Impact Evaluation of the Land Certification Program in Ethiopia: Challenges and Opportunities Daniel Ayalew Ali 19 April 2011
Policy Environment The federal proclamation devolved responsibility for land policy to the regions, resulting in differences in key features of the land legislation across regions
Key differences Land Redistribution Inheritance Restriction on rental market Risk of losing land Amhara Oromia SNNPR Tigray Requires support from landholders Family member (residency is required) No restriction (25 years) If earns permanent income from non-ag activities Prohibited except in irrigable land Family member, no residency requirement Only up to half of the land, restriction on duration (3/15) If leaves village of residence Do not explicitly rule out Family member Only up to half of the land, restriction on duration (5/10) Improper use (e.g., left fallow for more than 2 years) Do not explicitly rule out Only to landless children and parent Only up to half of the land, restriction on duration (2/20) If leaves village of residence for more than 2 years
Program characteristics Main features of first stage certification Democratically elected LAC has responsibility Field-based adjudication process (elders to resolve conflict) No spatial element included Very cost effective (<1US$/parcel, 3.2 US$/ha) Significant variation across regions
Key differences in program Amhara Oromia SNNPR Tigray Format of certificate/regis try book Green booklet with rights and obligations Green booklet with rights and obligations Green booklet with rights and obligations Blue cardstock Women s rights Joint ownership Joint ownership Joint ownership Under the name of the head, and spouse is listed as hh member Photograph Both head and spouse Only head Both head and spouse No photograph Differences were also observed in actual field measurement and adjudication process!!
Opportunities for IE Huge program quite rapidly implemented with little outside support Highly decentralized process and low cost Heterogeneity across regions (inter regional comparison) Policy environment Program implementation
Challenges for IE No impact evaluation strategy embedded into the program Lack of baseline data designed for the purpose of assessing the impact of the program Post program impact evaluation requires innovative solutions to separate out the impact of the program from other compounding factors
Practical Solutions Focus on specific regions building on existing panel data conducted before the start of the program WB studies: Amhara regions Holden, et al.: Tigray region Lost opportunity: heterogeneity in policy and program implementation across regions has not been captured Important for the process of harmonization
Data and general approach of the WB Study Data sources 4-round panel (1999, 2001, 2004, 2007) from East Gojam Some 900 households with 4,000 plots in each round 3 rounds before implementation, some certified in 4 th round Expected short-term impacts Perceived tenure security Land-related investment (soil conservation measures) Rental market participation Identification strategy Define villages as treated if certified > 12 months ago Control for household specific effects Conservative estimate for two reasons Even in treated villages not all households have certificates Many households in untreated ones had certificates
Identification Strategy: certificates were issued 12 months before the survey (July-Aug 2007) No (Control) Villages Gozamin District (2 villages), Enemay District (2 villages) Yes (Treatment) Machakal District (1 village), Gozamin District (2 villages) Program introduced May 2005 June 2005 Feb 2003 Feb 2004 Registration completed Jan-Aug 2006 1 village not completed Start of certificate distribution Sep 2006 Not yet started in 2 villages June 04 May 2005 Feb 2005
Outcome variables considered Perception of administrative land size changes in the next 5 yrs Both loss and increases are relevant Investment in soil & water conservation structures Expect a direct investment effect Maintenance and new construction No of hours spent in maintaining/constructing Land rental market participation Amhara is the only region where no constraints in terms of length Look at both participation and area transferred
VAR 1: Expect a change in holding size (%) Year Control villages Treatment villages 1999 0.75 0.78 2002 0.62 0.64 2004 0.38 0.38 2007 0.39 0.24 Number of households 477 356
VAR 2: Investment in soil and water conservation over the past 12 months (%) Year Control villages Treatment villages 2004 0.36 0.12 2007 0.24 0.25 Number of plots 2415 2086
VAR 3: Participation in rental market (rented-out land - %, area) Year Control villages Treatment villages % Area (ha) % Area (ha) 1999 0.24 0.20 0.17 0.14 2002 0.21 0.21 0.15 0.16 2004 0.29 0.28 0.26 0.32 2007 0.34 0.34 0.33 0.45 Number of hhs 477 356
Equation to be estimated Risk of land loss y it w x γ c u, (1) t it it i it use Chamberlain s method
Table 6: Impact of certification on perceived land tenure security: Marginal effects from Chamberlain random effects probit model Expect an Increase Expect a Decrease T Village. T Household T Village. T Household Certificate issued -0.135 *** -0.099*** -0.095 *** -0.095*** (4.56) (2.808) (4.74) (4.350) Relative land size -0.106 *** -0.106*** 0.041 *** 0.042*** (6.37) (6.364) (3.83) (3.939) Head s age -0.016 *** -0.016*** 0.010 ** 0.011** (3.22) (3.197) (2.36) (2.446) Other household level controls, time dummies, and mean values of time varying household level variables are included in all the specifications. Number of observations 3,042 3,042 3,042 3,042 Number of households 882 882 882 882
Investment & rental market Plot-level investment equation y jit ψ w t it x γ it p jit δ x i ξ a i u jit, Household-level rental market participation y it ψ w x γ y y x ξ t it it a u i, t 1 0 i0 i i it,
Table 7: Impact of certification on propensity and magnitude of investment in soil & water conservation Repairs & new invest. probit model Hours spent tobit model T Village. T Household T Village. T Household Certificate issued 0.290*** 0.191*** 1.382*** 0.788*** (6.096) (3.699) (6.257) (3.524) Parcel size in hectares 0.061*** 0.063*** 0.496*** 0.492*** (3.810) (3.916) (4.669) (4.686) No of years possessed 0.000 0.000 0.008 0.006 (0.260) (0.075) (1.025) (0.859) Other plot and household level controls, time dummies, and mean values of time varying household level variables are included in all the specifications. Number of plots 8,671 8,671 8,671 8,671 Number of households 856 856 856 856
Table 8: Certification impact on rental market participation Participation (Rented out): probit model Size of land rented out: tobit model T Village. T Household T Village. T Household Certificate issued 0.126** 0.114* 0.092** 0.102*** (1.977) (1.954) (2.237) (2.662) Total owned land in ha 0.054** 0.057** 0.075*** 0.076*** (2.364) (2.493) (3.143) (3.287) Lagged participation/size 0.338*** 0.338*** 0.079** 0.078** (6.891) (6.928) (2.049) (1.998) Includes controls for household level characteristics, time dummies, and mean values of time varying household level variables. Number of observations 1,302 1,302 1,302 1,302 Number of households 657 657 657 657
The way forward: IE strategy for LALUDEP Include IE strategy early on to draw lessons for scaling up as well as implementing similar program in other countries Rigorous impact evaluation at least requires: Establishing counterfactual: comparable control and treatment areas before the program Data: baseline and follow-up surveys
Example: second stage certification in selected districts (demand driven) Set selection criteria to identify eligible districts Randomly assign the eligible districts/villages into early and late intervention areas Stratify by region to capture heterogeneity Plan for baseline survey before the start of the program and follow up survey before the start of the program in the late intervention areas This approach thus combines randomization with panel data
Determinants of value of crop output: household fixed effect estimates Value of crop output (log) Plot has soil and water conservation structures 0.091*** Plot size (hectares) 0.276*** Male adult labor (days) 0.334*** Female adult labor (days) -0.001 Hired labor (days) 0.023 Oxen (days) 0.128*** Chemical fertilizer (kg) 0.155*** Manure (kg) 0.027** Other controls (plot characteristics and time dummies) Number of observations 11,689 Number of households 844