Trinity College Dublin Royal Economics Society Conference April

Trinity College Dublin Royal Economics Society Conference April 19 th, 2011 Income Shocks and Household Risk. Coping Strategies: Evidence from Vietnam Carol Newman and Fiona Wainwright Department of Economics, Trinity College Dublin

Motivation (1) Literature highlights fact that exposure to income risk remains a serious cause of poverty in developing countries (Fafchamps, 2009) Nature of the shock incurred impacts on ability to cope (Dercon, 2002) Natural disasters have been linked to: Idiosyncratic shocks Spatially covariant natural shocks Declines in agricultural yields and incomes (Porter, 2008) Persistent effects on nutrition of children (Alderman et al. , 2006) School attendance and enrolment (Jacoby and Skoufias, 1997) Evidence suggests that idiosyncratic shocks are more common but can be smaller and easier to deal with (Townsend, 1994; Mordoch, 2001; Udry, 1991)

Motivation (2) Idiosyncratic shocks can be insured at community level or through formal insurance contracts However, in developing economies formal insurance markets are not well developed Aggregate shocks are difficult to collectively insure and are rarely covered by formal insurance contracts Households must find alternative ways to smooth consumption in the face of shocks Risk management – managing ex ante the riskiness of the income generating process (income smoothing) – formal insurance, precautionary saving, income diversification Risk coping – dealing ex post with the consequences of income loss (consumption smoothing) – insurance claims, depletion of the stock of saving or borrowing

What we do. . . Present an ex post theoretical model within an inter-temporal utility maximizing framework which we use to explain household’s decisions to insure against idiosyncratic risk and save to protect against uninsurable aggregate risk Using panel data from rural households in Vietnam we develop a test of the extent to which households smooth consumption over time and consider how this depends on the presence of insurance and savings instruments We consider savings stocks in the form of liquid assets as a form of self-insurance or risk-coping strategy against aggregate and idiosyncratic shocks

Conceptual Framework (1) HH is given a wealth endowment W in each period such that With no insurance HH maximizes: = precautionary savings level = random variable representing exogenous natural shock = exogenous idiosyncratic shock with known probability distribution FOC: Approximation for A:

Conceptual Framework (2) With insurance HH maximizes: = proportion of idiosyncratic risk that is covered FOCs: Approximation for A:

Conceptual Framework (3) Summary of theoretical predictions: Households save and purchase insurance to smooth consumption and so the realization of adverse income shocks should not affect consumption levels between one period and the next Households with insurance will precautionary save less than those without given that the uncertainty associated with income due to idiosyncratic risks is eliminated Policy implication: Model predicts that households will precautionary save due to the uncertainty associated with future income levels and so they will consume less than they would in the absence of this uncertainty and will save less in the form of other more productive types of investments.

Empirical Considerations Step 1: Test ability of households to consumption smooth in presence of idiosyncratic and aggregate shocks Obtain parameters for a general consumption function that represents households consumption smoothing preferences For households that experience no shocks estimate: We use the estimated coefficients to predict household consumption in each period using the observed data on each explanatory variable in that period. Test the hypothesis that there is no difference between the predicted and actual observed level of consumption Failure to reject the null hypothesis provides evidence of consumption smoothing

Empirical Considerations Step 2: Empirical exploration of mechanisms of consumption smoothing Definition of shocks: Exogenous spatially covariant shocks: 1) economic (for example, crop price changes, key input price changes/shortages) 2) natural (for example, floods, typhoons, droughts etc) Idiosyncratic shocks: 3) insurable (for example, illness, injury or death of household member) 4) uninsurable (for example, crime/theft, divorce, family disputes etc) Fixed Effects model

Data Vietnamese Access to Resources Household Survey (VARHS) 2006, 2008 and 2010 – 12 rural provinces in Vietnam

Data Vietnamese Access to Resources Household Survey (VARHS) 2006, 2008, 2010; 12 provinces of Vietnam covering 2, 000 households in a balanced panel Table 2 a: Household Shocks and Recovery Statistics Any Shock Spatially covariant Idiosyncratic only Both 2006 -2008 Shock 56% 73% 13% 2008 Recovered 45% 48% 49% 30% Change in asset values Savings Livestock 2010 Shock 50% 71% 13% 16% 2010 Recovered 53% 57% 38% 60% Crops Borrowing No shock 3, 482 5, 755 1, 890 23 Shock 2008 -2010 No shock -1, 078 8, 053 3, 787 3, 906 7, 923 -1, 054 -2, 902 -1, 718 Shock 3, 715 -1, 903 -1, 859 2, 103

Data 55% of households have insurance in 2010 - half of which are voluntarily purchased Profile of households with formal insurance (selected results) Dependent variable: Time Exogenous: Natural Shock Exogenous: Economic Shock Idiosyncratic: Insurable Shock Idiosyncratic: Uninsurable Shock Number of shocks suffered Recovered from prior shock Stock Of Savings Dummy Livestock Dummy Borrowings Dummy Crop Stores Dummy Total Wealth (VND) Income (VND) Education Level of Household Head Purchased Voluntary Insurance 2008 2010 0. 2996** 0. 2099 0. 5389*** -0. 0531 0. 2011 0. 0296 0. 1130 0. 0410 -0. 2464*** -0. 0872 0. 0375 0. 0837 0. 2241*** 0. 3469*** -0. 0383 0. 0740 -0. 0772 0. 1266* -0. 0723 0. 1097 0. 1115*** 0. 1861*** 0. 0007** 0. 0005** 0. 1891*** 0. 2193***

Empirical Results (1) Stage 1: Consumption Smoothing t-tests (2008) Overall No Insurance With Insurance < Median Liquid Assets > Median Liquid Assets < Median Income > Median Income 2008 Full sample Actual 6. 37 Predicted 6. 16 Difference 0. 215*** n 2, 039 No shock Actual 6. 29 Predicted 6. 12 Difference 0. 172*** n 892 Any shock Actual 6. 43 6. 22 6. 74 6. 25 6. 62 6. 01 6. 90 Predicted 6. 18 6. 06 6. 35 6. 07 6. 29 5. 94 6. 45 Difference 0. 248*** 0. 154*** 0. 388*** 0. 173*** 0. 332*** 0. 072** 0. 448*** 1, 147 687 460 604 542 609 538 n

Empirical Results (2) Stage 1: Consumption Smoothing t-tests (2010) Continued Overall No With Insurance < Median Liquid Assets > Median Liquid Assets < Median Income > Median Income 2008 Idio shock Actual 6. 51 6. 36 6. 73 6. 25 6. 89 5. 99 7. 00 Predicted 6. 23 6. 13 6. 39 6. 10 6. 43 5. 89 6. 56 Difference 0. 275*** 0. 231*** 0. 343*** 0. 152** 0. 453*** 0. 093 0. 447*** 235 142 93 138 93 114 121 Actual 6. 35 6. 14 6. 68 6. 13 6. 57 5. 97 6. 85 Predicted 6. 16 6. 04 6. 35 6. 05 6. 26 5. 94 6. 45 Difference 0. 190*** 0. 098*** 0. 337*** 0. 073* 0. 306*** 0. 035 0. 395*** 880 542 338 441 501 379 n Nat shock n

Empirical Results (3) Stage 1: Consumption Smoothing t-tests (2010) Overall 2010 Full sample Actual 6. 62 Predicted 6. 23 Difference 0. 390*** n 2, 036 No shock Actual 6. 64 Predicted 6. 19 Difference 0. 442*** n 1, 014 Any shock Actual 6. 60 Predicted 6. 26 Difference 0. 339*** n 1, 022 No With < Median Insurance Liquid Assets > Median Liquid Assets < Median Income > Median Income 6. 36 6. 10 0. 258*** 535 6. 80 6. 41 0. 394*** 502 6. 28 6. 01 0. 271*** 523 6. 93 6. 52 0. 409*** 499 6. 86 6. 44 0. 428*** 487 6. 40 6. 11 0. 284*** 519

Empirical Results (4) Stage 1: Consumption Smoothing t-tests (2008) Continued Overall No With < Median Insurance Liquid Assets > Median Liquid Assets < Median Income > Median Income 2010 Idio shock Actual 6. 53 6. 29 6. 92 6. 20 6. 93 6. 13 6. 97 Predicted 6. 21 6. 10 6. 38 6. 01 6. 44 5. 94 6. 49 Difference 0. 323*** 0. 195** 0. 531*** 0. 188** 0. 489*** 0. 184** 0. 477*** 231 143 88 127 104 121 110 Actual 6. 63 6. 39 6. 87 6. 47 6. 78 6. 32 6. 94 Predicted 6. 28 6. 10 4. 46 6. 15 6. 41 6. 03 6. 54 n Nat shock Difference n 0. 348*** 0. 290*** 0. 407*** 0. 324*** 0. 370*** 0. 294*** 0. 404*** 864 440 424 425 438 441 423

Empirical Results (5) Main result from Stage 1: Households smooth consumption when faced with income shocks Motivation for Stage 2: What role do precautionary savings and insurance play? Estimate:

Empirical Results (6) Total Liquid Assets (Table 6 selected results) (1) Levels (3) Levels (4) Levels Exog: Nat -0. 036 -0. 034 -0. 033 -0. 005 Exog: Econ -0. 185* -0. 186* -0. 183* -0. 185* -0. 134** -0. 155** -0. 134** -0. 072 -0. 070 -0. 071 0. 019 0. 020 0. 036 Income Shock (2) Levels -0. 166*** Exogenous Shock -0. 099** Idiosyncratic Shock -0. 160*** -0. 153*** Idio: Insurable Idio: Uninsurable Transfers (5) (6) Interaction: : Insurance Trans 0. 022 0. 024 0. 020 Transfers*Nat Insurance Claim Ins*Insurable Shock -0. 041 -0. 096 -0. 092 -0. 124* 0. 159 -0. 092

Empirical Results (7) Total Livestock Holdings (Table 7 selected results) (1) Levels Income Shock (2) Levels (3) Levels (4) Levels -0. 034 Exogenous Shock (5) (6) Interaction: : Insurance Trans 0. 023 - Exog: Nat 0. 001 0. 007 0. 106 Exog: Econ 0. 050 0. 053 0. 064 0. 057 -0. 184** -0. 266*** -0. 181** -0. 069 -0. 064 -0. 062 0. 026 0. 027 0. 096 Idiosyncratic Shock -0. 154* -0. 155* Idio: Insurable Idio: Uninsurable Transfers 0. 023 0. 024 0. 026 Transfers*Nat Insurance Claim Ins*Insurable Shock -0. 150 -0. 279** -0. 281** -0. 282** -0. 281** -0. 398*** 0. 632** -0. 281**

Empirical Results (8) Total Savings (Table 8 selected results) (1) Levels Income Shock (2) Levels (3) Levels (4) Levels -0. 243*** Exogenous Shock (5) (6) Interaction: : Insurance Trans - -0. 191***. Exog: Nat -0. 194*** -0. 192*** -0. 193*** -0. 334*** -0. 067 -0. 081 Idio: Insurable -0. 156* -0. 154* Idio: Uninsurable -0. 001 -0. 132** -0. 133** -0. 206*** Exog: Econ -0. 074 Idiosyncratic Shock Transfers -0. 137* -0. 133** -0. 132** -0. 137* -0. 130** Transfers*Nat Insurance Claim Ins*Insurable Shock 0. 206* -0. 179** -0. 178** -0. 175** -0. 174** -0. 172** -0. 013 -0. 170**

Empirical Results (9) Cash Savings (Table 8 a selected results) (1) Levels Income Shock (2) Levels (3) Levels (4) Levels -0. 253*** Exogenous Shock (5) (6) Interaction: : Insurance Trans -0. 173**. - Exog: Nat -0. 165** -0. 160** -0. 294*** Exog: Econ -0. 208 -0. 192 -0. 2089 -0. 225** -0. 226** -0. 222** -0. 011 -0. 010 Idiosyncratic Shock -0. 200** -0. 198** Idio: Insurable Idio: Uninsurable Transfers -0. 209*** -0. 210*** -0. 213*** Transfers*Nat Insurance Claim Ins*Insurable Shock -0. 276*** 0. 188* -0. 166* -0. 161* -0. 159* -0. 158* -0. 160* 0. 009 -0. 154*

Empirical Results (10) Borrowing (Table 10 selected results) (1) Levels Income Shock (2) Levels (3) Levels (4) Levels 1. 073*** Exogenous Shock (5) (6) Interaction: : Insurance Trans 0. 878*** - Exog: Nat 0. 655*** 0. 657*** 0. 640** Exog: Econ 0. 899*** 0. 897*** 0. 901*** 0. 871*** 0. 901*** 0. 570* 0. 573* 0. 570* 0. 235 0. 236 0. 227 Idiosyncratic Shock 0. 903*** 0. 883*** Idio: Insurable Idio: Uninsurable Transfers 0. 235 0. 222 0. 235 Transfers*Nat Insurance Claim Ins*Insurable Shock 0. 021 0. 010 0. 018 0. 001 -0. 002 -0. 050 0. 229 -0. 002

Summary of findings Households smooth consumption in the face of adverse income shocks Precautionary savings are an important mechanism Households deplete their total stock of liquid assets in response to exogenous economic shocks and idiosyncratic insurable shocks Financial savings, particularly cash and gold held at home, act as important buffers in the face of spatially covariant natural shocks Insurance markets play an important role in easing the depletion of livestock holdings in response to idiosyncratic shocks while external transfers are important for risk-coping in the face of natural disasters Borrowing is increased when households are faced with idiosyncratic and spatially covariant shocks, particularly for wealthy households.

Conclusions and Future Work Savings and insurance instruments are important for consumption smoothing where there are income risks Savings for precautionary purposes may lead to lower welfare outcomes in the long run given that they result in lower levels of consumption and use up resource that could be put to more productive uses This is exacerbated by the fact that, as predicted by our theoretical model, the uncertainty associated with income shocks may lead households to save excessively in these forms The presence of insurance instruments reduces the need to draw down some forms of precautionary savings in the event of adverse income shocks but there is strong evidence to suggest that insurance markets do not fully cover idiosyncratic risks Future work will investigate the interplay between consumption and investment decisions by risk averse households where income is uncertain and insurance markets fail.

Thank you Questions and comments welcome

Trinity College Dublin Royal Economics Society Conference April 19 th, 2011 Income Shocks and Household Risk. Coping Strategies: Evidence from Vietnam Carol Newman and Fiona Wainwright Department of Economics, Trinity College Dublin