Narrowing the ideological gap over the federal budget

rust in government is pivotal for gaining support for government policies, especially regarding taxing and spending (Beck, Rainey, & Traut, 1990; Beck & Dye, 1982; Lowery & Sigelman, 1981; Sears & Citrin, 1982; Rudolph, 2009; Hetherington, 2005). This is because citizens are more likely to support an expansion of government and the use of tax money if they believe that their government is trustworthy (Hetherington, 2005). Moreover, taxing and spending are central components of the ideological positions that individuals take (Huckfeldt, Levine, Morgan, & Sprague, 1999; Jacoby, 2000; Jacoby, 1991). Taken together, trust then acts as a heuristic device that provides clear signals to individuals to either support or oppose public policies, which is activated depending on the degree of ideological sacrifice an individual is asked to make (Rudolph, 2009; Rudolph & Evans, 2005; Jacobs & Matthews, 2017). Various forms of the trust-as-heuristic theory have been used to explain both policy preferences and behavioral outcomes across different issue-areas (Scholz & Pinney, 1995; Scholz & Lubell, 1998; Hetherington, 1999; Davis & Silver, 2004). In the contexts of government budgets, scholars have provided some evidence that the gap in ideological differences over the federal budget narrows among individuals with higher levels of trust in government. On the spending side, Rudolph & Evans (2005) argue that trust is needed for conservatives to support spending policies because such policies require greater ideological sacrifice among conservatives than liberals. Using data from the 2000 National Annenberg Election Survey, they find that trust in government is moderated by ideology, with trust mattering more to conservatives. On the revenue side, Rudolph (2009) apply the same theory to explaining attitudes for Bush tax cuts and find that trust in government increases support only T Abstract: Do liberals and conservatives who trust the government have more similar preferences regarding the federal budget than liberals and conservatives who do not? Prior research has shown that the ideological gap over spending increases and tax cuts narrows at high levels of trust in government. We extend this literature by examining whether the dampening effect of trust operates when more difficult budgetary decisions (spending cuts and tax increases) have to be made. Although related, a tax increase demands greater material and ideological sacrifice from individuals than tax cuts. The same logic can be applied to support for spending cuts. We test the trust-as-heuristic hypothesis using measures of revealed budgetary preferences from a population-based survey containing an embedded budget simulation. Our findings show that trusting liberals and conservatives share similar preferences toward spending cuts and tax increases, adding an important empirical addendum to a theory based on sacrificial costs.

amongst liberal respondents.Together, the literature on trust, ideology, and budget policies suggests that trust in government dampens the ideological positions toward budget policies.But such a conclusion may be premature without further testing of the trust-as-heuristic theory in regards to support for tax increases and spending cuts.
In this paper, we ask: Do liberals and conservatives who trust the government have more similar preferences regarding spending cuts and tax increases than those with less trust in government?Although fundamentally linked, voters may interpret increases and decreases to each side of the ledger as distinct debates.Arguably, spending cuts and tax increases are the more politically contentious sides of budget policy because they reduce important services and require voters to chip in more.In the minds of voters, the decision to support tax cuts (spending increases) may be perceived differently than decisions about tax increases (spending cuts) (Franko, Tolbert, & Witko, 2013).Simply put, support for tax increases demands greater material and ideological sacrifice from voters, especially conservative voters, than tax cuts.While tax cuts may be seen as a difficult ideological sacrifice for liberals, the material sacrifice associated with a tax cut is less apparent because the costs are highly diffused.For this reason, Bartels (2005) argues that there was wide support for the Bush tax cuts despite the policy disproportionately benefitting wealthy voters.The same logic can be applied to support for spending cuts, reducing the services provided by the government.Though conservatives make an ideological sacrifice in support of policies that increase spending, it is not necessarily the case that they will personally incur a material sacrifice as a direct result of these policies.Therefore, an examination of this untested extension-determining whether trust dampens the ideological divide over spending cuts and tax increases-serves as an important, and arguably, harder test of the trust-asheuristic theory, a theory that relies on the concept of sacrificial costs.
We test this theoretical extension using newly collected data from an interactive budget simulation in which respondents are asked to submit a budget that they would support.Rather than using general attitudinal questions as measures of budget preferences (e.g.do you support tax increases?),our measures are generated from the behavioral choices that respondents make during the simulation.Our simulation draws upon Bonica (2015) and extends it by providing respondents with both sides of the federal budget.In our simulation, respondents were only given the options to either increase taxes or cuts spending, as a way of forcing them to make difficult choices which would require an ideological and material sacrifice.We can thus test whether liberals and conservatives who trust the government have more similar preferences regarding spending cuts and tax increases than liberals and conservatives who do not.
This paper makes two contributions to the ongoing development in this research program.Theoretically, we extend the trust-as-heuristic perspective to explain attitudes toward the budgetary tradeoff between tax increases and spending cuts.The trust-as-heuristic framework is grounded in individuals making material and ideological sacrifices.However, creating a sense of sacrifice has been a difficult task for researchers, especially when the sacrifice is seemingly abstract and emotionally distant from the participant (i.e. the national budget).In our simulation, we created these sacrificial costs by showing respondents that there was a budgetary deficit and then provided them with hard choices, increasing revenue and/or decreasing spending.Methodologically, we leverage budget simulations as a new way for social science researchers to study and measure budget preferences, as well as tradeoffs, by examining both sides of the budget sideby-side.We believe our measure provides stronger external validity than the traditional survey since respondents are making interactive choices.Finally, this study has important policy implications.As a society, we face a number of challenges, such as an aging population and climate change, which will require governments to make difficult trade-offs that are not always popular with their most ardent supporters.Our study demonstrates, however, that the ideological gap on important budgetary trade-offs narrows among individuals who express trust in the government, which could make addressing these challenges palatable for even the most ideological individuals.

Research Design
Between June 7 and 20, 2017, we conducted a population-based online survey with 1,991 voting-age Americans using a general population-based panel provided by Survey Sampling International (SSI).The survey respondents were first presented with a consent form and asked basic demographic information such age, employment status, gender, and income.
Then, before starting the budget simulation, respondents were given the following two paragraph prompts about the national deficit and instructions on how the budget simulation works: "The U.S. Federal Government is in debt.In 2017 the government is going to spend approximately $519 billion more than it will collect in taxes and other revenues.This 1/2 trillion dollars will be added to the debt total which is currently approximately 19 trillion dollars.To put this in perspective, the current debt of the U.S. Government is equal to approximately $45,000 for every American citizen.With the size of the deficit in mind, we would like to see how you personally would change Federal taxes and/or spending.We have created a "budget model" based on the actual Federal budget.In the model you will be able to cut spending in any category where you believe that the Federal government spends too much money and/or increase revenues in any category where you think taxes should be raised.The basic idea is to allow you to change taxes and/or spending in the ways you prefer.The model will only allow spending cuts and/or increases in Federal taxes.You will have 10 minutes to complete the budget model.After completing the model, you will be returned to the survey and asked some follow-up questions.Please click on the link below to begin the budget model." It is important to note that we did not provide any cues or signals to respondents that they should submit a particular kind of budget, such as a balanced budget.We provided information about the budget deficit to motivate respondents to submit a budget that they support.In one section of the instructions, for example, we state "We would like to see how you personally would change Federal taxes and/or spending."Later in the simulation we reinforce this point when we tell respondents, "the basic idea is to allow you to change taxes and/or spending in the ways you prefer." The budget simulation contains both tax and spending items.Figure 1 is a screen capture of the online budget simulation used for this study.The tax-side contains six income brackets and an additional six categories including social security payroll

Figure 1 Screen Capture of Online Budget Simulation
and corporate income tax.Effective tax rates and revenue amounts were given for each tax item. 1 On the spending-side of the budget we included nine broad spending categories and a total of 27 subcategories.Each budget item included an information bubble with a short description to help clarify the contents of the item.Respondents could change any item in increments of 1% by clicking on "+" and "-" signs.
Finally, though previous research by Rudolph & Evans (2005) and Rudolph (2009) has applied the interaction between the trust-as-heuristic and ideology to support for tax cuts and spending increases, we believe that the results could be limited by the fact that the sacrifices related to tax cuts and spending increases are not great enough to elicit a strong enough response.Instead, we allow respondents to only increase taxes and/or cut spending, and as such, we believe we can elicit a greater ideological and material sacrifice.
The budget simulation has several desirable attributes as a research instrument to measure preferences.First, it provides respondents with information about both sides of the budget, and all the major categories and subcategories that constitute each side.Second, respondents receive instantaneous feedback about how their individual choices on a single item impact the overall budget.Third, the budget simulation logs how respondents interact with the individual lines in the federal budget, producing a wealth of fine-grained information for researchers.In particular, the simulation stores information about which items of the budget are adjusted and the degree of adjustment for each item.Using the spending item "foreign military aid" as an example, data was collected on whether a respondent reduced foreign military aid, and the actual dollar amount to which it is reduced.

Hypothesis
In studying spending and tax preferences separately, Rudolph & Evans (2005) and Rudolph (2009) pose specific hypotheses about the moderating effect of ideology on trust.Here, we generalize these hypotheses to test whether trust in government, as Rudolph & Evans (2005) put it, "erases the ideological gap" between liberals and conservatives.We formally state our hypothesis as: H1: At higher levels of trust, there is no difference between liberals and conservatives in their preferences toward tax increases.
H2: At higher levels of trust, there is no difference between liberals and conservatives in their preferences toward spending cuts.

Dependent Variables
The budget simulation collects data on the number of items respondents adjust and the extent which they adjust each item.In general, data on whether an item (or items) has been adjusted is useful for studying budget preferences at the macro-level, such as explaining mass support for tax increases or spending cuts, while the dollar amount adjustment on a particular item is useful for budgetary questions at the micro-level, like explaining support for health care spending.The reason for this macro-micro distinction relates to how the budget simulation is designed.Respondents are able to adjust items at 1% increments, but in terms of actual dollar amounts, a 1% incremental change varies by item depending on the initial allocations in the budget.For example, in terms of the actual dollar size of the federal budget, a 1% decrease in defense spending roughly equates to an 8% cut in education spending because spending on defense is currently about eight times larger than education spending.Therefore, measures using the total dollars adjusted (e.g. total spending cuts or total revenue increase) to summarize budget preferences at the macrolevel are more sensitive to small adjustments on big budget items like defense and less sensitive to substantively important adjustments on smaller items like education.
Considering the macro-micro distinction in the data, we measure respondents' support for tax increases and spending cuts at both levels.Starting at the macro-level, we use a simple count of the number of tax items each respondent adjusted (increased) in the budget simulation, ranging from 0 to 12 (labeled Tax Count).We create an equivalent count variable for the spending side, ranging from 0 to 27 (labeled Spending Count).Without imposing assumptions on the data, all items in the budget simulation are included in the measures.Tax Count and Spending Count are general measures of how respondents interacted with the budget simulation.
Additionally, we analyze three spending variables and two revenue variables at the micro-level.These variables are continuous measures and expressed in terms of billions USD.On the spending side, we examine the amount of cuts respondents made to health care, economic security, and defense. 2 Higher values on these spending variables indicate greater cuts to these line items in the budget.On the revenue side, we focus on two measures of income tax.One measure is the increase in total revenue from all six income tax brackets in our simulation.The second measure is the increase in total revenue from only the top two income categories (151k-741k bracket and over 741k bracket).Higher values on these income tax variables indicate increased revenue from these categories.
These spending and revenue categories were chosen based on the budgetary literature.Prior work has shown that liberals and conservatives tend to hold divergent positions on defense, health care, welfare, and progressive taxes. 3Correspondingly, these expected differences bear out in our sample, finding statistically significant differences between liberals and conservatives across of these categories. 4Therefore, we believe that these categories represent an appropriate test for the argument that trust in government can narrow (even well-established) gaps between liberals and conservatives.

Independent Variables
The key independent variables in our analysis are trust in government and ideology.Similar to the "trust in government" question in the ANES, we asked, "How often can you trust the federal government in Washington to do what is right?"The five-point item included the follow responses: "always", "most of the time", "about half the time", "sometimes", "never".Responses to this question are coded such that higher values indicate higher levels of trust in government. 5We also modeled our seven-point ideology identification question after the ANES, "When it comes to politics, do you think of yourself as: very liberal, liberal, somewhat liberal, moderate or middle of the road, somewhat conservative, conservative, or very conservative?"We label this variable as Liberal, with higher values indicating more liberal.
We also include a variable measuring the difficulty of the simulation because scholars have shown that respondents face cognitive constraints and optimization problems when asked to perform complicated tasks such as creating budgets (Benartzi & Thaler, 2007).This 5-point measure is labeled Budget Difficult, with higher values indicating that the simulation was more difficult for the respondents than they had anticipated.Finally, several socio-demographic control variables are included to account for respondents' income level, educational attainment, age, gender, race, and party ID.The exact question wording and the coding of these variables are included in the supplementary material.

Results for Macro-Level Measures: Spending and Tax Count
We begin with presenting the macro-level results.
We model Spending Count and Tax Count using negative binominal regressions with standard errors clustered by state.Table 1 presents the results for these count measures. 6 On the spending side (Table 1, column 1), our results show that liberals (-0.664) are less likely to cut spending items compared to moderates (reference category) while conservatives (0.303) are more likely to cut spending.These findings are consistent with general expectations about ideological differences over government budgets.Our central interest is whether both interaction terms take on the opposite sign as the ideology dummy variables.Though not statistically significant at the 0.05 level, the results conform to our expectations with the coefficient on LiberalxTrust obtaining a positive coefficient (0.103) and ConservativexTrust obtaining a negative coefficient (-0.019).
Figure 2a shows the predicted count (y-axis) of spending items reduced for liberals and conservatives at different values of trust (x-axis).At the lowest levels of trust, conservatives are predicted to decrease about 10 spending items while liberals are predicted to decrease 4.4 items.As trust increases, the predicted difference in count between liberals and conservatives decreases and becomes statistically insignificant.In contrast to Rudolph & Evans (2005) who would expect the effect of trust to be more pronounced among liberals because spending cuts requires more ideological sacrifice on their part, our results suggest that trust dampens both sides of the ideological divide over spending cuts.
On the tax side (Table 1, column 2), conservatives take a reverse position (-0.503) and are significantly less likely to adjust revenue items compared to moderates.Notably, the difference between liberal and moderates is not statistically significant at the 0.05 level, but the estimated difference between liberals and conservatives (0.480) is significant at the 0.001 level.The interaction term Con-servativexTrust is positive and statistically significant at the 0.10 level, indicating a dampening effect of trust on ideological differences.We assess the overall effect of trust among liberals and conservatives by examining the predicted counts.In Figure 2b, we can see that for respondents who never trust government, liberals are predicted to increase 4.13 tax items while conservatives increase only 2.78.These ideological differences, in terms of tax items increases, become statistically indistinguishable at higher levels of trust.These results are analogous to the findings by Rudolph (2009) who finds that trust matters more to liberals over the issue of tax cuts because such policies demand more ideological sacrifice from them.Consistent with the trustas-heuristic hypothesis, we find that trusting conservatives are willing to raise taxes. 7

Results for Micro-Level Measures: Allocations to Spending and Revenue Categories
Does the attenuating effect of trust on ideological differences hold at the micro-level, that is across specific line items in the budget?We investigate this question by using ideologically divisive budgets items from both sides of the ledger.On the spending side, we examine the amount of cuts respondents made to health care, economic security, and defense.On the revenue side, we examine the increased revenue from income tax.The spending and revenue items are modeled using OLS regression because the measures are continuous (expressed in billions USD).We apply the same model specification as for the count models described above.
We present the spending models in Table 2.The results on the ideological variables comport with general expectations.Conservative and liberals are statistically different from moderates across all three spending models.At low levels of trust, Conservatives are more likely to reduce spending on health care (100.721) and economic security (47.357) while liberals are more likely to reduce spending on defense (82.245).In health care and economic security models, the interaction terms on ConservativexTrust are negative and statistically significant (-22.966, -10.100), indicating that higher levels of trust reduce spending cuts among conservatives.In the defense model, we find paralleling results with the interaction term on LiberalxTrust negative and statistically significant at the 0.10 level.
In Figure 3, we plot the predicted amount of spending cuts for conservatives and liberals at different levels of trust.At low levels of trust, the difference between liberals and conservatives is statistically significant.The figures show that the group  differences narrow at higher levels of trust and become statistically insignificant at the highest levels.
The results on revenues from income tax are presented in Table 3 and provide further evidence of the attenuating effect of trust on ideology.For both measures of income tax revenue, low trusting conservatives are statistically less likely to raise revenue from income taxes than low trusting moderates and liberals.The coefficient on Conservativex-Trust obtains a positive sign, indicating that conservatives at higher levels of trust are more likely to raise revenue from income.We observe a similar dampening effect for liberals with a positive coefficient on liberal (113.210, 94.407) and a negatively signed coefficient on the interaction term .Figure 4 is a graph of the predicted revenues from income taxes for liberals and conservatives across different levels of trust.
Similar to the prior results, we observe a narrowing difference in budgetary preferences among trusting liberals and conservatives. 8 Overall, our findings show that trust in government narrows the ideological divide over the federal budget with trusting liberals and conservatives expressing similar preferences toward spending cuts and tax increases.

Discussion
In recent years, research on both budget preferences and trust in government has grown.Large government debt and deficits have led scholars to ask citizens about the extent to which they are willing to pay taxes in order to fund these high levels of spending, while the expansion of trust literature  corresponds to the great increase in public distrust and polarization.We have demonstrated that these issues are not unrelated.
Our results have important implications.Theoretically, we demonstrate the trust-as-heuristic theory extends to issues that are arguably more contentious than support for tax cuts and spending increases.Moreover, our results demonstrate that trust in government is a valuable mechanism for decreasing public policy divisions among individuals who are the most ideologically opposed to those policies.Our findings also have important political implications.Our article demonstrates that by harnessing trust as an important resource, policy makers can garner support for public policies that might be unpopular amongst the more ideological segments of their base, which could lead to more moderate policy outputs.
This study sits well within an increasing body of literature in public administration that notes a large decline in public trust since World War II, while at the same time offering solutions as to how good public administration can improve trust (Wang & Wan Wart, 2007;Thomas, 1998;Tolbert & Mossberger, 2006).We believe that our paper not only sheds light on how trust in government is important for garnering wide support for public policies, but we also provide our interactive budget tool as one way that governments can engage citizens, provide transparency, and improve trust.
Finally, budget simulations provide a great deal of flexibility for scholars studying preference for taxes and spending.For future research we will remove the restrictions on respondents so that they are able to increase spending and decrease taxes.This allows us to makes our theories more generalizable and sophisticated.Future research can also begin to examine how partaking in these types of exercises can increase political efficacy and civic engagement.The possible utility of these tools is numerous, providing benefits to policymakers, policy advocates, and scholars endeavoring to understand mass attitudes toward budgetary policies.

Figure 4a
Figure 4b used in our budget simulation, we refer to welfare spending cuts as Economic Security Cuts.Economic Security Cuts is calculated as the total reduction in spending from the following categories: Family and Nutrition Assistance, Housing Assistance, and Unemployment Insurance.Defense Cuts is calculated as the total reduction in spending from the following categories: Military, Veterans Benefits.See Appendix Table 3 for a detailed description of each line item.3. See Bartels (1994); Jacoby (1994) 2014) also find that other attitudinal questions on trust (e.g.crookedness and wastefulness) are time-and context-sensitive meaning that they may be interpreted differently depending on when the questions were asked.Rudolph (2009) uses the same measure of trust in their analysis on tax cuts.6.The full model results presented here are robust to varying model specifications.Additional model specifications are omitted here for space consideration and can be found in the supplementary material.Following the guidelines provided by Lenz & Sahn (2017), the supplementary material includes results without additional covariates.7. Our results are robust to different measures of spending and revenue.In the supplementary material, we provide additional robustness checks by substituting the count measures with total spending cuts and total revenue increases.8.With the exception of social security spending, we find that trust attenuates the ideological gap found in education spending, infrastructure spending, foreign aid and diplomacy.In our sample, we do not find a statistically significant difference on social security cuts between liberals and conservatives, a result consistent with the expectation that tax payers are not willing to cut social security spending.On the revenue side, we also find a statistically significant difference between the ideological groups over corporate tax and find that trust mitigates this difference.We do not find significant differences between liberals and conservatives on payroll tax, gas tax, estate tax, other taxes, or the cap on social security taxes.We interpret the non-significant differences on these other tax categories as tax payers' strong aversion to taxes in general, regardless of whether the respondent leans liberal or conservative.The supplementary material (Table 13) presents the results for these budgetary categories where we observed differences between liberals and conservatives.
Figure 2 Predicted Count of Tax and Spending Items Adjusted by Ideology Figure 3 Predicted Spending Cuts on Health Care, Economic Security, and Defense (in billions USD) Figure 4Predicted Revenues Increases from Income Tax (in billions USD)

Table 2 Spending Cuts (in billions USD) by Category
Entries report OLS regression coefficients and standard errors clustered by state in parentheses.Dependent variables are expressed in terms of the US dollar amount reduced in each category.Statistical significance denoted as: * p<0.05 ; ** p<0.01.

Table 3 Explaining Support for Spending Cuts and Tax Increases (Increased Revenue from Income Tax)
; Henderson & Hillygus (2011); Ballard-Rosa, Martin, & Scheve (2017).4. Bivariate models comparing liberals and conservatives to moderates are shown in the supplementary material.The results show that liberals are statistically different than both conservatives and moderates across these budget items.5. Following Poznyak, Meuleman, Abts, & Bishop (2014), we chose this measure of trust over other ANES measures because it represents the broadest assessment of overall trust.Poznyak et al. (