A central question in development economics is how to fund public goods. Informal taxation, whereby households make direct contributions to local public goods (such as water resources, roads and schools) outside of the formal tax system, is an important source of funding for public goods in many low-income countries, especially Kenya (Olken and Singhal 2011, Ngau 1987, Barkan and Holmquist 1986). Informal taxes are coordinated and collected by local leaders and enforced via social sanctions rather than the state. In a formal tax system, legal statutes dictate how taxes change with household income. But how does informal taxation respond to changes in household income?
My job market paper first quantifies informal taxation in Kenya. Using household panel data, I estimate informal tax schedules over the income distribution and test whether informal taxes respond to changes in earned income. Second, I estimate how informal taxation and public goods respond to a large, one-time increase in income from a randomized unconditional cash transfer program targeting poor households.
While a large literature finds positive benefits from cash transfers for recipient households (Bastagli et al. 2016, Banerjee et al. 2016, Evans and Popova 2017), less attention has been paid to how these programs interact with local public finance. If one-time cash transfers to poor households can lead to an increase in public goods investment, this provides a mechanism for long-term and spillover benefits to non-recipients.
Contribution 1: Quantifying Informal Taxation
The first contribution of my paper is to quantify informal taxation in rural Kenya, where local leaders collect informal taxes via community fundraisers (known as harambees), village meetings, and going door-to-door. Payments are made in cash, labor or in-kind contributions, though most households report paying cash. Households that do not pay are subject to social sanctions such as public announcements, letters, and home visits (Miguel and Gugerty 2005).
I find informal taxation is widespread, with over 40% of households participating in the last 12 months, twice the rate for formal taxes. Both informal tax participation and the amount paid are increasing in household income but declining as a share of household income, implying informal taxes are regressive. In fact, informal taxes are more regressive than formal taxes. Using panel data from control households, I also find that informal taxes respond to changes in earned income: a one decile shift up the earned income distribution is associated with a statistically-significant increase of KES 33, or about 10 percent of the mean informal tax amount. This suggests that leaders are aware of household income changes, and are able (and willing) to change tax payments in response to changes in household income.
Contribution 2: Exploiting Randomized Income Shock
The NGO GiveDirectly provides large, one-time unconditional cash transfers to poor households in rural Kenya meeting a basic means-test. Existing research documents large, positive effects for recipients of GiveDirectly’s program (Haushofer and Shapiro 2016). As part of a study on the general equilibrium effects of cash transfers, close to $11 million in transfers was distributed across 653 villages in rural western Kenya (Haushofer, Miguel, Niehaus and Walker (forthcoming)). For this study, households with a grass-thatched roof, an indicator for poverty, were eligible for the program (33% of households). Treatment was randomly assigned at the village level, and all eligible households in treatment village received a cash transfer, while no households in control villages received transfers. (In addition, treatment intensity was randomized at the sublocation level (the administrative unit above a village, containing an average of 10 villages), with 2/3 of villages in high saturation sublocations assigned to treatment and 1/3 of villages in low saturation sublocations assigned to treatment. I control for this in the analysis). Cash transfer treatment households received a series of 3 payments over 8 months totaling almost $1,000 (nominal) via the mobile money system M-Pesa.
Two aspects of the transfer program are especially notable: first, the magnitude of the transfer is sufficiently large to temporarily shift all cash transfer treatment households above the 90th percentile of the baseline income distribution; the median cash transfer treatment household moves to the 97th percentile. Second, local leaders are informed GiveDirectly is working in their village. The fact a) eligibility is based on observable characteristics of the households and b) all eligible households receive a transfer makes it easy for leaders to infer who has received transfers.
Data comes from surveys I specifically designed to measure informal taxation and public goods of over 7,000 households (including households both eligible and ineligible for the cash transfer) and 800 local leaders. The average endline survey occurred 1.5 years after households received their first transfer.
Main Finding 1: No effects on informal taxes or public goods
I find no significant effect on informal tax amounts and participation for cash transfer treatment households, nor ineligible households in treatment villages. Figure 1, Panel A plots the mean effect for eligible and ineligible households in treatment versus control villages. The observed point estimate of KES 14 for cash transfer treatment households is 0.01 percent of the total transfer value, and significantly less than the predicted change of KES 166 if transfer income was taxed the same as earned income for control households.
Given the informal tax results, it is unsurprising that I find no increase in the number of village public goods projects (Figure 1, Panel B), both overall and for water resources, for which informal taxation is especially important. Notably, while the cash transfers do not appear to increase public goods, they do not have a negative effect.
Main Finding 2: Local leaders exempt the cash transfer from informal taxation
The null effect on informal taxes is surprising, given that i) leaders are aware of transfers and ii) informal taxes respond to changes in earned income. To further explore this, I compare endline informal tax amounts by pre-treatment household income deciles (Figure 2). The gray line plots the informal tax “schedule” by income decile for control village households. As previously noted, cash transfer treatment households were temporarily shifted into the top income decile, so the dotted line represents the expected amount paid if households were taxed inclusive of their cash transfer income.
Actually, cash transfer treatment households pay similar informal tax amounts as control village households in the same pre-treatment income decile (the solid black line). I cannot reject these tax schedules are the same. This is consistent with leaders exempting the transfer income from informal taxes, and suggests that leaders favor equity considerations (as the cash transfers are targeted towards poor households) over the sizable potential public goods investment. That said, the fact even high-income cash transfer treatment households do not pay more informal taxes exacerbates the regressive nature of informal taxation.
Taken together with the fact informal taxation responds to changes in earned income for control households, this provides suggestive evidence informal taxation responds to permanent, rather than temporary, changes in income. However, as the cash transfer is both temporary and targets poor households, I cannot fully disentangle these channels.
Alternatively, leaders may exempt the cash transfer to allow households to make productive investments now in return for larger payments in the future. However, I find no heterogeneity in effects on informal taxes by the number of months since transfers began, though these are all relatively short-run effects (less than 2 years). I also do not find that leaders extract larger contributions for community fundraisers for social insurance, such as funerals and weddings, nor for bribe payments, from cash transfer treatment households
I do find some evidence of “kinship taxation” (Jakiela and Ozier 2016, Squires 2017), as cash transfer treatment households send about 25 percent more in inter-household transfers compared to eligible households in control villages, predominately to other family members. The magnitude of this increase is still less than 1 percent of the cash transfer value.
Implications of Results
My findings suggest that cash transfer treatment households are not overtaxed by local leaders, but I find no increase in public good investment, tempering possible spillover or long-term benefits from one-time cash transfers. Importantly, I do not find negative effects on public good provision, and these findings do not negate the positive benefits to recipient households documented in the literature. To the extent informal taxation is based on permanent rather than temporary income, changes in permanent income are required in order to increase public goods provision via informal taxation.
Michael Walker is a Ph.D. candidate in economics at UC Berkeley.