Source: Authors’ analysis based on data from the Kagera Health and Development Survey (KHDS) and the International Coffee Association.

Note: The price received is in Tanzanian shillings per kg; quantity sold is in kg. Coffee revenues and all expenditure variables are in Tanzanian shillings. “Coffee grower” is a dummy variable that equals 1 if a household reported harvesting coffee in that wave. “Harvest area under coffee” is the number of acres harvested in the last 12 months if the household is surveyed in the first wave, and the number of acres harvested in the last six months if the household is surveyed in any subsequent wave. The sample sizes reflect the number of household-year observations in which the household reports nonmissing values of the dependent variables (e.g., the number of household-year observations in which the household reports having farm acreage under coffee cultivation, observations for which the household reports a specific quantity, revenue, and/or price for coffee harvests, etc.). Columns (1)–(3), (7), and (8) include households who harvested coffee in any of the four waves, and columns (4)–(6) include all households. Expenditures are trimmed at the 1st and 99th percentiles. Untrimmed results are presented in the supplementary online appendix. SD |$=$| standard deviation. Robust standard errors are given in parentheses. Standard errors are clustered at the enumeration cluster level. |${}^{*\!\!}\, p\lt .1$|⁠,  |${}^{**\!\!}\:p\lt .05$|⁠,  |${}^{***\!\!}\:p\lt .01$|⁠.