Allocative Inefficiency, Tenure Systems and Poverty in Irrigated Agriculture in Pakistan by Dr. Ariel Dinar and Mr. Sanval Nasim, University of California-Riverside, USA
Allocative Inefficiency, Tenure Systemsand Poverty in Irrigated Agriculture inPakistan Ariel Dinar, Steven Helfand, and Sanval Nasim Water Science and Policy Center University of California, Riverside The Pakistan Strategy Support Program Competitive Grants Conference, Islamabad 9th February, 2013
Introduction• Ineffective water-management policies in the past have affected water availability• Institutional constraints also affect the degree of utilization of water in Pakistan’s agricultural sector. • Tenure arrangements could lead farmers to misallocate inputs• Differences in incentives across tenure systems may explain part of Pakistan’s water-management problems.
Research Question and ObjectivesResearch Questions• How are farmers using irrigation water? • Water-use efficiency • Economic approach to measurement of water-use is through estimation of allocative efficiency• Does this efficiency differ across land tenure systems and to what degree? • Tenure systems affect the incentives farmers face to utilize resources efficiently
Objective• We seek to obtain estimates of allocative inefficiency so that the degree of over (or under) utilization of water across tenure systems can be quantified in an empirically consistent manner.• We will use our analysis to evaluate possible water policy reforms conditional on land tenure systems and political and economic feasibility, and to compare the impact of these reforms on agricultural incomes and poverty across land tenure systems.
Stochastic Profit Model• Translog specification (Kumbhakar and Lovell, 2000):Production function and input demand system ln y = b0 + å bn ln xn +åg n ln zq n q 1é ù 1é ù + êåå b nk ln xn ln xk ú + êååg qr ln zq ln zr ú 2ë n k û 2ê q r ë ú û +åådnq ln xn ln zq n q +v - u• X: variable inputs• Z: quasi-fixed inputs• u: technical inefficiency• ξ = input-specific allocative inefficiency
ConstraintsA: vector of allocative inefficiency explanatory variablesH: vector of technical inefficiency explanatory variablesEstimation• Iterated nonlinear seemingly unrelated regression • Maximum likelihood
Data• Pakistan Rural Household Survey II (PRHS-II) • Information on two seasons: 2003 kharif (autumn harvest) and 2004 rabi (spring harvest). • 887 agricultural households and 1690 plots• Initially planned to use a panel regression with PRHS-I and PRHS-II. However, not all variables are measured consistently across datasets. Will need more time to sort this out for the final report. • We lose fixed-effects and price variation • Fixed-effects still possible with PRHS-II as long as sample is constrained to households with multiple plots • Ease of estimation
VariablesY: weighted output quantity index (wheat, irri-rice, basmati rice, cotton,and sugarcane)X (variable inputs):Labor (N), fertilizer (F), and groundwater (G)Z (quasi-fixed inputs):Capital (K) and Land (L)Technical inefficiency explanatory variables:Relative farm size size (small if less than 10 acres)years of cultivation experienceInitially, we proposed to include surface water. However, surface water allocationsand price are fixed. Consequently, it is not a variable factor of production and itsallocative inefficiency cannot be estimated in the current framework. We intend toinvestigate explore this issue further in the final report.
Allocative inefficiency explanatory variables:Tenure dummies; total plot size; years of cultivation experienceControl variables:Land value; access to surface water dummy; location on watercoursedummies; groundwater quality dummies; district dummies; and seasondummyAlso control for zero input quantities (Battese, 1997)
Initial Estimation ResultsProduction function parameter estimates
Groundwater Allocative Inefficiency Distribution Across PlotLevel Characteristics
Moving Forward• These are preliminary results. Much more to be accomplished in the future.• Include more inputs • Separate male-female and owned-hired labor. • Include an index of minor inputs • Include surface water in a more systematic manner• Test different model specifications• Control for fixed-effects• Consider system level inefficiencies (rather than the current farm level inefficiencies)