Using individual-level randomized treatment to learn about market structure

Out­line

The de­gree to which in­ter­me­di­aries com­pete is a long-stand­ing ob­ject of in­ter­est in stud­ies of agri­cul­tur­al mar­kets in de­vel­op­ing coun­tries. Com­pe­ti­tion shapes how price sig­nals prop­a­gate along sup­ply chains, and the wel­fare im­pli­ca­tions of tax­es and sub­si­dies for pro­duc­ers and con­sumers. How­ev­er, the ev­i­dence is re­mark­ably thin. In the lim­it­ed cas­es where con­clu­sions can be drawn, the re­searchers see a high de­gree of com­pe­ti­tion. On the oth­er hand, gov­ern­ments and in­ter­na­tion­al in­sti­tu­tions of­ten em­pha­size the monop­son­is­tic pow­er of traders. Pre­vi­ous stud­ies of com­pe­ti­tion in these mar­kets have pri­mar­i­ly re­lied on ob­ser­va­tion­al data, an­a­lyz­ing trad­er price-cost mar­gins, price dis­per­sion across space, or the pass-through of in­ter­na­tion­al prices along the sup­ply chain. In con­trast, we pro­pose an ex­per­i­men­tal ap­proach for es­ti­mat­ing the de­gree of com­pe­ti­tion.

Shed­ding light on mar­ket struc­ture

Our ex­per­i­ment is based on the ran­dom­iza­tion of unit sub­si­dies to com­pet­ing traders for their pur­chas­es from farm­ers. By com­par­ing prices that sub­si­dized and un­sub­si­dized traders pay to farm­ers, we can re­cov­er the rate of dif­fer­en­ti­a­tion be­tween traders, the key mar­ket struc­ture pa­ra­me­ter in a stan­dard mod­el of oligop­son­is­tic com­pe­ti­tion. By com­bin­ing the ex­per­i­men­tal re­sults with qua­si-ex­per­i­men­tal es­ti­mates of the pass-through rate, we also es­ti­mate mar­ket size, or the ef­fec­tive num­ber of traders com­pet­ing for farm­ers’ sup­ply.

In the con­text of the Sier­ra Leone co­coa in­dus­try, our re­sults point to a com­pet­i­tive agri­cul­tur­al trad­ing sec­tor and sug­gest that the mar­ket size is sub­stan­tial­ly larg­er than the vil­lage. The method­ol­o­gy de­vel­oped in this study uses pure­ly in­di­vid­ual-lev­el treat­ment to shed light on mar­ket struc­ture. This ap­proach may be use­ful for the many cas­es where mar­ket-lev­el ran­dom­iza­tion is not fea­si­ble.

Re­search Team

Author

Lorenzo Casaburi

Associate Professor of Development Economics

Zurich ZCED

An­drea Guar­iso

Trin­i­ty Col­lege Dublin