Persistent Effects of a Temporary Input Subsidy Evidence from a - - PowerPoint PPT Presentation

Persistent Effects of a Temporary Input Subsidy

Evidence from a Randomized Controlled Trial in Mozambique Michael R Carter, Rachid Laajaj & Dean Yang

University of California, Davis Paris School of Economics University of Michigan . IGC Africa Growth Forum 2014, Accra

June 17, 2014

Carter, Laajaj & Yang Input Subsidies

The Yield Gap Problem

Sub-Saharan Africa largely sat out the seed-fertilizer ’Green Revolution’ that swept most of the developing world over the 1960-2000 period with tiny increases in the yields of basic grains In 2009, SSA farmers used an average of 13 kg/hectare of fertilizer, compared to 94 in other developing countries Mozambique is no exception to this pattern–at the national level, most maize farmers use no fertilizer and yields are less than 1 ton per-hectare International Fertilizer Development Center program in Mozambique identified a gaping 2-3 ton/hectare yield gap between what is possible with existing technologies and what farmers currently achieve

Carter, Laajaj & Yang Input Subsidies

Input Subsidies: Temporary or Permanent?

Governments & (sometimes) aid agencies have responded to this apparent yield gap challenge with input subsidies Malawi, beginning with its “starter pack” program in the early 1990s has led this approach in sub-Saharan Africa Across 10 countries implementing input subsidy programs, 2011 expenditures totaled $1.05 billion, or 28.6% of public agricultural spending

Carter, Laajaj & Yang Input Subsidies

2011 Expenditures on Input Subsidies

8.4% 10.4% 18.1% 25.7% 26.0% 26.1% 29.9% 39.9% 46.0% 58.3% 0% 10% 20% 30% 40% 50% 60% 70% Burkina Faso Ethiopia Mali Kenya Nigeria Senegal Ghana Zambia Tanzania Malawi Expenditures as % of public agricultural spending

Source: Jayne and Rashid (2013)

Carter, Laajaj & Yang Input Subsidies

Input Subsidies: Temporary or Permanent?

High opportunity cost of these funds raises question, “Why subsidize a privately profitable input?” Subsidies can arguably make sense as they can break a low technology poverty trap they:

Make technology affordable for low income farmers (i.e., relax liquidity constraints) Share the risk of experimentation Reduce learning costs & break the ’let someone else experiment’ equilibrium

These are all arguments for temporary subsidies But will a one-time subsidy work & will its impacts persist over time? If not, must input subsidies become a permanent feature of the agricultural and budgetary landscapes?

Carter, Laajaj & Yang Input Subsidies

Possible Short Term Impacts of Subsidies

Carter, Laajaj & Yang Input Subsidies

Possible Long Term Impacts of Subsidies

Carter, Laajaj & Yang Input Subsidies

Mozambique Test of Temporary Input Subsidies

Impacts on Fertilizer Use: Control Group

Carter, Laajaj & Yang Input Subsidies

Mozambique Test of Temporary Input Subsidies

Impacts on Fertilizer Use: Short term

Carter, Laajaj & Yang Input Subsidies

Mozambique Test of Temporary Input Subsidies

Impacts on Fertilizer Use: Long-term

Carter, Laajaj & Yang Input Subsidies

Mozambique Test of Temporary Input Subsidies

Impacts on Maize Production

Carter, Laajaj & Yang Input Subsidies

Summary of Results

As we shall see, these simple ’intention to treat’ estimators understate impacts for those who actually used the subsidies We will also see that input subsidies also have impacts on:

Agricultural production (not only maize) Household consumption Assets Housing investments

Impacts of one-time subsidies persist into two subsequent post-subsidy seasons Evidence for learning: positive impacts on estimated returns to modern inputs Positive spillovers to control group Based on random variation in share of social network in treatment group

Carter, Laajaj & Yang Input Subsidies

Mozambique Voucher Program

Voucher program funded by European Union, implemented by Mozambican government, FAO, and IFDC Followed the AIMS program that expanded agro-dealer network in study area 2 year program 2009-2010 and 2010-2011, 25,000 vouchers of maize and rice in 5 provinces Evaluation focuses in Maize vouchers in the Manica Province Voucher gave 73% subsidy on package of 12.5 kg of improved seeds (either OPV or hybrid), and 100kg of fertilizer Market price of package was about USD 113, with the individual having to pay about USD 30 to use coupon

Carter, Laajaj & Yang Input Subsidies

Research Design

94 localities randomly assigned to one of three treatment arms “Subsidy only” localities are the subject of this paper Eligibility rules at household levels (0.5-5 hectares in maize; willing & able to make voucher co-pay) Subsidies assigned by random lottery to eligible households within 41 villages

Carter, Laajaj & Yang Input Subsidies

Theoretical Model of Subsidies

Paper details a model of technology adoption by credit-constrained, risk averse households with imperfect information on returns to fertilizer: Without input subsidy, non-adoption is possible if initial wealth is low; exacerbated by low beliefs about returns to fertilizer One-time subsidy can lead some non-adopters to adopt Persistence of adoption (beyond subsidy period) is possible via wealth effect

Wealth effect is stochastic, dependent on realized production

• utcomes

Sustained adoption may be fragile: poor outcomes in future may lead to disadoption

Sustained adoption more likely (and stable) if there is also a learning effect

Carter, Laajaj & Yang Input Subsidies

Theoretical Model of Subsidies

Other explanations of low fertilizer use tend to imply that a temporary subsidy would at most have a temporary effect:

Bio-physical limitations, leading to negative returns at unsubsidized prices (Marenya & Barrett 2009) Behavioral biases: Self-control problems (Ashraf et al 2006) or naïveté about stochastic temptation (Duflo et al 2011) Savings constraints, e.g. due to egalitarian norms and demands from social network (Platteau 2000)

Our finding of persistence effects suggests that these explanations are not the whole story Let’s now look at results in more detail

Carter, Laajaj & Yang Input Subsidies

Voucher Uptake & Use

Carter, Laajaj & Yang Input Subsidies

Impacts on Input Use

Carter, Laajaj & Yang Input Subsidies

Impacts on Production

Carter, Laajaj & Yang Input Subsidies

IV Estimates of Returns to Fertilizer

Given prices, estimates imply a benefit cost ratio of 220% in first year, higher in later years

Carter, Laajaj & Yang Input Subsidies

Impacts on Consumption & Savings

Carter, Laajaj & Yang Input Subsidies

Impacts on Learning

Carter, Laajaj & Yang Input Subsidies

Learning from Others

We collected data at outset on social network links with other study participants Definition of social network member: individuals with whom you discussed agriculture in the 2010 season (prior to the project) “moderately” or more Can estimate effect of fraction of social network randomly assigned to treatment group Regression equation: Yiv = α +βZiv +δ1SNSiv +δ2SNTRiv +θv +εiv where SNS is the size of the individual’s social network and SNTR is the fraction of the social network that won the voucher lottery

Carter, Laajaj & Yang Input Subsidies

Learning from Others

Carter, Laajaj & Yang Input Subsidies

Conclusions

On average, fertilizer appears to be profitable (heterogeneity, however) Impact of temporary input subsidies persist, a finding that is consistent with common economic models, but less so with behavioral explanations Strong evidence that learning is an important part of the story Overall uptake was quite modest –> other constraints Important to emphasize that intervention took place against the background of a program to expand fertilizer dealer network (AIMS program) Forthcoming work on Matched savings intervention

By itself, appears to have same impacts as vouchers But no added impact of combining them

Interactions between behavioral and material

Finding that interventions extend planning time horizon for poorer respondents, suggesting either a further learning or aspirational effect

Carter, Laajaj & Yang Input Subsidies

Regression Approach

Reduced form (ITT) for individual i in community v with treatment indicator Ziv: Yiv = α +βZiv +θv +εiv (1) Estimate in levels of outcome variables and using Inverse hyperbolic sine transformation: sinh−1(x) = ln(x +(x2 +1)0.5)

Carter, Laajaj & Yang Input Subsidies