Endogenous Matching and Agricultural contract Choice among Rice farmers in Bangladesh Sharmina Ahmed, PhD student School of Economics, University of Adelaide AARES Annual Conference, 2012
Research Question What are the channels by which landlords and tenants are matched in the informal agricultural tenancy markets?
Background and Motivation Some of theoretically relevant characteristics are partially observed (i.e. proxied) or unobserved. Under the presence of imperfect/missing markets there may be incentives to ‘matched’ with each other according to economic variables. ‘Endogenous matching’ generates correlation between observable characteristics of one party and proxy errors of the other party- causing biased estimation.
Background and Motivation In Bangladesh: Role of agriculture ─ Major industry (employment) : 62% nationally (2005) Agricultural land- major source of income and asset Informal land leasing is major source of land input in rural Bangladesh ─ 45.8 % farmers cultivate their own land, 20.4% lease out their land and 33.8% of farmers operated as tenants in 2005
Conceptual Model (ref. Ackerberg and Botticini, 2002) A simple contract choice equation: Where, c = binary choice variable (= to 1 if a share contract is observed; 0 if it is a fixed-rent contract t and l = tenant’s and landlord’s characteristics, includes risk aversion r= crop variety The regression results in bias under two conditions: An inaccurate proxy (tenant's wealth)for the risk aversion. Using proxies (o) for such unobserved characteristics:
Conceptual Model (contd.) Association between crop types/crop practices and the tenants Risk averse agents may match with crops that are less risky. Estimation concern: if such endogenous matching exists, omitted variable bias arises because of the correlation between the regressors and the residuals.
Estimation Methodology Step 1 : Identify the existence of possible sources of matching ─ Matching between the tenant farmer and crop practice ─ Matching between the tenant (the landlord) and the landlord (the tenant) Step 2: Estimating fixed-effect model with binary contract choice dependent variable with and without using IV methods and compare the results.
Data Sample Survey, 2000 and 2004 - collected by IRRI Sample size – 2000: 1882 farming households 2004: 1927 farming households (includes 92.5% households of 2000’s) Data include both household level and plot level characteristics
Results: Endogenous variables Crop practice ( debt/capita–multi-cropping) *Poorer tenants choose safer crop practice L_debt/income ratio ( L_debt/income – T_age) L_wealth ( L_wealth – T_age ) *Less risk averse landlord match with less experienced tenants (details in the paper)
Instruments Instruments: divisional dummies access to electricity dummy Instruments that affect the matching equation but do not affect the contractual choice equation or proxy error. – divisional dummies represent temporal markets where distribution of crop practice and tenants and landlords characteristics differ – access to electricity dummy affects level of risk preferences but do not affect contract choice
Contract Choice Equation
Findings Tenant’s level of risk aversion affect the choice of contracts between share-cropping and fixed-rent. Evidence that particular types of crop practices chosen by particular types of tenants and particular type of landlords matched with particular types of tenants. Naïve estimates ignoring this matching can give misleading results.
Future Work Absence of well-defined theoretical framework Find out the process through which the landlords and the tenants match with each other and build a fully specified model of both matching and contract choice.
Conclusion risk-sharing considerations have an impact on the choice of share contracts. With imperfect insurance or capital or land markets (especially in the case of developing countries), the findings of this paper can provide to current policy debates on the role of agrarian arrangements