Land Consolidation Results from a Randomized Experiment Preliminary Quy-Toan Do Trung Dang Le
The context Vietnam moved from a centralized economy in 1986 Farmers become residual claimants in 1988 Land law of 1993: Land Use Certificates instituted
The question Land reallocation in 1988 based on equity High levels of fragmentation Market institutions in 1993 aimed at reducing fragmentation via market forces Land consolidation did not happen spontaneously
Why? Lack of information Lack of coordination Bureaucratic incentives Transaction costs Fragmentation is just optimal as it is
Methodological problems Observational data are problematic: differences in land market activity, land market outcomes, or land right regimes can be due to (unobservable) factors that affect socio-economic outcomes Comparing before and after change is also problematic: change in land governance usually comes with a lot of other changes that affect growth
Description of our intervention End 2006 – beginning 2007: design of an intervention to trigger land consolidation Three packages: Information campaigns Local cadres training Combined package
Description (contd) Sample universe: communes in the North in volunteering provinces that did NOT undertake land consolidation at the time of the intervention Randomized allocation of the packages across four groups (3 treatments and one control group) Communes are communes in the Vietnam Household Living Standards Study panel survey
Empirical methodology Baseline data: VHLSS 2004 Household survey: representative at province-level, and detailed agricultural module Intervention: Follow-up data: Preliminary data from VHLSS 2008 Household survey: panel structure at the commune-level (100%)
Empirical methodology (contd) Quantitative impact evaluation: Y ijt = a + X ijt b + cT j + d Year t + g T j *Year t + u ijt Where Y is the outcome of interest X is a vector of household characteristics T is the treatment variable, Year is the year dummy variable u is the error term (allowing for HH, commune FE)
Preliminary results Summary statistics Differences at baseline Preliminary look at the results Effect on land fragmentation Effect on land area Effect on agricultural production Effect on income sources
Summary statistics means.d.N means.d.N Total number of plots Land area Agricultural costs Income sources gender age household size ethnicity
Balance Treatment communesControl communes Household size Poverty headcount Per-capita expenditure # income sources Total number of plots Number of annual land plots Cost of ag. Production Income diversification (H)
Results I: land fragmentation # of plots of all land types# of plots of annual land Dummy for treatment *** *** [0.29][0.30][0.73][0.28][0.29][0.67] Dummy: 2008 = 1, 2004 = 01.44***1.74***1.69***1.26***1.52***1.55*** [0.39] [0.43][0.37] [0.39] Treatment*Year 2008== [0.64][0.61][0.72][0.60][0.58][0.67] Household characteristicsNoYes NoYes Commune fixed-effectsNo YesNo Yes Observations R-squared
Results II: land area Areas of all land types in logAreas of annual land in log Dummy for treatment *** *** [0.11][0.09][0.22][0.10][0.08][0.07] Dummy: 2008 = 1, 2004 = ** [0.10][0.09] [0.08] Treatment*Year 2008== [0.16][0.14][0.13][0.14][0.12][0.11] Household characteristicsNoYes NoYes Commune fixed-effectsNo YesNo Yes Observations R-squared
Results III: agricultural production Cost of agricultural production in log Dummy for treatment *** [0.08][0.07] Dummy: 2008 = 1, 2004 = 00.83***0.88***0.86*** [0.08] [0.09] Treatment*Year 2008== [0.12][0.11][0.12] Household characteristicsNoYes Commune fixed-effectsNo Yes Observations2533 R-squared
Result IV: Income sources Herfindahl index of income sources Dummy for treatment *** [0.01] [0.02] Dummy: 2008 = 1, 2004 = [0.01] Treatment*Year 2008==10.06*** [0.02] Household characteristicsNoYes Commune fixed-effectsNo Yes Observations2673 R-squared
Conclusion The main contribution is methodological Results are preliminary: no definitive conclusion to be drawn at this early stage