![Motivation ,[object Object],[object Object],[object Object],[object Object],[object Object],[object Object]](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
Recommended
More Related Content
Viewers also liked
Viewers also liked (16)
Similar to The Brokerage Institution & The Development of Agricultural Markets: New Evidence from Ethiopia
Similar to The Brokerage Institution & The Development of Agricultural Markets: New Evidence from Ethiopia (20)
More from essp2
More from essp2 (20)
The Brokerage Institution & The Development of Agricultural Markets: New Evidence from Ethiopia
- 6. Surveys considered Traders' Surveys Survey 1996 E. Gabre-Madhin dissertation survey ("Market Institutions, Transaction Costs, and Social Capital in the Ethiopian Grain Market") Number of respondents 169 Number of markets 13 Survey 2002 IFPRI and ILRI “Survey of grain and coffee traders” Number of respondents 561 Number of markets 45 Survey EDRI and IFPRI “ECX Trader survey 2007 ” Number of respondents 457 Number of markets 21
- 23. Marginal Effects for the Heckman selection model Conditional Marginal Effects (for traders already using brokers) Unconditional Marginal Effects (for the overall sample of traders) Continuous variables : McDonald and Moffitt decomposition(1980) Discrete variables : Cong (2000)
- 24. Data & Methodology “ ECX Trader Survey 2007” (Gabre-Madhin, EDRI and IFPRI): 457 wholesalers in 21 markets recalled their activity from October/November 2006 to April/May 2007
- 27. Heckman model for Buyers: Regression Results Variables Outcome equation Selection equation Conditional Marginal Effects Marginal Effect for Coeff. Coeff. ASSETS Social Capital No. of Trading Contacts in the Main Market -0.109 (0.066) -0.039 * (0.024) No. of Regular Suppliers -0.087 *** (0.025) 0.101 (0.062) -0.096 *** (0.024) 0.036 (0.022) Human Capital No. of Employees Engaged in Search 0.684 *** (0.161) 0.246 *** (0.057) No. of Trader’s Substitutes 0.400 ** (0.171) 0.144 ** (0.062) Financial Assets & Access to Credit Working Capital 0.043 ** (0.017) -0.060 (0.065) 0.048 *** (0.018) -0.021 (0.023) Credit Access -0.106 ** (0.053) -0.270 * (0.151) -0.082 (0.050) -0.096 * (0.053) CONTRACTUAL PERFORMANCE: Costs Annualised Physical Marketing Costs -0.018 (0.043) -0.006 (0.016) Fixed/Operational Costs 0.205 *** (0.062) 0.073 *** (0.022) TRADING PRACTICES Distance from the Base to the Main Market 0.102 ** (0.043) 0.117 (0.149) 0.092 ** (0.040) 0.042 (0.054) ACCESS TO PHYSICAL INFRASTRUCTURE Storage Capacity -0.018 (0.013) -0.135 *** (0.049) -0.007 (0.013) -0.049 *** (0.018) Asphalted Roads -0.687 *** (0.254) -0.984 (0.886) -0.595 ** (0.243) -0.275 (0.170) Dry-Weather Roads -0.606 *** (0.207) -0.044 (0.883) -0.602 *** (0.195) -0.016 (0.311) All-Weather Roads -0.654 *** (0.240) 0.144 (0.866) -0.667 *** (0.220) 0.053 (0.326)
- 29. Heckman model for Buyers: Regression Results (cont.) As distance between base and main markets increases, buyers’ location matters less in determining the probability of brokerage use
- 30. Heckman model for Buyers: Regression Results (cont.) As distance increases, the predicted probability of using brokers is more than double when roads are all-weather or dry-weather roads than when they are asphalted roads. The worst the condition of the road linking the base to the main markets of each buyer, the more it is likely that buyers turn to brokers.
- 31. Heckman model for Buyers: Regression Results (cont.) Variables Outcome equation Selection equation Conditional Marginal Effects Marginal Effect for Coeff. Coeff. ACCESS TO FINANCIAL INSTITUTIONS Bank in the Main Market MAIN CROP TRADED Cereals Coffee Pulses LOCATION: Agricultural Development Domains Moisture-Reliable Areas Drought-Prone Areas Pastoral Areas Base Market – Low-Market Access & High-Population Density Base Market – High-Market Access & High-Population Density Constant 0.264 *** (0.095) 0.136 (0.128) 0.196 * (0.119) -0.150 * (0.077) 0.074 (0.080) -0.206 (0.193) -0.177 ** (0.078) -0.027 (0.063) 0.272 (0.213) 1.208 *** (0.400) -0.312 (0.246) -0.673 ** (0.297) -0.292 (0.335) 0.832 *** (0.260) 1.665 *** (0.274) -0.033 (0.436) -0.105 (0.219) -0.130 (0.197) -2.759 *** (0.786) 0.291 *** (0.093) 0.198 (0.124) 0.222 * (0.114) -0.222 *** (0.076) -0.055 (0.067) -0.203 (0.188) -0.168 ** (0.077) -0.016 (0.061) 0.310 *** (0.061) -0.114 (0.091) -0.207 *** (0.074) -0.098 (0.104) 0.291 *** (0.086) 0.593 *** (0.079) -0.012 (0.155) -0.037 (0.076) -0.046 (0.070) Number of observations Of which uncensored Log pseudolikelihood Wald test of indep. eqns. (rho = 0)/χ^2 (1) p-value for the Wald test /athrho /lnsigma rho sigma Lambda 449 162 -241.673 5.49 0.019 0.450 ** (0.192) -1.261 *** (0.068) 0.422 0.283 0.120 Ycond=0.703 Psel=0.323
- 32. Heckman model for Sellers: Regression Results Variables Outcome equation Selection equation Conditional Marginal Effects Marginal Effect for Coeff. Coeff. ASSETS Social Capital No. of Regular Customers Human Capital -0.089 *** (0.027) 0.049 (0.059) -0.079 *** (0.026) 0.010 (0.011) No. of Employees Engaged in Search -0.560 *** (0.167) -0.109 *** (0.033) No. of Years of Operation -0.292 (0.224) 1.156 ** (0.483) -0.056 (0.198) 0.225 ** (0.091) Square of the No. of Years of Operation 0.052 -0.237 ** (0.049) (0.108) 0.004 (0.044) -0.046 ** (0.021) Financial Assets Working Capital 0.173 ** (0.080) 0.034 ** (0.016) CONTRACTUAL PERFORMANCE: Costs & Trading Disputes Annualised Physical Marketing Costs 0.098 * (0.056) 0.019 * (0.011) Fixed/Operational Costs -0.102 (0.079) -0.020 (0.015) Trading Disputes with Customers: No Payment 0.452 ** (0.189) 0.101 ** (0.047) TRADING PRACTICES Distance from the Base to the Main Market 0.043 (0.043) -0.031 (0.062) 0.037 (0.042) -0.006 (0.012)
- 33. Heckman model for Sellers: Regression Results (cont.) Variables Outcome equation Selection equation Conditional Marginal Effects Marginal Effect for Coeff. Coeff. ACCESS TO PHYSICAL INFRASTRUCTURE Storage Capacity 0.042 ** (0.019) -0.070 (0.052) 0.027 (0.017) -0.014 (0.010) Asphalted Roads -0.066 (0.238) 0.128 (0.371) -0.040 (0.229) 0.026 (0.080) Dry- & All-Weather Roads MAIN CROP TRADED Cereals Coffee Pulses LOCATION: Agricultural Development Domains Moisture-Reliable Areas Drought-Prone Areas Pastoral Areas Base Market – Low-Market Access & High-Population Density Base Market – High-Market Access & High-Population Density Constant 0.055 (0.314) 0.222 * (0.133) 0.336 ** (0.169) 0.206 (0.196) -0.454 (0.296) -0.553 * (0.297) -0.171 (0.359) -0.172 (0.113) 0.055 (0.107) 1.370 *** (0.465) -0.081 (0.570) -0.570 * (0.305) -0.678 * (0.383) -0.582 (0.437) 0.960 ** (0.472) 1.234 *** (0.472) 0.722 (0.643) -0.013 (0.245) 0.031 (0.235) -3.981 *** (1.071) 0.038 (0.295) 0.108 (0.109) 0.194 (0.153) 0.084 (0.173) -0.258 (0.262) -0.312 (0.257) -0.031 (0.320) -0.175 * (0.104) 0.062 (0.100) -0.015 (0.101) -0.128 * (0.078) -0.094 ** (0.037) -0.084 * (0.044) 0.183 ** (0.083) 0.321 ** (0.138) 0.196 (0.216) -0.002 (0.047) 0.006 (0.046) Number of observations Of which uncensored Log pseudolikelihood LR test of indep. eqns. (rho = 0)/χ^2 (1) p-value for the Wald test /athrho /lnsigma rho sigma Lambda 414 69 -160.766 5.53 0.019 -0.943 ** (0.400) -1.089 *** (0.187) -0.737 0.337 -0.248 Ycond=0.561 Psel=0.115
- 35. Heckman model results: Decomposition of Unconditional Marginal Effects for Buyers
- 36. Heckman model results: Decomposition of Unconditional Marginal Effects for Sellers
- 40. Thank You!
- 48. Variables used in the Analysis – Exclusion Restrictions/Selection Instruments
- 49. Variables used in the Analysis – Variables in both the Outcome & the Selection Equation
Editor's Notes
- Chamberlin et al. (2006) found that the highest commercialisation rate for cereals in moisture-reliable areas was found for farmers based in districts with lower market access . The authors suggested that “in high potential areas, cereals are likely less profitable than higher value commodities such as vegetables in areas of high market access, but may have a strong comparative advantage in areas of low market access”. Contextually, “In more drought prone highland areas and low rainfall lowland areas , cereals may be the most profitable and/or least risky option for farmers with relatively good market access (unless they have access to irrigation)”
- Information costs A 1 percent surge in search labour (the number of employees that participate in collecting price information , which is an indicator of human capital ) raises the predicted probability that buyers use brokers by 0.246 percentage points (p.p.). Human Capital Positive relationship between the probability of brokerage use and human capital is also indicated by the significant marginal effect for the number of trader’s substitutes (family helpers and permanent workers authorised to buy in the name of the enterprise) whose cost surges as their number grows. On the contrary, brokers are paid with a flat fee that varies across regions but is fixed across time (fixed costs that does not change at the margin after a change in the share of purchases dealt by brokers). Negotiation costs If the number of regular suppliers in the main market ( social capital ) increases by 10 percent, the share of brokered transactions for traders using intermediaries will approximately decrease by 0.01 percentage points. Social capital A 1 percent increase in the number of trading contacts in the main market (another indicator of social capital availability) reduces the predicted probability that buyers use brokers by 0.039 p.p. (significance at the 10 percent level). Financial assets ( working capital indicator for the size of the trading business ) Analysis on the actual/observed quantities (i.e. on the conditional marginal effects) reveals that bigger trading firms use brokers for a greater percentage of their purchases. Access to credit (incentive to investment) Accessing credit and getting the chance to become bigger makes buyers more likely to buy personally (marginal effect for the probability that buyers that have access to credit use brokers is nonetheless significant at only 10 percent level). Contractual performance , as measured by marketing and fixed costs , positively and significantly affects the probability that buyers use brokers only in its operational costs component (ex. rental of shops and storage facilities, maintenance of vehicles, municipality taxes etc.). Attempts to minimise depreciation costs induce buyers to use brokers. Distance The conditional marginal effect of distance on brokered transaction is 0.092, meaning that a 10 percent increase in distance approximately raises the share of brokered purchases by 0.01 percentage points for buyers already using brokers. THIS MAY SEEM MARGINAL, BUT SEE POSTESTIMATION RESULS Access to infrastructure – Roads’ availability (either asphalted or dry-weather or all-weather roads) linking the base to the main markets lowers the share of brokered purchases. Thus, traders that have access to road infrastructure are more able to travel personally for a bigger share of their transactions, and they rely less on the transportation services offered by intermediaries. Access to infrastructure – Capacity of storage facilities (in quintals) under wholesalers’ exclusive control . A 1 percent increase in the storage capacity reduces the predicted probability to use brokers by 0.049 percentage points. A commodity exchange , which formalises the activities conducted by brokers and is supported by warehouses , might benefit in particular smaller wholesalers that cannot afford the costs of a storage facility with adequate capacity under their exclusive control . These traders are consequently more likely to use brokers as well as the services of a grain exchange. Other infrastructure and telecommunications (landline and mobile networks and electricity) were excluded from the model because of the very limited sample variation (i.e. their presence was found in almost all main markets).
- Post-estimation analysis reveals that distance impacts in particular on the decision process of buyers of cereals based in drought-prone areas . As distance increases from 7 to 230 kilometres, an average buyer with these characteristics would be 23.42 percentage points more likely to use brokers than the rest of traders ( significance at the 2 percent level ). Furthermore, after the travelled distance reaches and overcomes 160km, all buyers of cereals based in drought-prone domains would be likely to ask brokers to manage (some or all of) their long-distant transactions (see figure 1).
- For the whole sample of traders and as distance raises , we observe a decrease in the gap between the higher predicted probability that buyers based in drought-prone areas use brokers and the lower predicted probability that buyers located in other areas (i.e. moisture-reliable areas, pastoralist areas and the central market of Addis Ababa) use brokers. Nonetheless, the confidence interval for the estimated probability difference tends to become larger at a distance of around 150 kilometres, which indicates an increase in the uncertainty about the true value of the estimated gap in predicted probabilities.
- Post-estimation analysis shows that, as distance increases, the predicted probability to use brokers is more than double when available roads are all-weather roads than whey they are asphalted roads. The predicted probability for all-weather roads significantly exceeds that of asphalted roads for values of distance between 80 and 660kilometers approximately, as shown by the non-overlapping 95 percent confidence intervals of the two predicted probability lines
- Financial institutions Dummy of whether one or more banks are operating in wholesalers’ main market towns . The predicted probability that traders use brokers increases by 31 percentage points (highly significant at the 1 percent level) if there is a bank in the traders’ main markets. Crops Considering wholesalers at their average characteristics, the share of brokered transactions for buyers of cereals was greater than that of oil seeds traders in 2006/07 harvest year. At the same time, an average coffee trader was 20.7 percentage points less likely to use brokers for his/her purchases than an oil seeds trader (base category). This is most probably because coffee is marketed in Ethiopia through cooperatives that organise transportation from producers to city warehouses (Chamberlin et al., 2006). Agricultural Domains The predicted probability (computed at the mean value of all other explanatory variables) that buyers of any crop use brokers is 0.593 for traders based in drought-prone areas (i.e. these buyers are 59.3 percentage points more likely to use brokers), compared to 0.291 for wholesalers located in moisture-reliable domains. Post-estimation reveals that the predicted 0.593 likelihood would further increase by 16.4 percentage points for traders based in markets with limited access and low population density . If the subsample of buyers of cereals was considered, this increase would be of 23.4 percentage points instead (Confidence intervals for post-estimation are computed using the delta method; only significant results are mentioned). These outcomes suggest the existence of an inverse relationship between the commercialisation rate , which measures how much smallholder farmers participate into marketing their crops, and the probability that traders buying those crops are using brokers . The more farmers sell their crops on the market the more trading opportunities arise, the less it is likely for traders to turn to brokers. In other words, brokers seem to facilitate commercialisation especially where it would be otherwise difficult, given the predominant aspects of smallholder-relevant agricultural domains . Population density An average buyer based in a market with low access and high population density has a share of brokered purchases smaller than a buyer located in a market with low access and low population density (base category). Thus, high population density in the markets where traders are located reduces their need to use brokers . We can infer this is because the larger the population in the base markets, the wider the existing network of business relationships spreading to distant markets.
- Information Costs Sellers seem to be less concerned than buyers about the minimisation of information costs: a 1 percent increase in the number of employees engaged in search (human capital) reduces the predicted probability of using brokers by 0.109 percentage points (the more sellers can afford to employ manpower to engage in price information gathering, the less they are likely to use brokers). Human Capital Another indicator of human capital availability is the number of years of operation of the trading business . This number has a positive marginal effect on the probability that wholesalers use brokers, whereas the marginal effect is negative for the number of years of operation squared. This implies an inverted U-shaped pattern for the effect of this human capital indicator on the probability of brokerage use . In other words, the relationship between years of operation and probability to use brokers is not linear ; new trading businesses are more likely to use brokers and this probability decreases while trading experience is cumulated. The number of years of operation was included in the analysis for sellers instead of the number of trader’s substitutes (an alternative indicator for human capital availability which was found statistically insignificant, and was thus excluded from the final model formulation). Meantime, the number of trading contacts in the main market (proxy for social capital) was excluded from the Heckman model for sellers as associated with a non-significant marginal effect on both the probability to use brokers and the share of brokered transactions. Negotiation costs An increase in the number of regular customers (social capital) decreases the amount of brokered out of total sales for the sellers currently using brokers. Working Capital A 1 percent expansion of the working capital at trader’s disposal raises the probability to engage the brokerage services by 0.034 percentage points. This result somehow contradicts the finding for buyers, according to which the predicted probability to use brokers decreases as credit access becomes available. This is most probably because working capital is specifically used for trading purposes and it is regularly available, while credit ‘is cumulative and can be used for purposes other than buying and selling grain; for instance, it can be applied to the fixed costs of the business.’ (Gabre-Madhin, 2001a: 24). Contractual performance A 1 percent surge in annualised physical marketing costs (mainly transportation costs), increases sellers’ likelihood to turn to brokers by 0.019 percentage points (significance at the 10 percent level). Contractual risks The predicted probability to use brokers is 0.101 for those sellers who experienced, at least once since the start of the production year, lack of payment from their customers (i.e. these sellers are 10.1 percentage points more likely to use brokers). Distance The distance between the base and the main markets, determinant of higher transaction costs, seems to have no effect on sellers’ decision to use brokers when road-quality dummies are included in the model. If road-quality binary variables were excluded from the model, a 10 percent increase in distance would raise the current shares of brokered sales by 0.0036 percentage points (significant at the 5 percent level).
- Physical infrastructure Both the access to asphalted, dry- and all-weather roads and the capacity of storage facility/ies under wholesalers’ exclusive control show insignificant marginal effects. Thus, infrastructure availability and road quality impact on sellers and buyers’ decisions in significantly different ways. Access to financial institutions was not introduced in the sellers’ model specification as it did not vary greatly across respondents (98 percent of sellers reported the presence of a bank in their main markets), while access to formal and informal credit , partly associated with financial institutions’ availability, was excluded as found to have no significant impact on sellers’ decision making process. Traded crops and for a significance level of 5 percent, coffee sellers are less likely to use brokers. Yet, the Heckman estimated coefficient for the outcome equation is positive and significant. Hence, the expected (i.e. potential and desired) shares of sales covered by brokers , for the full sample of sellers, marginally increase when coffee is the traded crop. Smallholder agricultural development domains The predicted probability of brokerage use is 0.321 for sellers located in drought-prone areas and 0.183 for sellers based in moisture-reliable domains (both probabilities are significant at the 5 percent level). Population density Similarly to what was found for buyers, the higher the population density in the base market the lower the share of brokered sales (at the 10 percent significance level).
- Contextually, for the subsample of traders of cereals and pulses, an average seller in a drought-prone area was 19.4 percentage points more likely to use brokers if s/he was based in a highly-accessible and highly-populated market (significance at 10 percent level).
- The decomposition of unconditional marginal effects on actual quantities transacted reveals that: Unconditional marginal effects for buyers are mainly associated with a significant quantity component (the part due to current participants) than a participation component (the part of the total effects due to new participants)
- The decomposition of unconditional marginal effects on actual quantities transacted reveals that: For sellers, the participation component (the part of the total effects due to new participants) is usually significant and larger than the quantity component (the part due to current participants)
- Nekemte and Nazret (in Oromiya region, respectively for the trade to the South-West and to the East and South-East of the country) and Bahir Dar (in Amhara region, for the transactions towards the North). In addition, Bure and Dessie, both in Amhara region, are key transit points from where traders redirect food to the North of the country (USAID and FEWS NET, 2009).