Testing for Poverty Traps: Asset Smoothing versus Consumption Smoothing in Burkina Faso (with some thoughts on what to do about it)

Transcription

1 Testing for Poverty Traps: Asset Smoothing versus Consumption Smoothing in Burkina Faso (with some thoughts on what to do about it) Travis Lybbert Michael Carter University of California, Davis

2 Risk & Why It Matters In the presence of full and complete financial markets, risk presents no particular problem But without those market or other forms of mitigation, risk may make & keep people poor by distorting income generation & asset accumulation strategies In this context, notion that individuals can mitigate risk by low cost asset destabilization/consumption smoothing strategies is especially important But how and for whom does consumption smoothing work?

3 Canonical Model of Savings & Consumption max c i 0, M t1 E t t 0 u c it subject to: c F( ) ( M M )(1 r) t it it it 1 it M it 0 t FONC for this problems implies a constrained form of the Permanent Income Hypothesis: u ( c ) max{ u ( x ), E [ u ( c )]}, t t t t 1 1 r where and x M y 1 t t t

4 7% Solution to 10% Problem Coefficient of variation of income: 10% Coefficient of variation of consumption: 4% Smoothing Ratio: 40%(=4/10)

5 Testing the Efficacy of Consumption Smoothing Consumption should be unresponsive to transitory shocks (under interior solution) Alternatively, asset liquidation should follow shocks In Burkina Faso, we might look to see if shocks to crop income are matched by livestock sales:

6 Livestock Sales & Shocks Fafchamps, Marcel, Chris Udry and Katherine Czukas (1998). Drought and saving in West Africa: are livestock a buffer stock? Journal of Development Economics 55:

7 Pooled Regression on Sales Fafchamps, Marcel, Chris Udry and Katherine Czukas (1998). Drought and saving in West Africa: are livestock a buffer stock? Journal of Development Economics 55:

8 Rethinking Theory of Savings & Consumption Problem with empirical literature is that no clear alternative model against which the consumption smoothing hypothesis is being tested (if coefficient relating shocks to assets does not equal -1, then what?) There are some ad hoc empirical approaches in the literature Consumption smoothing as luxury for non-poor (Jalan & Ravallion) But this does not tell how to look and further reifies arbitrary (nonbehavioral) poverty lines (Kazianga & Udry on Burkina) Nor does it tell us much about what it means & how to fix it So when does consumption smoothing not make sense? Deaton-esque answers (AR(1)) apply to everyone and make it hard to explain inconsistent evidence of Fafchamps et al. Poverty trap models with richer economic environment: Multiple assets (productive & buffer) Asset price risk, especially with covariant shocks Non-convexities irreversabilities Common property of Micawber Threshold Let s look at contrasting results from Zimmerman & Carter analysis:

9 Income and Consumption (Log Scale) Income and Consumption Under Optimal Stable Strategies Entrepreneurial Consumption Income Defensive Time Zimmerman, F. and M. Carter (2003). Dynamic Portfolio Management under Risk and Subsistence Constraints in Developing Countries J of Dev Econ.

10 Zimmerman, F. and M. Carter (2003). Dynamic Portfolio Management under Risk and Subsistence Constraints in Developing Countries J of Dev Econ.

11 Threshold Econometrics Net Livestock Sales ivt P T U 1 yivt 2 yivt 3 yivt β z z i ivt if Livt L% ivt P T U 1 yivt 2 yivt 3 y ivt βz zivt i ivt if Livt L% Note that this model inadequate from perspective of ~ Barrett-Carter-Ikegami model which conditional L( i ) But need to wait for these econometrics In the meantime, how recover estimates of permanent & transitory income components?

12 Recovering Income Components α z +α F X + F ivt 1 ivt 2 vt ivt v vt i ivt Permanent Income y α z P ivt ivt 1 ivt i Transitory Income y α F X + F Unexplained Income T ivt ivt 2 vt ivt v vt U ivt ivt ivt Note that could decompose transitory into (more easily insured) idiosyncratic & covariant components y

13

14

15

16 Threshold Regression Results Note cattle costs 30-40,000 CFA almost perfect c-smoothing for upper regime!

17 Threshold Regression Results Note cattle costs 30-40,000 CFA almost perfect c-smoothing for upper regime!

18 Conclusions 1 Whose assets are really smoothed by asset smoothing? asset smoothing implies an attempt to preserve assets, but consumption is an input into the formation and maintenance of human capital. [Thus] the distinction between consumption and asset smoothing, while useful as a descriptive tool, may be somewhat misleading. Rather, household responses to adverse shocks are effectively changes in their asset portfolio, with a critical issue being the extent to which the draw down of a given asset has permanent consequences. (Hoddinott 2006) Costs of unmitigated risk for asset smoothers can thus be quite high Insurance instruments that mitigate some of this risk would thus seem to be a promising development policy instrument: Sustain human capital investment Positive moral hazard effects on asset portfolio Social benefits of reduced indigency But can it be done?

19 Financial Instruments to Mitigate Smallholder Risk So can insurance be made to work for smallholder sector? Conventional (individual) insurance unlikely to work: Transactions costs Moral hazard/adverse selection However, index insurance avoids problems that make individual insurance unprofitable for small, remote clients: No transactions costs of measuring individual losses (payouts based on a single index for a location) Preserves effort incentives (no moral hazard) as no single individual can influence index. Adverse selection does not matter as payouts do not depend on the riskiness of those who buy the insurance So let s look at how this might work in the case of Burkina Faso

20 NDVI-based Index for Grains in Burkina

21 NDVI-based Index for Grains in Burkina

22 NDVI-based Index for Grains in Burkina

23 Conclusions 2 So can this be made to work, will there be uptake, and will it sufficiently mitigate risk such that costly asset smoothing can be offset? Related pilots underway stay tuned!

24 Thank you for your time, interest and comments!

25 High Quality Data NDVI-based Livestock Mortality Index NDVI February 2009, Dekad 3 Deviation of NDVI from long-term average February 2009, Dekad 3

26 Livestock mortality index One possible index is based on area average livestock mortality predicted by remotely-sensed (satellite) information on vegetative cover (NDVI):

27 Estimate separate response functions for distinct clusters (Marsabit District) Geographic Clusters

28 Index performance

29 Index performance Index predicts large-scale losses very well Performance of mortality index in predicting insurance trigger Location Strike Correct decision Incorrect decision False positive False negative Chalbi 10% 71% 13% 17% 15% 81% 6% 13% 20% 88% 4% 8% 25% 85% 10% 4% 30% 94% 4% 2% 35% 92% 6% 2% 40% 94% 6% 0% Laisamis 10% 80% 9% 11% 15% 88% 3% 9% 20% 84% 9% 6% 25% 81% 14% 5% 30% 84% 13% 3% 35% 94% 6% 0% 40% 95% 5% 0%

30 How will IBLI work?