects or co-manage projects with ngo s. a nal Y sis of p otential Determinants of a gricultural e xtension Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 10 t able 3.1: r egression r esults Determinants of usage of agricultural extension services Dependent variable: use of extension services = 1 if use at least one extension service, 0 otherwise Regressor All South East Center South Female 0.0135 –0.08 0.123 0.0008 (0.0439) (0.0495) (0.0803) (0.0498) Education None omitted omitted omitted omitted — — — — Literate 0.1927*** 0.0639 0.1433 0.369*** (0.0738) (0.0627) (0.1242) (0.1235) Elementary 0.0183 –0.0992 –0.022 0.1526 (0.0787) (0.1413) (0.1603) (0.1164) High School 0.0752 0.0245 0.0661 0.1525 (0.0624) (0.0972) (0.1381) (0.1024) Professional 0.1115 0.0173 –0.5746*** 0.4518*** (0.1236) (0.2078) (0.1037) (0.1143) University 0.1231* 0.2117* –0.0106 0.2026** (0.0672) (0.1272) (0.2003) (0.0895) Agricultural Training Empirical omitted omitted omitted omitted — — — — Occasional 0.9046*** 0.6633*** 0.9499*** 1.0046*** (0.1032) (0.1804) (0.0921) (0.1673) Technical 0.9257*** 1.3219*** 0.5434*** 0.881*** (0.1599) (0.3451) (0.0709) (0.1784) University 0.0113 0.2258 0.0863 –0.1214 (0.1275) (0.2245) (0.1532) (0.18) Farm Size (hectares) Less than 0.15 omitted omitted omitted omitted — — — — From 0.15 to 0.3 0.1187 0.0012 0.1019 0.2254*** (0.0758) (0.1057) (0.2272) (0.0769) From 0.3 to 0.6 0.2525*** 0.1416 0.4005 0.3354*** (0.0844) (0.1134) (0.2504) (0.1006) From 0.6 to 1.2 0.2359** 0.0372 0.5151 0.3005*** (0.1012) (0.1501) (0.2777) (0.1018) From 1.2 to 2.4 0.1519 0.0166 0.3559 0.2694*** (0.1149) (0.1699) (0.3343) (0.0986) More than 2.4 0.1372 –0.0823 0.3664 0.2688* (0.1284) (0.1715) (0.3075) (0.1561) Crops Maize 0.1653 –0.0075 –0.1178* 0.2729 (0.1085) (0.1415) (0.0656) (0.1829) ( continued on next page) Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 11 g en D er Women play an important role in Haitian agriculture. o ne fourth of heads of household are women in the s outh and Center departments, and in the s outh e ast department, the proportion is even larger (34 percent). Moreover, a recent survey conducted by the Conseil National de Sécurité Alimentaire (2011) indicates that the proportion of female-headed households (pooling data from the s outh e ast, Center, and s outh) is 45 per- cent. According to Lastarria-Cornhiel (2006), the pro- portion of rural female-headed households for the late 1990s across 13 countries in Latin America reached nearly 23 percent (Lastarria-Cornhiel, 2006). Hence, it can be argued that the proportion of female-headed house- holds in Haiti is higher than the regional average. This is in accordance with s aito and s purling’s (1992) argu- ment that it is increasingly common for women to man- age or operate farms on a daily basis in all parts of the world, as men leave farms in search of paid employment. i t is important, therefore, to examine if there are any systematic differences between men and women in terms of their receipt of extension services. As illustrated by g raph 2, there is no systematic trend regarding the degree to which male- or female-headed households receive extension services. Moreover, in the Center department, a larger proportion of female- headed households received extension services com- pared to male-headed households. n evertheless, these results may be hiding other variables correlated to both the gender of the head of household and the likelihood of receiving extension services, introducing a bias in the interpretation of the unconditional relationship between gender and receipt of extension services. For instance, being a female-headed household can be correlated with farm size. i f female-headed households had larger farms on average, and larger farms tended to receive more extension services, they would likely receive equal or more extension services than men, not because of their gender, but because of the size of their farms. n evertheless, on average, female-headed households have smaller farms than men. Table 3.2 tells us that for the three departments analyzed in the data, female- headed farms are much smaller than male-headed ones. For instance, in the Center department, which seems to be the area where farmers have the biggest farms, the size of male-headed farms is, on average, 1.33 hectares, while the size of female-headed farms is 1.12 hectares. The differences are fairly similar in the s outh e ast and the s outh departments and even larger in the Center depart- ment when we look at the median values of farm size. i n Tables A.1, A.2, and A.3, we examine the proportion t able 3.1: r egression r esults Determinants of usage of agricultural extension services Dependent variable: use of extension services = 1 if use at least one extension service, 0 otherwise Regressor All South East Center South Beans –0.0573 0.0282 –0.1616 0.025 (0.1108) (0.117) (0.2752) (0.1774) Bananas 0.02 –0.0453 0.2337* –0.2122** (0.0814) (0.1373) (0.1223) (0.084) Coffee 0.2728** 0.445*** 0.0166 –0.0379 (0.1204) (0.1633) (0.118) (0.1404) Mangoes –0.0285 –0.184 0.0719 0.0802 (0.0797) (0.1198) (0.1411) (0.1283) Intercept –1.6973*** –1.7464*** –2.3228*** –1.9243*** (0.3777) (0.2301) (0.4424) (0.4018) Observations 300100 85763 123814 90523 Pseudo R2 0.1817 0.0757 0.2997 0.1527 s ource: Authors. * p-value < 0.1, ** p-value < 0.05, *** p-value < 0.01. ( continued) Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 12 of female-headed and male-headed households by de- partment for each bracket of plot size (not farm size). We observe clearly that as plot size increases, the propor- tion of female-headed households decreases, except for the last bracket size in the Center and s outh, where the proportion of female-headed households slightly increas- es in comparison to the previous bracket. We further examine if the underlying features present in each commune that are affecting receipt of extension services interact differently with male-headed and fe- male-headed households. Maps 2 and 3 demonstrate the level of extension services reception across communes for each type of household according to the gender of its head. The maps are fairly similar, yet there are im- portant differences in relation to Map 1. i n almost every commune, the proportion of households who received extension services is lower than the commune average if the head of household is female. i nterestingly though, when the average rate of extension reception is high, female-headed households receive more extension ser- vices than male-headed households. For instance, in the Cerca La s ource commune in the Center department, the average rate of extension reception is 53 percent, yet for female-headed households it is 64 percent. i t seems that when the supply of extension services is scarce, men are favored over women; when supply is fairly high, the supply of extension services may be the same for both male-headed and female-headed households, thus the quantity of services allocated is solely demand-driven. i n other words, when extension is widely available, receipt of extension services may depend primarily on the de- mand for extension services in both female-headed and male-headed households, which appears to be higher for female-headed households. This observation has important implications for the interpretation of equilibrium between the supply of and demand for extension services—the rather small differences between men and women in terms of their receipt of extension services may be explained by a higher demand for extension services by female- headed households, and perhaps less access. Hence, the equilibrium would misleadingly appear to be the same for male-headed and female-headed households. g raph 2: h ousehol D s receiving extension services b Y gen D er of househol D hea D South East 16 Male Female 14 14.08 12.38 14.03 14.99 14.1 12.82 12 10 8 6 4 2 0 Center South Percentage (%) Department s ource: Agricultural Census 2008−2010. Authors’ calculations. t able 3.2: a verage (me D ian) farm size in hectares b Y gen D er of househol D hea D Gender South East Center South Male 0.91 (0.65) 1.33 (0.97) 0.94 (0.65) Female 0.68 (0.48) 1.12 (0.81) 0.68 (0.48) Percentage Difference 34% (35%) 12% (20%) 38% (35%) s ource: Agricultural Census 2008–2010. Authors’ calcula- tions. Farm size is calculated at the household level (where each can have more than one plot), whereas Tables A1, A2, and A3 are calculated at the plot level. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 13 m ap 2: a gricultural extension services at the commune level – m ale-hea D e D househol D s s ource: Authors. m ap 3: a gricultural extension services at the commune level – f emale-hea D e D househol D s s ource: Authors. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 14 According to our econometric model, gender itself is not important in explaining supply and demand equilibrium levels of extension services. The fact that gender is not significant when controlling for these covariates and location means that the initial rather small differences in extension reception observed in g raph 2 were not the result of underlying differences in education, agricultural training, farm size, type of crop produced, and location between male-headed and female-headed households. We also ran two separate regressions (results not shown): one only for farmers located at Cerca La s ource (a loca- tion with a high level of extension) in the Center depart- ment and the other for those farmers located at s t. Louis Du s ud (a location with a low level of extension) in the s outh department. i n the case of Cerca La s ource, the coefficient on the female dummy is positive and signifi- cant at the 0.1 level. i n that commune, a female-headed household has an 11.48 percent greater chance of receiv- ing extension services than a male-headed household controlling for education, agricultural training, farm size, and type of crop. i n s t. Louis Du s ud, the coefficient on the female dummy is not significant. Therefore, the relationship between the gender of the head of house- hold and the receipt of extension services, if any, may favor women. i n those places with a high overall availabil- ity of extension services, women receive systematically more extension services than men. i n those places with an overall low availability of extension services, there are no significant differences between men and women after controlling for other covariates in the model. r ecall that we are observing the equilibrium of demand for and supply of extension services, which means that even when female-headed and male-headed house- holds receive the same level of extension services (pro- vided they have the same education level, agricultural training, farm size, and produce the same type of crop), the interaction between supply and demand by which they receive the same services can be different. For instance, extension services in a particular commune may be provided primarily to male-headed households, yet the demand from female-headed households could be significantly higher than that from male-headed ones, resulting in the receipt of the same number of extension services. s ometimes, in this context, a female-headed household may receive a lower amount, as we previously observed in Maps 2 and 3, in locations where overall ac- cess is low. i f we assume that the aforementioned house- holds were being offered the same amount of extension services, we may conclude that no further interventions are necessary to correct the tendency to favor men, when in reality, discrimination may be latent—factors such as the time of the day services are offered, night travel, and long distances, among others, have been documented in Haiti as issues that prevent women from accessing services. eD ucation Haiti faces challenges of both supply and demand in the education marketplace. These challenges are compounded in rural areas by high poverty and difficult access. o n the supply side, there are simply not enough spaces for children to enroll in school. i t is estimated that 400,000 to 500,000 children aged 6 to12, the major- ity of whom live in rural areas, are not attending school. o n the demand side, the average cost of us $70 tuition per child per year is prohibitive for poor families, especially for those living in rural areas characterized by poverty rates of 82 percent (77 percent living in extreme poverty). 1 e ven when schools are accessible, the quality of the educa- tion offered is uneven, and often very low. This is demon- strated by the findings of the recent e arly g rade r eading Assessment ( egr A), carried out in 2008 and 2009 in Haiti. o n average, children in g rade 3 are able to read fewer than 23 words per minute. 2 For those students studying in Creole, 29 percent were unable to read a single word by g rade 3. r eading comprehension is even weaker, with children able to answer less than 10 percent and 17 per- cent of reading comprehension questions correctly, in French and Creole respectively. 3 o pportunities to improve small farmers’ competitiveness are reduced as extremely poor levels of education ham- per the implementation of new productivity-enhancing agricultural technologies. According to the Agricultural Census (see Table 3.3), 57.09 percent of heads of house- hold are illiterate. i f we further discriminate by gender, the 1 The World b ank. e ducation for All Project – Phase ii (APL). o ctober, 2011. 2 s ixty words per minute is standard for early primary reading fluency. 3 r esearch Triangle i nstitute. Haiti e arly g rade r eading Assessment ( egr A): r apport pour le M en FP et la b anque Mondiale. Avril 2010. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 15 level of illiteracy in female heads of household reach- es 65.88 percent. i n Table 3.3, we clearly observe how the proportion of male-headed households increases as the level of education increases, indicating that women are less favored than men in terms of education. For example, the proportion of female-headed households in the three departments analyzed is 26.83 percent; however, among those heads of households with university-level education, the proportion of female-headed households is only 12.57 percent. g raph 3 provides useful insights that may clarify the mechanisms through which receipt of extension ser- vices is influenced by education. Moving from being an illiterate to a literate farmer seems to have a positive effect on receiving extension services for all departments, presumably, as a result of required reading material. How- ever, even if the ability to read is not necessary to receive extension services, literate people are more likely not only to be aware of the benefits of receiving agricultural extension services, but also to understand the procedures for receiving extension services and how to implement what they learn or what they receive as inputs for their farms. b erger et al. (1984) points out that “education enhances the ability of farmers to acquire accurate information, evaluate new production processes, and use new agricultural inputs and practices efficiently. b etter educated farmers are twice as likely to be in contact with t able 3.3: eD ucation b Y gen D er of hea D of househol D Male-Female Ratios Total Male Female Male Female Education N % N % N % % % None 172,011 57.09 118,758 53.86 53,253 65.88 69.04 30.96 Literate 60,865 20.20 47,137 21.38 13,728 16.98 77.45 22.55 Elementary 46,386 15.39 36,490 16.55 9,896 12.24 78.67 21.33 High School 18,866 6.26 15,366 6.97 3,500 4.33 81.45 18.55 Professional 2,305 0.76 1,960 0.89 345 0.43 85.03 14.97 University 875 0.29 765 0.35 110 0.14 87.43 12.57 Total 301,308 100.00 220,476 100.00 80,832 100.00 73.17 26.83 s ource: Agricultural Census 2008–2010. Authors’ calculations. g raph 3: h ousehol D s receiving extension services b Y e D ucation level s ource: Agricultural Census 2008–2010. Authors’ calculations. None Literacy Elementary South East Center 0 High School Professional University Percentage (%) Education Level South 5 10 15 20 25 30 Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 16 agricultural extension agents, indicating that farmers with higher levels of education benefit most from extension services.” i n addition, “educated farmers may push the extension system to deliver what they need and make sure the knowledge is appropriate to their resources.” n evertheless, elementary schooling appears to have a negative effect on extension reception relative to mere literacy. i t is important to note that as people become more educated, they acquire skills that can be better re- warded in non-farm activities. Hence, the more educated a person is beyond literacy, the lower their demand for agricultural extension services may be. However, if wage jobs are scarce, or the opportunity costs related to leav- ing their farms are fairly high, then we would presumably see an increasing relationship between education and the receipt of extension services, as seems to be the case in the s outh e ast department. When controlling for other factors, the positive effect of being literate is smaller than that observed in g raph 3. i n particular, literacy increases the likelihood of receiving extension services by only 3.44 percent. This trend is mainly driven by the s outh department, where literacy increases this likelihood by 7.43 percent. i n the other departments, the effect is not even statistically significant. Moreover, in the Center department, having professional education decreases the likelihood of receiving extension services, whereas, in the s outh, it increases the likelihood by 10.23 percent, and university-level education increases the like- lihood of receiving extension services by 2.21 percent. o n the supply side, there is the possibility that for low levels of education, access to extension services is still extremely low as extension agents may prefer to provide extension services to more educated farmers where the possibility of implementing newly acquired knowledge is higher. At the same time, on the demand side, farmers with more education are less prone to demand extension services as they are able to learn and apply new tech- nologies or knowledge by themselves, what we could call the “knowledge effect.” Furthermore, the possibility of looking for non-farm jobs is higher for those with bet- ter education. Therefore, on the one hand, there may be three education-based forces affecting the demand for extension: the “awareness effect,” the possibility of finding a non-farm wage job that is economically more convenient than the farmer’s agriculture-related activity, and the “knowledge effect.” o n the other hand, there is one clear education-based force affecting the supply of extension: the eagerness of extension agents to pro- vide services to more educated farmers. The dynamics of these forces may explain the different levels of exten- sion services received depending on a farmer’s level of education. For instance, at first glance, it might seem strange that the positive effect of education fades be- yond mere literacy and then returns after university-level education. However, if we assume both that extension agents tend to favor educated farmers and that demand for extension services is lower for higher levels of educa- tion, this result can be reasonable. i t can also be argued that the demand for extension services can even in- crease at high levels of education as farm owners might hire farm workers that receive extension services. These forces can also explain the apparent heterogeneity that we observe across departments. For example, in the s outh department, education has a more consistently posi- tive effect overall on receipt of extension services com- pared to in other departments. Presumably, in the s outh department, the opportunity costs of leaving agriculture as a main activity are higher than in other departments. Furthermore, as we observe in Table A.4 (see Annexes), the s outh department has more farmers reporting livestock and fisheries as their main economic activities. These activi- ties may be more difficult to leave behind, which means that they may be more profitable than agriculture. a gricultural t raining As demonstrated by g raph 4, there seems to be an in- verted u-shaped relationship between agricultural training and the receipt of extension services. e ven after controlling for other covariates, the results confirm the concavity and show that having “occasional agricultural training” ( o AT) increases the likelihood of receiving extension services by 23.98 percent compared to having just empirical train- ing. Furthermore, having technical agricultural training increases the likelihood by 25.12 percent, which means that the positive effect of agricultural training is decreas- ing. Apart from the “awareness effect” and the “knowledge effect,” which were also discussed in the case of education (and which may be even more pronounced in this case), receiving o AT from specialized agencies, such as a DDA or ngo s, may create an enabling environment for farmers, putting forward adequate channel factors for both farmers Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 17 demanding extension services and extension providers supplying the services. Furthermore, extension agents may naturally target farmers with high agricultural training since adoption of new technologies and knowledge received is more likely and thus their work can be properly measured and rewarded. i t is also important to note that the positive effect of having o AT in terms of the receipt of extension services diminishes significantly when the head of house- hold is a woman (see Table A.6 in Annexes) since women may benefit less from the opportunities brought about by the channel factors mentioned above. o ther possible explanations are that the “knowledge effect” may be more pronounced in the case of female-headed households, or the supply of extension services to female-headed households may be low even when they have high-level agricultural training. The positive impact of agricultural training on the uptake of extension services starts to sink in at the technical level. i t is possible that within agricultural training, the “knowl- edge effect” discussed previously is dominating the dynamics of receiving extension services. i n other words, people with technical agricultural training might per- ceive the benefits of receiving extension services as mini- mal, or even non-existent. For example, the FA o found that 40 percent of extension personnel used in developing countries had only secondary school education (Feder et al., 1999). Hence, not surprisingly, uptake of extension services is significantly diminished as people get more informed and knowledgeable about agricultural topics— and may even be more knowledgeable than extension facilitators themselves. o ne public sector supply of o AT is the Ecoles Moyennes Agricoles ( e MAs) for Vocational and Farmer Field e duca- tion on a nationwide scale. Having the proper channels through which extension services are delivered not only increases the supply of extension, but also stimulates the demand for these services. The e MAs are well-known agricultural training institutions supported by the World b ank, Canada, us A i D/ us DA, and other development organizations working in Haiti. The MA rn D r is seeking to leverage and strengthen the e MAs as part of the na- tional strategic plan (PDVA) to expand extension services in Haiti. s ome might reasonably argue that occasional agricul- tural training is so statistically significant in explaining ag- ricultural extension services because o AT and extension services are being perceived by the farmers interviewed as being the same thing. However, if this is true, then the correlation between receiving extension services and having occasional training should be nearly one. i n order to assess the possibility that o AT and extension services might be perceived as being the same, we present Table 3.4, which shows the relationship between o AT and receipt of extension services, based on data pooled from the three departments. g raph 4: p roportion of househol D s receiving extension services b Y agricultural training Empirical 60 South East Center 50 40 30 20 10 0 Occasional Technical University Percentage (%) Agricultural Training South s ource: Agricultural Census 2008−2010. Authors’ calculations. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 18 According to Table 3.4, among those who did not re- ceived o AT, the ratio between those who received exten- sion services and those who did not is 0.15 (=12.97/84.59), yet within those who did receive o AT, the ratio is 0.62. Therefore, there is a positive correlation between o AT and extension services, however the correlation is rather low (0.11). Moreover, there is a significant proportion of the population who did not receive o AT and who did receive extension services. Hence, we cannot conclude that o AT and extension services are exactly overlapping events. f arm s ize g raph 5 plots the relationship between receiving exten- sion services and farm size. b oth empirical and theoreti- cal studies suggest that farmers with larger farms adopt extension services more quickly (Fischer, 1985); thus we would expect a greater use of extension services in larger farms. As we can see in g raph 5, the relationship between farm size and receipt of extension services is in- deed positive, yet the relationship is concave, meaning that the rate of receipt of extension services decreases as farm size increases. Moreover, for the s outh and s outh e ast departments, these curves correlate very well with those of the previous graph. i t seems that the marginal benefits of implementing extension services might be con- stantly reducing as farm size increases, ultimately affect- ing the demand for extension services. Feder (1999) shows that the effectiveness of extension investment is highly contingent on relaxing wider barriers to the successful development of the agricultural sector as a whole, includ- ing such potentially limiting factors as credit, technology stock, input supplies, price incentives, institutions, and hu- man resource constraints. Therefore, it may be reasonable to argue that extension services in Haiti are not highly ef- fective, and so the demand for these services is rather low for farmers with access to other alternatives for acquiring knowledge (such as fee-based extension). However, again, we are observing the equilibrium between the supply of and demand for extension services. These preliminary results may be explained not only by issues of the marginal benefit of implementing extension advice, but also by issues related to the marginal propensity to offer extension advice. i n other words, it is possible that the sup- ply of extension services is more targeted to smaller farms. n evertheless, the literature suggests that the opposite is true. Feder et al. (1999) stresses that there is a tendency of extension agents to favor more responsive clients, who are typically better endowed and more capable of under- taking risks. Consequently, this reinforces the possibility that the concavity of the relationship between farm size and receiving extension services is better explained by a low de- mand for extension services from farmers with larger farms. The concave relationship described above between receiving extension services and farm size is somewhat supported by the results of the regression; yet if we dis- criminate by department, we observe that only in the s outh is the relationship significant. Having a farm of be- tween 0.3 and 0.6 hectares increases the likelihood of receiving extension services by 4.57 percent compared to having a farm of less than 0.15 hectares; however, hav- ing a farm of between 0.6 and 1.2 hectares decreases the probability of receiving extension services by 0.44 percent in relation to the previous size bracket. i n the s outh, the concavity is even more pronounced. Therefore, larger farms either received proportionally (to size) fewer exten- sion services or received fewer extension services in abso- lute terms. Taking into account the tendency of extension agents to favor more responsive clients, who are typically better endowed and more capable of undertaking risks t able 3.4: o ccasional a gricultural t raining ( oat ) an D extension services OAT Extension Received Not Received Total N % N % N % Received 2,819 0.93 4,562 1.51 7,381 2.44 Not received 39,160 12.97 255,461 84.59 294,621 97.56 Total 41,979 13.9 260,023 86.1 302,002 100 s ource: Authors. Determinants of a gricultural e xtension s ervices: t he c ase of h ait i 19 (Feder et al., 1999), this result may be driven by low de- mand rather than by a lack of adequate supply. i n summary, up to a certain farm size, the receipt of ex- tension services increases as farm size increases, possibly because of a greater supply for larger farmers, but also because of economies of scale, making the implemen- tation of new technologies more feasible, which in turn increases the demand for extension services. However, beyond that point, it is likely that demand for extension services decreases as farmers with larger farms have more leverage to acquire new knowledge from more efficient sources (such as fee-based extension). i t is also important to note that for large farms, the effect on the likelihood of receiving extension services is not significant; however, for female-headed households it is significant and positive. Assuming that the supply of extension services is not higher for female-headed households, a feasible explanation for this result may be that women are generally more risk averse (see for example e ckel and g rossman (2008)) and prefer not to in-