Welfare shocks, coping strategies and safety nets: a study on rural households of Cachar District, Assam

Abstract

Effective shock management is an integral part of any well-defined and comprehensive welfare strategy. As the type of shocks facing rural households in developing countries depends, inter alia, on local geographical, socio-economic and climatic conditions, micro-level studies profiling the exposure to shocks in specific regional contexts become important for designing and delivering programmes of economic development at the grass roots. Using primary data collected from 444 rural households, this study attempts to study various facets of household-level welfare shocks in Cachar District of Assam. Specifically, the research tries to identify the various sources of shocks confronting rural households in the study area along with the factors that determine exposure to different types of shocks. Also, factors influencing shock severity, frequency and choice of coping strategies are analysed using suitable econometric methods. Based on the finding that the incidence of shocks is very high in the rural areas of the district, the study also takes stock of the current coverage of insurance programmes available to rural households. It is found that while benefits of social insurance programmes are available only to a miniscule percentage of surveyed households, private insurance remains out of reach of most households due to the costly premiums payable on such schemes. The study underscores the need to improve social infrastructure to reduce the prevalence of certain types of shocks while designing effective and comprehensive social insurance programmes to provide the necessary safety cushion to rural households in times of distress.

Appendix

Cluster analysis for livelihood identification

On the basis of data collected on the occupational choice of adult working members of households (aged 21 and above), cluster analysis was carried out to identify the livelihood strategies of households. Such an analysis was deemed necessary for analysing whether exposure to shock differs by livelihoods. For conducting the cluster analysis, the proportion of total household workers engaged in different occupations was computed and k means clustering technique was employed. On the basis of the elbow plot, seven livelihood clusters were chosen. However, the cluster analysis could be done only for those households which had adult working members. There were sixty-four households in the sample which had no adult workers and were entirely dependent on pensions, remittances or other types of transfers. So, these households were included as a separate group (Figure 7 and Table 15).

Figure 7

Elbow plot

Table 15 Distribution of sample households across livelihood clusters

Economic Infrastructure Index (EEI)

The EEI was computed for all villages in Cachar District using Population Census data of 2011 and by employing multiple correspondence analysis (MCA). MCA was employed as all the variables used in compiling the index were binary in nature. For the regression analysis only the index for the particular village to which the household belonged was taken. The components of the EII are shown below (Table 16).

Table 16 Components of Economic Infrastructure Index

Classification of shocks (by nature of shocks)

Health shock	Economic shock	Natural shock	Social shock
Death of working members	Job loss	Flood	Theft/robbery
Death of Family members (other than working members)	Decrease in sales	Storm	Divorce/family Break-up
Illness of working members	Death of livestock	Fire/house burnt	Strikes, Lockouts
Illness of family members (other than working members)	Increase in input prices	Animal attack
	Asset and crop loss	Earthquake
	Bankruptcy/monetary loss	Droughts

Econometric specification of regression models

I. Probit model for determining factors influencing exposure to specific types of shocks:

The probit regression(s) fitted to assess the effect of household characteristics and locational factors on the probability of exposure to specific shocks is

$${y}_{1}^{*}=X\beta +\mu$$

(1)

where ${y}_{1j}^{*}$ is the latent variable corresponding to the actual, observed outcome ${y}_{1j}$ such that

$${y}_{ij}=1\; \left({\mathrm{the\; house\;hold \;experienced\; the\; }} j{\mathrm{th \;shock}}\right)\; {\mathrm{if}}\; {y}_{1j}^{*}=X\beta +\mu >0$$

$$=0\; ({\mathrm{house\;hold \;did \;not \;experience\; the\; }} j{\mathrm{th\; shock}})\; {\mathrm{if}}\; {y}_{1j}^{*}=X\beta +\mu \le 0$$

X = matrix of regressors and μ is the stochastic error term.

ii. Ordered logit model (OLM) determining factors influencing severity of the primary shock:

The specification of the OLM is

$${y}^{*}=X\eta +\varepsilon$$

where ${y}^{*}$ is the latent dependent variable corresponding to the observed variable y measuring the severity of the primary shock, η is the vector of coefficients, X is the vector of independent variables, and ε is the stochastic error term such that

$$y=1 ({\mathrm{primary\, shock\, is\, less\, severe}}) {\mathrm{ if }}-\infty \le X\eta +\varepsilon \le\mu 1$$

$$y=2\; \left({\mathrm{primary\; shock\; is \; moderately \;severe}}\right)\; {\mathrm{ if }}\; \mu1\le X\eta +\varepsilon \le\mu 2$$

$$y=3\; \left({\mathrm{primary\; shock\; is \;highly\; severe}}\right) \;{\mathrm{if}}\; \mu 2\le X\eta +\varepsilon \le \infty$$

where the threshold values μ₁ and μ₂ are unknown parameters to be estimated.

iii. Poisson model for determining factors influencing exposure to multiple shocks:

The Poisson model is

$${{y}}={{\exp}}(X\nu +\Omega )$$

where y is the vector showing the number of shocks faced by the households and X is the matrix of regressors, ν is the vector of coefficients and Ω is the stochastic error term.

iv. Probit model for determining factors that influence choice of coping strategy:

The model is

$${y}_{2}^{*}=X\lambda +\theta$$

(3)

where ${y}_{2}^{*}$ is the latent outcome variable which captures the type of coping strategy used by the households in the event of the primary shock, X is a set of household and other characteristics, λ is the vector of coefficients, and θ is the stochastic error term. The observed outcome variable is

$${\text{y}}2{{i}} = 1({\mathrm{household adopts good coping}}) {\text{if}} {y}_{2}^{*}=X\lambda +\theta >0$$

$$=0 ({\mathrm{household adopts bad coping}}) {\text{if}} {y}_{2}^{*}=X\lambda +\theta \le 0$$