Why do Taxes and not Windfalls Motivate Political Action?

While the microfoundational proposition underpinning the tax mechanism suggests that taxes give citizens stronger incentives than windfalls to monitor, participate, and sanction, we still know little about why this would be the case. Implicit in the literature are three distinct aspects of revenue that might drive motivation at the individual level: the tax burden, the share of taxes (relative to windfalls) in total revenue, and the level of taxes (relative to windfalls) vis-á-vis spending.

Whereas windfalls are often described as “manna from heaven,” taxes could motivate political action because they require a contribution on the part of citizens. One reason why taxes provoke more political action than windfalls might simply be a distaste for higher tax burdens (taxes as a share of income) in the absolute (Ross Reference Ross2004). Another possibility is that paying taxes makes citizens care more about $1 of misspent tax revenue than $1 of misspent windfall revenue. Research on the endowment effect in psychology and behavioral economics shows that people are more averse to out-of-pocket losses than to foregone gains of an equivalent value (Kahneman, Knetsch, and Thaler Reference Kahneman, Knetsch and Thaler1991). As Sandbu (Reference Sandbu2006) explains,

While the tax burden could mobilize citizens by provoking anger or ownership over the budget, citizens could alternatively be motivated by the share of total government revenue that comes from taxes relative to windfalls. The share of taxes in total revenue has traditionally been regarded as an indicator of the state's fiscal dependence on society, and “the higher that dependency, the more likely it is that governments will have to listen to their citizens” (Waterbury Reference Waterbury and Entelis1997, 157). In contrast, rentier states are characterized by a large share of budget revenue from windfalls and a low degree of fiscal dependence on society. If citizens are more likely to succeed in eliciting better governance as fiscal dependence increases, then this could be reflected in feelings of efficacy and empowerment.

Finally, as noted in the Introduction, it is widely believed that windfalls give leaders the resources to buy political support through patronage or goods at little or no cost to citizens. The corollary claim is that citizens are motivated to take action not by higher taxes in the absolute but by the “price” of public goods. As Ross (2004) states: “High taxes would not produce greater demand for representative government if the taxes were offset by greater benefits. Nor would a small tax bill necessarily lead to political quiescence. Both the size of the tax burden, and the quality and quantity of government spending matter; citizens ultimately care about the ‘price’ they pay for the government services they receive.” As the ratio of taxes to spending rises—and perhaps especially when taxes exceed the benefits provided in return—taxation becomes increasingly perceived as unfair. This could motivate action in a way that windfalls—which keep the price of public goods low—could not.

Revenue and the Link between Information and Action

Predictions on why different sources of revenue affect political action also have important implications for understanding the relationship between information and accountability. To clarify how the motivation and information microfoundation stories relate, I consider how different sources of government revenue could affect the ways in which citizens become politically informed.

One possibility, captured in H1 on monitoring, is that taxation motivates citizens to acquire more information, either by seeking it out or by demanding transparency concessions from government. For many—especially in developing countries where access to information is low—the costs of information are high and contribute to political ignorance (Downs Reference Downs1957). If paying taxes increases the perceived benefits of acquiring information, then taxation could help to explain when a better information environment emerges and why windfall-dependent governments tend to be so opaque (Ross Reference Ross2012). This is important because few studies have identified the origins of fiscal transparency.Footnote ⁶ Another implication is that, if taxation mitigates the resource curse by motivating information acquisition, transparency could be an alternative to taxation.

Another possibility is that revenue conditions how individuals process information on government performance. If taxation induces an endowment effect and makes citizens care more about waste, then learning about misspent revenue should provoke a stronger political reaction in a tax environment. Or, if citizens are sensitive to the “price” of public goods, then information revealing that citizens get less for their taxes should motivate political action. In other words, information that reveals bad spending should be more likely to provoke political action in a tax than a windfall environment.Footnote ⁷ If this were true, then where a government gets its revenue from could determine the extent to which information is effective at getting citizens to hold politicians accountable. Importantly, this would suggest that improving access to information would not be a substitute for taxation in combatting the detrimental effects of windfalls on political action. To test whether revenue conditions how citizens evaluate information on government spending, I add the final hypothesis:

1. H4: The misuse of public funds provokes more monitoring, participation, and sanctioning when citizens pay taxes.

For a graphical summary of the four main hypotheses, see Appendix A online.Footnote ⁸

The Experiments

To test the hypotheses, revenue and information experiments were overlapped in a cross-cutting design and embedded in a public awareness campaign conducted one-on-one by canvassers in participants’ homes. The basic campaign, which all participants received, was designed to familiarize citizens in Blora with the concept of the district budget and the role of elected officials in managing it. The campaign began with illustrations reminding participants that they vote for their representatives, that their representatives make decisions about public services, and that they should think about the future of the district when making political decisions (see the complete campaign script and illustrations in Appendix B online).

All participants then took part in a household budgeting exercise. Upon giving consent, participants earned a small income in exchange for their time, equivalent to about a half a day's wage for an agricultural laborer. Specifically, the windfall group earned 10,000 rupiah (about $1) and the tax group earned 14,000 rupiah (about $1.40). The canvasser then asked participants to take part in an exercise in which they used this income and a household budget game-board to show how they planned to allocate their income across different common household expenditures. This exercise helped to familiarize participants with the concept of a budget while setting the stage for the revenue experiment that followed.

Revenue Experiment

The canvasser next introduced the district government budget, explaining that just as households get income that they allocate across expenditures, so does the government. In the windfall treatment, participants were asked to return their 10,000 rupiah in income to their pockets and the canvasser took out a separate 10,000 rupiah representing district government income per capita. Using the windfall board to denote district government revenue (Figure 1, top), the canvasser divided the amount into 8,000 rupiah representing windfall revenue, primarily from central transfers and natural resources, and 2,000 rupiah representing revenue from other, unspecified sources.

FIGURE 1.

Experiment Game-Boards

Note: The windfall game board (top); the tax game board (center); and the information experiment spending game board (bottom).

In the tax treatment, rather than asking participants to return their income to their pockets, the canvasser reminded them that they (or citizens like them) often have to pay taxes that go to the district government. To represent these tax obligations, the canvasser permanently transferred 4,000 of the 14,000 rupiah income from the participant's household budget game-board to the tax board representing district government revenue (Figure 1, middle). The canvasser then added 6,000 rupiah to the tax board from a separate pot, representing government income from other sources. Both windfall and tax groups thus ended with the equivalent of 10,000 rupiah on the board representing government income per capita.Footnote ¹¹

A Note on Design

The design merits some discussion as one of the first attempts to use an experiment to compare the effects of windfalls and taxes on citizen political behavior.Footnote ¹⁴ Importantly, the experiment aims to test whether political behavior varies depending on whether citizens view government as managing “their taxes,” defined as tax revenue from citizens living within the jurisdiction, or “windfalls,” defined as natural resource revenue and external revenue in central transfers. In defining transfers as windfalls, I follow on Gervasoni (Reference Gervasoni2010), who argues that central transfers are windfalls for poor subnational governments because they are far bigger than what they could obtain through their own taxing efforts. In other words, transfers produce “fiscal rents” when they contain a substantial amount of tax or nontax revenue that originates outside the jurisdiction. This is consistent with discourse in Indonesia, which suggests that district governments exhibit poor spending performance and accountability because they rely on “other people's money” in transfers rather than on locally generated tax revenue (Auracher and Sitepu Reference Auracher and Sitepu2011). Since this indicates that Indonesians plausibly view transfers as windfalls, the experiment was designed to test whether citizens are more motivated to take action when district revenue comes from “their” taxes rather than from natural resources and external revenue in central transfers.Footnote ¹⁵

In doing so, the experimental design sought to employ the strengths of a “lab-in-the-field” study by marrying a high degree of researcher control with a more naturalistic context than a pure lab experiment would provide (Morton and Williams Reference Morton and Williams2010). One main advantage of the revenue experiment is that it isolated taxation's hypothesized incentive effect from its possible information effect by equalizing the information held by both the windfall and tax groups. Both the windfall and tax treatments presented public finances in per capita terms and discussed revenue and spending as proportions of the total budget, not in terms of magnitudes, which are difficult to comprehend. Both treatments also minimized fiscal illusion by establishing clear linkages between revenue, spending, and the district government. These design features should increase confidence that any effect from the tax treatment derives from stronger incentives, not better information.

Another important feature of the design is that it varied the three distinct and theoretically motivated aspects of revenue discussed in the section on theory. Table 1 illustrates how the revenue and information experiments altered the tax burden, the perceived share of local taxes (relative to windfalls) in total revenue, and the perceived ratio of local taxes (relative to windfalls) to spending. By shifting these in a controlled way, the design gains purchase not only on whether taxes motivate more action than windfalls, but also why. While it was not feasible to vary each aspect independently, I present survey evidence below to shed light on which drives results.

TABLE 1.

Detailed Summary of the Treatments

Notes: The panels present details of the three aspects of revenue varied exogenously by the experiments. Panel A shows the tax burden (the ratio of taxes paid to income earned in the experiment). Panel B depicts the shift in the perceived share of taxes in total revenue. The assumption (later verified) is that the windfall treatment caused participants to believe that locally generated revenue constituted ≤20 percent of the district budget. Panel C shows the effect of the shift in priors induced by the information treatment, which revealed that the government spent 30 percent of total revenue on citizens. Panel D captures the “price” of public goods, obtained by putting the numerator in Panel B over the numerator in Panel C.

The tax treatment primed participants to think of the district government as managing “their” money by simulating a 4,000 rupiah tax payment (panel A of Table 1). The design strove for realism in imitating the core components of an actual income tax, conceptualized as a mandated transfer from an individuals’ earned income to the government at at an exogenous rate.Footnote ¹⁶ To underscore that the simulated tax was on earned income, canvassers emphasized at the outset that the payment was in exchange for the participants’ time and effort and the household budget exercise was designed to deepen perceived ownership over the income. The role of the canvasser can be seen as akin to that of an employer withholding income. Reassuringly, when participants were asked during the pilot how they felt about paying the simulated tax, a common response was, “biasa saja,” implying “same as [I] always feel” (when [I] pay taxes). Such a response is consistent with the fact that many citizens in Blora have experience paying taxes.Footnote ¹⁷

Second, the revenue experiment primed participants to think of the share of “their” taxes or of windfalls in the district budget. In particular, the tax group was primed on the share of taxes in total revenue that came from citizens in Blora and was shown a revenue composition in which 40 percent of total revenue came from taxes originating in the district. In contrast, the windfall group was shown that 80 percent of local government income came from central transfers and natural resources (see panel B of Table 1).Footnote ¹⁸ To prime but not deceive, participants in both groups were informed that this revenue composition of the budget was hypothetical and illustrative only.

Finally, the revenue and information experiments jointly affected the perceived “price” of public goods, defined here as the ratio of local tax revenue to spending on direct programs and services for citizens in the budget. As Ross (Reference Ross2004, 247) notes, “A rise in the price of government services can either take the form of a rise in taxes for a constant set of government services, or a constant level of taxes with a cut in government services.” While the tax experiment altered the numerator of that ratio as just described, the information experiment altered the denominator. Specifically, the spending experiment informed participants that 30 percent of the budget was spent on citizens, whereas beliefs in the low information group are participants’ priors (panel C). Panel D shows how the “price” of public goods changes across the experimental conditions. When taxes increase (relative to windfalls) and true spending decreases vis-á-vis prior beliefs on average, it is straightforward to see how the price of public goods rises across experimental conditions.Footnote ¹⁹

Sampling and Randomization

Participants in the campaign were randomly sampled using multistage cluster sampling from the adult population in Blora. The target population was all individuals between the ages of 17 (the voter eligibility age) and 65 who had resided in Blora for at least six months. First, 93 of 295 villages were randomly sampled within strata formed by subdistrict and urban-rural status. Within each village, one subvillage unit (dusun) was randomly sampled, followed by the random selection of 20 households from an updated list of all households in the dusun. Canvassers then sampled one participant in each household on arrival using simple random sampling from a full list of eligible household members made in consultation with a head of household. Random assignment was blocked at the village level so that, of the 20 participants per village, five were randomly assigned to each of the four experimental conditions. A number of steps were also taken to minimize design, canvasser, and timing effects, including blocking treatment assignment by canvassar and randomizing the order in which villages were visited. Evidence from a one-way anova of baseline survey data shows that, as expected, randomization produced a balance in pretreatment covariates across the four experimental conditions (see Appendix E online).

Data

Data come from two sources: Participation in a postcard campaign and an original survey. Survey data were collected by canvassers during the same visit. Upon consent, canvassers implemented a pretreatment module that inquired into individual and household characteristics, such as demographics, public goods usage, political participation, and experience with taxation. Post-treatment modules were implemented following both the revenue and information experiments to provide measures of the main outcomes as well as of underlying attitudinal change. The survey was identical for all participants. The sequencing for the campaign and survey is available in Appendix C online.

The concern with surveys is that they do not provide reliable measures of political behavior because responses are costless and subject to social desirability bias. I therefore use participation in a postcard campaign as a revealed preference measure of political behavior. At the end of the campaign, canvassers gave all participants the opportunity to return postcards. The postcard asked participants to indicate whether they were “satisfied with the district government in Blora and don't want to change anything about how it works” (a reward for good performance) or whether they “want the district government in Blora to do a better job” (a sanction for bad performance) (The postcard can be seen in Appendix D online). Participants were informed that the results of the postcard campaign would be shared with candidates in the lead up to the district head elections to encourage them to respond to public opinion.Footnote ²⁰

Returning the postcard intentionally entailed a small cost. On average, participants lived about a 10-minute walk from a specially designated mailbox located in their community and they had up to 24 hours to deposit their postcards. This small cost is akin to various types of low cost political action, like signing a petition or contacting an official. The postcard campaign also resembles the voting decision in that participants had to decide both whether to turn out and whether to use their postcards to reward or punish the incumbent government. Descriptive statistics for the main outcome measures are available in Appendix F online.

Estimation and Hypothesis Testing

An unbiased estimate of the average treatment effect (ATE) can be obtained using the difference in mean outcomes for the tax and windfall groups as well as for the information treatment and control groups. To estimate the interaction between taxes and information, I take the difference-in-difference effect of the revenue and information treatments. The estimate of the ATE is based on treatment assignment rather than actual treatment status.Footnote ²¹

For the variance of the estimate of the ATE, I use standard errors developed by Neyman (Reference Neyman1923). This approach is conservative (produces bigger standard errors) because it assumes the (unobserved) covariance between potential outcomes at the individual level is zero.Footnote ²² For the main tests of the hypotheses, I use randomization inference p values from a Fisher exact test.Footnote ²³ Randomization inference p values require no modelling assumptions.

Additionally, because I have several outcomes and measures of each outcome, I employ a multiple inference correction for all tests of the main effects of taxes versus windfalls. I use the two-step procedure for correcting the false discovery rate (FDR), or the proportion of null hypotheses that are falsely rejected (Anderson Reference Anderson2008). Analogous to p values, FDR q values are the smallest ratio of falsely rejected null hypotheses to total rejected null hypotheses at which the hypothesis would be rejected. Additional robustness tests and validity checks for all results are available in Appendixes G–K online.

Does Revenue Motivate Political Action?

I begin by evaluating the effects of the windfall and tax treatments on the three main outcomes of interest: monitoring (H1), participation (H2), and incumbent sanctioning (H3). Table 2 reports the effects of the tax treatment for the full sample using the main measures for each outcome from both the survey and postcard campaign.

TABLE 2.

The Effect of the Windfall and Tax Treatments on Political Action (H1–H3)

Notes: Table reports (i) the mean in the windfall group, (ii) the average effect of the tax treatment, (iii) Neyman standard errors, (iv) the randomization inference p value, (v) FDR q values adjusting for multiple inference, (vi) sample size. Significance levels are denoted by *p < 0.10, **p < 0.05, and ***p < 0.01.

^aHow interested are you in learning more about how the district government spends money in the budget? (Interested = 1.)

^bHow interested are you in learning more about what the government of Blora is doing? (Interested = 1.)

^cYou should pay more attention to what the district government does. (Agree = 1.)

^dRegarding a problem or issue that was affecting your daily life or your community, would you in the future: contact a village or subdistrict official, contact the district head, contact a local legislator, contact the media or an NGO, take part in a demonstration? (Average of 5.)

^eReturned postcard (1), abstained (0).

^fWhich [candidate] would you say has your strongest support at present [in the upcoming district head elections]? (Incumbent = 1.)

^gWhich [candidate] would you say has your strongest support at present [in the upcoming district head elections]? (Challenger and former legislative chairman = 1).

^hReturned postcard and sanctioned incumbent (1), returned postcard and rewarded incumbent (−1), abstained from returning postcard (0).

I look first at how the revenue treatments affected participant willingness to monitor politicians, proxied by three survey questions reported in panel A of Table 2. The first question inquires into the willingness to learn more in the future about how the district government manages public funds. The tax group was five percentage points more willing to monitor the budget than the windfall group. This result is significant at the 99 percent confidence level (p = 0.003) and is robust to controlling for multiple inference. I also confirm that the finding is driven by a shift upward in the tax group rather than a shift downward of the windfall group. In the pretreatment survey, 71 percent of all participants wanted to learn more about the budget; the revenue experiment increased willingness to monitor in both the windfall and tax groups but more so in the latter. Substantively, the magnitude of the treatment effect is about half the size of having a primary school education; those who had completed primary school were 11 percentage points more willing to monitor the budget.

The following two questions in panel A ask about interest in monitoring government more broadly. There is weak evidence that the tax treatment caused a three percentage point increase in the proportion of partipants willing to monitor government more generally (row 2), although the more abstract question on whether participants agree they should “pay more attention” to what the district government is doing reveals no effect (row 3). Overall, the fact that the tax treatment had a positive effect on demand for information on the budget, if not government more broadly, supports H1.

There is no indication that the tax treatment increased willingness to engage in other forms of political participation, however. I evaluate the effect of the tax treatment on political participation (H2) using data from both the survey and the postcard campaign. To assess whether the treatments made participants more willing to take action to address issues affecting their lives or their communities, the survey asked about five types of nonelectoral political behavior. These included contacting a village or subdistrict official, the district head, a local legislator, and the media or an NGO. Participants were also asked if they would be likely to take part in a demonstration. The results, reported in panel B of Table 2, show no difference between windfall and tax groups in the mean number of types of political action participants were willing to take.Footnote ²⁴

Participation in the postcard campaign, as a revealed preference, provides a better measure of actual willingness to take political action. As can be seen in row 5 of panel B, however, the tax treatment did not cause higher levels of participation in the postcard campaign. While 78 percent of all participants returned their postcards, there is no detectable variation in return rates across experimental conditions.

The third outcome of interest is whether the tax treatment reduced support for the incumbent government (H3). Panel C of Table 2 presents survey data on self-reported vote choice as well as the overall sanctioning effect of the postcard campaign. The public awareness campaign took place in the three months leading up to district head elections. The three candidates for district head included the incumbent, the former chairman of the district legislature, and a third challenger with ties to the bureaucracy and a popular Muslim organization. As described earlier, the former legislative chairman was widely suspected of corruption at the time. Participants were asked which candidate they most supported at the time the survey was conducted.

While low statistical power makes it difficult to detect significant differences across groups, the signs on the treatment effects in Table 2 suggest that the tax treatment produced higher support for the incumbent district head and lower support for the former legislative chairman.Footnote ²⁵ One plausible explanation is that the tax treatment caused participants to prefer the incumbent district head relative to the challenger given the latter was viewed as more corrupt.Footnote ²⁶

The postcard campaign again provides the main revealed preference measure of willingness to take action to demand better performance from government. Whereas the previous measures capture support for the incumbent district head relative to a challenger, the postcard more closely resembles a referendum on incumbent performance. To provide a composite “net sanctioning effect” measure, I code all those who returned their postcard and signaled that they want government to do better as 1, those who returned their postcard and said they were satisfied with the status quo as −1, and those who abstained from participating as 0. The final row of Table 2 provides little evidence that the tax treatment produced a greater impetus to sanction the incumbent than windfalls. Yet, as will be shown in the next section, while the tax treatment had no effect in the high information environment, it caused a six percentage point increase in the use of the postcard to sanction the incumbent in the low information environment. The results below provide evidence for H3 while also pointing to an unanticipated relationship between revenue and information.

Does Revenue Condition the Impact of Information?

The results presented so far suggest that the tax treatment increased demand for information on the budget, consistent with the notion that taxation promotes information-seeking. Hypothesis 4 predicts another way in which revenue affects information: by conditioning how people evaluate information on government spending. To investigate whether citizens care more about misspent public funds when they pay taxes, I focus on whether the impact of the information treatment depends on whether participants were in the windfall or tax treatment groups.

In order to study the effect of information, the information treatment had to shift participant beliefs about government spending. To check first that this occurred, I use data from a district budget exercise conducted with participants just prior to the information experiment to measure participants’ initial ideals and beliefs about public spending. All participants were asked to the use the 10,000 rupiah from the revenue experiment (and the district budget game-board) to illustrate how they would spend the funds if they were the decision-maker (ideals) and how they think government actually spends the funds (beliefs). As can be seen in Figure 2, while actual spending on education was far greater than what participants wanted or expected, spending was lower than desired or expected for health, infrastructure, and farming. In general, participants estimated that government spent about 55 percent of its revenue on public services and direct programs for them; the campaign revealed the true amount in 2008 to be 30 percent.

FIGURE 2.

Spending Priors

Notes: Figure compares actual spending (as revealed by the treatment) to prior ideals and beliefs about government spending across the six categories highlighted in the information campaign. The two columns at the far right show the initial beliefs and the truth about the total share of the budget that government spent on all direct goods and services for citizens.

What participants learned made them view district government performance in a substantially worse light. Figure 3 presents results for three survey measures of attitudes towards government. The information treatment caused a 43 percentage point increase in the share of participants who felt government was doing worse than expected in the windfall group and a 45 percentage point increase in the tax group. Both of these effects are strongly significant. Yet, contrary to expectations, there is no evidence that the tax group reacted more strongly to the negative information. Questions on dissatisfaction with budget management (panel B) and distrust in the district head (panel C) reveal similar results; the pattern also holds for additional measures on dissatisfaction with district government overall and distrust in legislators (for tables of all results, see Appendix G online). These findings suggest that, while the information was indeed negative and caused participants to update beliefs, citizens in the tax treatment did not care more about misspent revenue.

FIGURE 3.

How the Revenue Treatments Condition the Impact of Information on Attitudes Towards Government

Notes: Panel A shows the proportion who believe that “elected leaders in the district government are doing a worse job than [you] thought they were”; panel B is the proportion dissatisfied “with the way the district government manages the budget” (panel B); and panel C is the proportion that distrusts “the district head to do the right thing for the people of Blora” as a result of the treatments. Significance is denoted by *p < 0.10, **p < 0.05, and ***p < 0.01 based on randomization inference p values.

I proceed to assess the main evidence for H4: That information provokes more monitoring, participation, and sanctioning when citizens pay taxes. To do this, I revisit the main outcomes presented above but now with attention to whether there was an interaction effect in the revenue and information experiments. I focus here on the main measure of each outcome.

I begin by examining whether the impact of the information treatment on willingness to monitor the budget was greater for the tax than the windfall treatment group. The results, shown in panel A of Figure 4, indicate that the information produced a seven percentage point increase in willingness to monitor the budget in the windfall group (p = 0.006) and an eight percentage point increase in the tax group (p = 0.001). Similar to the findings above, however, since the magnitudes of these effects are approximately the same, there is no support for the prediction that revelations of misspent funds provokes a stronger desire among taxpayers to monitor government in the future.

FIGURE 4.

How the Revenue Treatments Condition the Impact of Information on Political Action (H4)

Notes: Panel A is the proportion willing to monitor the budget; panel B is the proportion that participated in the postcard campaign; and panel C is the proportion that used the postcard to sanction the incumbent as a result of the treatments. All measures are defined in Table 2. Significance is denoted by *p < 0.10, **p < 0.05, and ***p < 0.01 based on randomization inference p values.

Next, I evaluate whether the effects of the information treatment on more overt types of political participation vary across windfall and tax groups. Notably, while we know that the information caused participants to take a more negative view towards government, panel B of Figure 4 indicates that this change in beliefs did not translate into higher participation in the postcard campaign for either the windfall or tax groups. This result further undermines support for H4.

Finally, I consider whether participants in the tax treatment were more likely than citizens in the windfall treatment to punish the incumbent. Panel C of Figure 4 presents the results for the net sanctioning effect of the postcard campaign. Whereas H4 predicts a significant positive interaction between the tax and information treatments, there is actually an unanticipated and significant negative interaction (p = 0.023). This occurs because the tax treatment caused a six percentage point increase in the willingness to use the postcard to sanction the incumbent but only in the low information environment (p = 0.041). Moreover, the information treatment caused a seven percentage point increase in incumbent sanctioning but only in the windfall environment (p = 0.014).Footnote ²⁷ The information treatment, however, had no significant effect on subjects in the tax treatment.

Two notable implications arise from these findings. First, they provide support for H3 in that the tax treatment clearly produced more incumbent sanctioning than the windfall treatment in a low information environment. Second, they suggest that participants sanctioned the incumbent either because of the tax treatment or information treatment but the benefit of receiving both was limited; this could be interpreted to mean that there is possibly a ceiling in the effect of the treatments on political action.Footnote ²⁸ All in all, while the data show three different patterns in how revenue and information relate, there is little evidence to support the idea that information provokes more political action when people pay taxes.

Summary of Main Results

In showing that the tax treatment caused a greater willingness to monitor the budget compared to the windfall treatment, the main results can be interpreted as support for H1. Evidence for H3 is also provided in showing that the tax treatment motivated participants in the low information environment to use their postcards to sanction the incumbent. There is less support for the hypothesis that different sources of government revenue create a differential willingness to participate in politics, as predicted by H2. One possible explanation for this pattern of results is that the tax treatment increased the perceived benefits of restraining government for participants but not sufficiently to exceed the perceived costs of action. This is consistent with the notion that citizens are most likely to protest taxes only when they become excessive (Scott Reference Scott1976). Interestingly, the fact that the tax treatment had a positive effect on monitoring but not on more overt types of political mobilization suggests a variation in how individuals assess the costs and benefits of different types of political engagement.

The findings also shed light on the possible relationship between taxation and information. In showing that the tax treatment made participants more willing to learn about the budget, it supports the notion that taxation enhances the demand for information. As such, the results provide microlevel corollary evidence for cross-national findings that governments that depend on windfall revenue tend to be less transparent (Ross Reference Ross2012). With respect to the effect of revenue on information processing, both the tax and windfall groups proved equally intolerant of revelations that public funds were not spent as well as expected. Contrary to H4, the results point to the effects of information on political action being as great, if not greater, in the windfall treatment.