Informal Taxation (Web Appendix)

As described in the text, the data used in this paper stems from household surveys from 10 countries around the world, which were selected because they included questions on payments to local public good provision. Since each of the surveys is somewhat di¤erent in terms of sampling and questionnaire design, this section brie‡y describes each of these surveys one by one, including details on the informal tax question, sampling design, and the de…nition of a community used for community …xed e¤ects. Since the derivation of VAT tax rates is somewhat more involved, Section 2 separately discusses the construction of implied VAT rates in each country in more detail. In constructing the data, we use several normalizations to standardize the data. To standardize units of time (e.g, for labor contributions) between surveys, we assume that each “day”worked is equivalent to 6 hours worked, and that there are 260 working days per year. When including “village”…xed e¤ects, we use the smallest geographic unit available in the data. When the smallest geographic unit includes both urban and rural areas, we interact the geographic unit with an urban/rural dummy, so that each “village”is entirely rural or entirely urban. We convert all local currency units to 2000 PPP dollars after obtaining the exchange rate by dividing each country’s GDP in local currency units for that year by the GDP in PPP constant 2000 international dollars. This data is available in the World Bank World Development Indicators. All our surveys (with exception of Ethiopia) provide a household level weight variable to account for the sampling procedure. We used these weights in calculating the descriptive statistics (Table 3), as well as in all our regression analysis.