Transaction Costs and Market Institutions: Grain Brokers in Ethiopia
Transaction Costs and Market Institutions: Grain Brokers in Ethiopia
Transaction Costs and Market Institutions: Grain Brokers in Ethiopia
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
17<br />
5. ESTIMATION RESULTS FOR TRADERS’ LONG-DISTANCE TRADE<br />
(STEP 1)<br />
Tobit estimations of the share of each trader’s long-distance trade are based on<br />
the shadow transaction costs obta<strong>in</strong>ed from both the net revenue <strong>and</strong> the gross<br />
value of sales IV estimations <strong>in</strong> Table 1. Shares of long-distance trade are<br />
estimated for purchases <strong>and</strong> sales separately, <strong>in</strong> order to control for the location<br />
effect on long-distance trad<strong>in</strong>g behavior, noted earlier. Unconditional or Tobit<br />
elasticities adjust the estimated coefficients by account<strong>in</strong>g for both the effect on<br />
the conditional mean of the dependent variable <strong>in</strong> the positive part of the<br />
distribution <strong>and</strong> the effect on the probability that the observation will fall <strong>in</strong> the<br />
positive part of the distribution (McDonald <strong>and</strong> Moffitt). 10<br />
Tobit elasticities are<br />
obta<strong>in</strong>ed by adjust<strong>in</strong>g the coefficients by the Φ proportion of the sample that has<br />
zero observations of the dependent variable. In effect, this adjustment lowers the<br />
marg<strong>in</strong>al effect by the probability that traders with zero distant transactions would<br />
engage <strong>in</strong> some amount of distant transactions as a result of marg<strong>in</strong>al changes <strong>in</strong><br />
the regressors. In both model specifications, transaction costs have a large <strong>and</strong><br />
10<br />
McDonald <strong>and</strong> Moffitt (1980) suggest a decomposition of the slope vector <strong>in</strong>to: ∂E[y i |x i ] /∂x i =<br />
Prob[y i * >0] ∂E[y i *|x i , y i *>0]/∂x i + E[y i *|x i , y i *>0] ∂Prob[y i * >0]/ ∂x i . The unconditional elasticity is<br />
obta<strong>in</strong>ed by scal<strong>in</strong>g the parameters of the Tobit regression by the probability <strong>in</strong> the uncensored<br />
part of the distribution. See also Greene (1993) for examples <strong>and</strong> Ravallion <strong>and</strong> Dearden (1988)<br />
for an application of this method.