Hermann Pythagore Pierre DonfouetCREM, UMR CNRS 6211University of Rennes I
Pierre Wilner JeantyKinder Institute for Urban Research
Rice University
Eric MalinCREM, UMR CNRS 6211
University of Rennes I, France
Accounting for Spatial Interactions in the Demand for Community-Based
Health Insurance: A Bayesian Spatial Tobit Analysis
Presentation outline
Introduction
Methodology
Survey design and data
Results
Conclusion
IntroductionLow-income households lack health insurance and
adequate health care services.
CBHI has been recognised as an efficient mechanism to finance the need for healthcare of the low-income households in developing countries (DC).
CBHI is a kind of insurance which is designed for low-income households who are totally excluded from formal insurance.The demand aspect of CBHI is important to policymakers.
Many studies conducted in DCs had revealed that the low-income households are willing to pay for CBHI (Ataguba et al., 2008; Bärnighausen et al., 2007; Dong et al., 2004; Dong et al., 2004b; Dror et al., 2007; Wang et al., 2005)
Introduction (cont.)
Lack of spatial interactions in the previous studies.
Spatial dependence can be ascribed to the situation where observations on the dependent variable (or the error term) at one location is correlated with observations of the dependent variable (or the error term) at other locations.
To the best of our knowledge, no previous studies have examined the factors determining the demand for CBHI while allowing for the spatial interactions. This present study is an attempt to fill this void.
MethodologyThe contingent valuation method (CVM) was used to
assess the demand for CBHI.
Elicitation format used:Closed-ended questionOpen-ended question
Spatial interaction was integrated in the two elicitation formats by defining a social network spatial weights matrix (Anselin and Bera, 1998) as follows: households are neighbors if they live in the same village.
Testing the existence of spatial interactions Closed-ended question Open-ended question : OLSTobit
Will you be willing pay X $? 1. Yes 2. No
Moran tests (Moran's_I)
What is your maximun amount ? __$
Spatial autoregressive Probit (Bayesian approach)
SAR or SEMBayesian spatial Tobit (SARBT)
Survey design and dataGovernment in Cameroon (Central Africa):
40% coverage with CBHI by 20156 villages in Bandjoun (November 2009)
by a two-stage sampling
Face-to-face interviews sponsored by the International Labour Organisation (ILO)
The most important part of the CV survey was the scenario
CBHI and their benefit was presented to the head of the households.
The monthly premium that they must pay.
ResultsTable 1: Description of the variables and summary statistics Variable Description Mean Standard
deviation Min Max
Logbid Logarithm of the first bid 6.15 0.40 5.52 6.68
Loginc Logarithm of income of the respondent
10.02 0.97 8.92 12.22
Age Age (number of years) expressed in thousand
0.04 0.01 0.02 0.08
Male Gender of the respondent (1 if the respondent is a male, 0 otherwise)
0.61 0.49 0 1
Healhstate Health status of the respondent (1 if the respondent has a poor or very poor health state, 0 otherwise)
0.26 0.44 0 1
Education Level of education of the respondent (1 if the respondent has been to secondary school, 0 otherwise)
0.24 0.43 0 1
Farmer Profession (1 if the respondent is a farmer/ seller, 0 otherwise)
0.53 0.49 0 1
Meanstreatment The means of seeking treatment when any member of the household falls sick (1 if conventional, 0 otherwise)
0.84 0.38 0 1
Involvement Participation of the respondent in an association (1 if yes, 0 otherwise)
0.52 0.50 0 1
Distance Distance between the house of the household and health public facility in kilometers (continuous)
1.60 1.26 0.01 8
Results (cont.)Table 2: Results of the regression model without spatial interactions Variables OLS Probit Tobit Loginc 0.06 0.09 -0.71 (0.04) (0.09) (52.60) Age 0.10 5.23 2304.99 (3.34) (6.86) (4169.18) Male -0.10 0.16 -138.62 (0.08) (0.17) (106.27) Healhstate 0.17* 0.16 199.43* (0.09) (0.18) (116.27) Education 0.20** -0.11 258.41** (0.10) (0.18) (118.53) Farmer -0.12 -0.04 -198.17** (0.08) (0.16) (100.60) Distance 0.01 0.01 59.67 (0.03) (0.07) (39.86) Meanstreatment -0.11 0.41** -130.70 (0.11) (0.20) (136.10) Involvement 0.12 0.49*** 128.71 (0.08) (0.16) (101.82) Logbid -0.60*** (0.20) Intercept 6.07*** 2.75* 825.32 (0.46) (1.50) (574.20) Notes: ***, **, and * significant at 1, 5 and 10% respectively. Standard errors are in brackets.
Results (cont.)Table 3: Diagnostic test for spatial interactions Tests Statistic P-Value Moran's_I 4.40 0.00*** LM_Error 11.13 0.00*** Robust_LM_Error 1.84 0.18 LM_Lag 13.64 0.00*** Robust_LM_Lag 4.35 0.04** Moran's_I for Probit Moran's_I for Tobit
-0.83 3.34
0.41 0.00***
Notes: ***, **, and * significant at 1, 5 and 10% respectively. Thus, the test reveals that there are spatial interactions in the demand for CBHI mostly for the spatial autoregressive (SAR) and Bayesian spatial Tobit (SARBT).
Results (cont.)Table 4: Spatial autoregressisve and Bayesian spatial Tobit
Variables SAR SARBT Loginc 0.04
(0.04) 0.14 (0.005)**
Age 0.4 (3.26)
3.60 (6.24)
Male -0.08 (0.08)
-0.11 (0.008)
Healhstate 0.16 (0.09)*
0.35 (0.01)**
Education 0.21 (0.09)**
0.45 (0.01)***
Farmer -0.13 (0.07)
-0.11 (0.008)
Meanstreatment -0.08 (0.11)
-0.12 (0.01)
Involvement 0.11 (0.07)
0.11 (0.008)
Distance 0.01 (0.03)
0.08 (0.003)*
Intercept 3.27 (1.12)***
1.08 (0.06)
Rho 0.43 (0.15)***(a)
0.54 (0.007)***(b)
Notes: ***, **, and *significant at 1, 5 and 10% respectively. Standard errors are in brackets. The P-value of (a) is 0.007 and (b) is 0.0002. Thus, (b) is more significant than (a). SAR and SARBT are respectively the spatial autoregressive for the OLS and spatial autoregressive Bayesian Tobit.
ConclusionThe test of spatial interactions in the SARBT
reveals that there are spatial interactions in the demand for CBHI. These spatial interactions thus affect the WTP for CBHI
As provided by table 4, the intensity of the spatial interactions is positive (rho>0) , implying that households buying behavior are strategic complements.
This externality (imitation effects) in the demand for CBHI may be explained by the social norms that rule many rural areas in developing countries.
Policymakers must be conscious that space matters a lot in the demand for CBHI and must take this in account when designing health insurance packages for rural households and their premium as well.
Thank you for listening Your comments are welcomed.
Top Related