Table 7.
Estimates of the effect of open defecation on child height from the literature, and implications for the height gap.
| context | identification | coefficient | 95% CI | % DHS gap | % JMP gap | |
|---|---|---|---|---|---|---|
| randomized experiments | ||||||
| Gertler et al. (2015) | India, Indonesia, Mali | IV from three RCTs | −0.46 | (-0.772, −0.148) | 98 | 85 |
| Clasen et al. (2014)a | India | IV from RCT, see note | 0.351 | (-0.460, 1.163) | ⋅ | ⋅ |
| difference in trends | ||||||
| Headey (2015) | Ethiopia | region and time FEs | −0.31 | (-0.53, −0.09) | 66 | 57 |
| Vyas et al. (2016) | Cambodia | province and time FEs | −0.502 | (-0.684, −0.320) | 107 | 93 |
| Spears (2013) | India | district and time FEs | −0.553 | (-0.798, −0.308) | 118 | 102 |
| cross-sectional | ||||||
| Jayachandran and Pande (2013) | India and Africa | cross-section, OD | −0.358 | (-0.382, −0.334) | 76 | 66 |
| Lin et al. (2013) | Bangladesh | cross-section, WASH | −0.54 | (-1.01, −0.06) | 115 | 100 |
| meta-estimates | ||||||
| all seven studies | −0.360 | (-0.384, −0.337) | 77 | 67 | ||
| excluding Jayachandran and Pande (2013) | −0.421 | (-0.550, −0.292) | 90 | 78 | ||
| excluding both cross-sectional and Clasen et al. based IV | −0.436 | (-0.572, −0.299) | 92 | 80 | ||
Note: Coefficients are, unless otherwise noted, coefficients on the local area fraction of households defecating in the open, predicting child height-for-age. The Lin et al. (2013) estimate compares households with dichotomized extreme sanitation environments, in data extracted from an experimental project. The Spears (2013) estimate is taken from Table 6 of section 3 of the World Bank Policy Research Working Paper version of this paper. The Jayachandran and Pande (J & P) (2013) estimate compares households with and without latrines in the child level dataset used in section 4 and is further discussed in supplementary appendix section B.1; it is excluded from some meta estimates because its very small standard errors dominate the averaging. The meta estimate weights estimates by the square of the inverse of their standard error. To explain the India-Africa height gap, the coefficient on local open defecation must be 0.54 using open defecation figures from the Unicef-WHO Joint Monitoring Programme or 0.47 using the same DHS data compilation used in section 4. None of these estimates consider the interaction of open defecation with population density.
The Clasen et al. (2014) RCT in Orissa, India did not find a large first stage effect on open defecation, and therefore did not find a statistically significant effect on child height. Quoted here with permission are preliminary results from an in-progress reanalysis of the data by Scovronick et al. to produce an IV estimate and confidence interval of the implied effect of local open defecation on child height; the confidence interval is large.