Why Do Red States Vote Republican While Blue States Pay the Bills?

Transcription

1 Why Do Red States Vote Republican While Blue States Pay the Bills? Dean Lacy Department of Government Dartmouth College 211 Silsby Hall, HB 6108 Hanover, Abstract: Every year more states receive more in federal spending than they pay in taxes to the federal government. Since the early 1980s, when data on tax burdens per state are first available, the states that receive more in federal spending than taxes paid have been increasingly Republican in presidential elections. This pattern persisted in the 2008 presidential election. The impact of federal spending on a state s Electoral Vote is increasing over the period 1984 to 2008 and remains statistically significant as a predictor of the vote when controlling for differences in ideology and opinions on issues across state populations and when controlling for per capita or median state income. Paper presented at the Annual Meeting of the American Political Science Association, Toronto, Ontario, September 3-6, 2009.

2 In the wake of the 2000 U. S. presidential election, pundits and journalists began writing much about the Two Americas: The red states on the Electoral College map that voted for Republican George W. Bush, and the blue states that voted for Democrats Al Gore in 2000 and John Kerry in The shading of the states on Electoral College scoreboards on election nights in 2000 and 2004 showed an unmistakable pattern: Bush won a swath of states through the South, Great Plains, and Rocky Mountains while his Democratic opponents won the Northeast, Great Lakes, and Pacific Coast continued the pattern with an interesting effect that the states added the Democratic column by Barack Obama were geographically connected to the states Democrats won in 2000 and The blue on the Electoral College map bled down from the New England and Mid-Atlantic states into the once red states of Virginia and then North Carolina. Indiana and Ohio, which had been a red gap between the blue states of Pennsylvania, Michigan, and Illinois, also became blue in The geography of the Electoral College changed some between 2004 and 2008, but the changes were at the margins of the seemingly stable pattern of a Democratic New England, Mid Atlantic, Great Lakes, and Pacific Rim bordering a Republican South, Great Plains, and Mountain West. The 2000 election created a wave of analyses and speculation about the differences between red states and blue states as pundits and political scientists set off in search of the singular characteristic that distinguishes inhabitants of red states from their fellow Americans in blue states. Thus far the cause or closest correlate for distinguishing red states from blue states has proved elusive. While it is clear to even casual observers 1

3 that the culture, attitudes, and lifestyle of red staters and blue staters is different in image, though perhaps not in substance, we have no idea why. One of the earliest and most controversial descriptions of the red-blue divide came from Boston Globe reporter Mike Barnicle, appearing on NBC after the 2000 election, when he dubbed the Bush states the family values states and the Gore states the sense of entitlement states. Former Clinton adviser Paul Begala responded that the Bush states were home to hate crimes, setting off a frenzy of op-ed pieces about the differences between Bush s America and Gore s America. Since then, innumerable books and op-ed pieces have taken up the claim of whether the United States is shaking on a political fault line (e.g., Fiorina 2004, Frank 2004). Mike Barnicle s characterization of the red and blue states provided an interesting and, depending on one s perspective, intuitive starting point for explaining differences between blue and red states. It would seem to make sense that the states that lose money to the federal government would be more likely to vote for the candidate who promised to cut taxes and reduce the scope of government, and that the states that gain from the federal government would support the candidate who would protect or increase federal spending. If Barnicle is correct that Democratic states are entitlement states, then we should expect that the states won by Democrats in the past few elections should receive the most in federal spending compared to the tax revenues they send to Washington. In short, Democratic states may be net beneficiaries of federal government spending while Republican states may be net contributors to the federal government. The evidence shows that such a story is exactly backwards. In a paradox of the Electoral College, Republican presidential candidates since when data on state tax 2

4 burdens are first available---have won most of the states that benefit from federal spending, while Democrats have won most of the states that bankroll the federal government. In every year during this 20-year period, between 25 and 32 states have gained more in federal spending programs than they have paid in taxes to the federal government, while the remaining minority of states has footed the bill. This political economy of redistribution plays out in the Electoral College as increasingly Republican states are increasingly dependent on federal spending. These paradoxical empirical patterns hold under several different perspectives on the data, including controlling for state and individual-level conservatism on social issues, military spending per state, and the partisan balance of a state s governorship and representation in Congress. I begin by disentangling the effects of federal spending on the 2000 election the election that started the controversy by showing that social spending not defense spending is related to Republican vote share. I then assess a number of different explanations of this paradox, finding none of them adequate. The 2004 and 2008 elections continued the pattern of red states voting Republican while blue states foot the bill. While the exact cause of this pattern is elusive, I disprove several feasible explanations and speculate about possible causes. Federal Spending and Elections Residents of a state may not realize exactly how many federal dollars they receive for every dollar in taxes they send to Washington, but the product of those federal dollars public works projects, military bases, salaries and wages, and even retirement and disability income are readily observable. People do not have to receive a check 3

5 from the federal government to benefit from federal spending. Business owners whose clients or customers receive income from the federal government are indirect beneficiaries of federal spending. Owners of property or businesses near national parks and military businesses also reap some benefits from federal spending. The flow of tax dollars into a state provides an overall estimate of how much the average citizen in that state receives from federal spending. Very little research has examined the historical relationship between federal spending and electoral votes, save for several studies of the allocation of New Deal programs (Anderson and Tollison 1991; Arrington 1970; Couch and Shugart 1998; Reading 1973; Wallis 1987, 1998; Wright 1974). The results of these often-contradictory studies suggest that the Roosevelt administration targeted some New Deal spending, particularly on agricultural programs, in pivotal Great Plains and Mountain West states. A larger body of research focuses on the relationship between federal spending and congressional elections. Stein and Bickers (1994) find some evidence that increased federal spending in a district is associated with greater chances of re-election for a member of Congress. Only electorally-threatened incumbents appear to seek added spending in their districts, and this additional spending is recognized by only an attentive subset of voters. Bickers and Stein (1996) further find that new members of Congress elected from open seats seek greater funding for their districts, particularly from small business grants. Levitt and Snyder (1995) find that federal expenditures on a limited set of programs from 1984 to 1990 are higher in congressional districts that have a larger proportion of Democratic voters. Their data do not include procurements, government 4

6 salaries, and some insurance and loan programs. They conclude that a solid majority in Congress appears to be necessary for initiating partisan spending programs. Parties appear unable to precisely target large amounts of federal funds to particular districts, not can they substantially alter the flow of federal dollars to districts over short time periods in response to changes in district circumstances. Bickers and Stein (1996), Stein and Bickers (1994) and Levitt and Snyder (1995) focus on congressional districts. Looking only at congressional districts may distort patterns of federal spending. Large portions of federal spending in states flow through state capitals or land-grant universities on their way to outlying areas. Measuring spending at the congressional district level makes spending appear higher in state capitals, which are mostly large cities, and in land-grant university towns, usually represented by Democrats. 1 Measuring spending by state may better focus the pattern of federal spending. The Senate must also approve federal expenditures, and Senators obviously care more about state-level spending than about district-level spending. Certainly presidential candidates care about states, which cast Electoral votes, more than congressional districts, which typically do not. 2 Mayer (1995), examining the 1988 and 1992 presidential elections, shows that some procurement spending, particularly civilian, is accelerated in October of election years, though the total amount of federal spending is unchanged in election years. Unlike 1 Levitt and Snyder include a control variable for dollars spent in state capitals. They also include a measure of grants flowing to districts to capture the boost to university towns, though their grants measures does not include agriculture spending funneled through land grant universities. 2 Maine and Nebraska are possible exceptions. 5

7 Levitt and Snyder, Mayer s study focuses on spending in states rather than in congressional districts. Only recently have political scientists examined the effects of federal spending on presidential politics. An earlier version of this paper (Lacy 2002) shows that states that receive the most in federal spending per tax dollar were more likely to vote for Bush in States that increased their ratio of federal spending to tax dollars during the Clinton years ( ) were even more likely to vote for Bush, controlling for the partisan composition of the state s congressional delegation, a state s leadership in Congress, and the party of the state s governor. In the overall federal budget, Democrats did worse in a state when they increased its federal receipts. Francia and Levine (2006) showed that states that are net recipients of federal spending were more likely to vote Republican in They also show that the Republican vote share increased from 2000 to 2004 in states that increased their ratio of federal spending to taxes. From this evidence, Francia and Levine conclude that big government and entitlement programs are important club goods that Republicans deliver to their states for electoral purposes, causing more funds to flow to red states in recent years. In contrast, Lacy (2002) shows that the same pattern Republican presidential candidates increase their vote share as federal spending per tax dollar increases in a state also held true from 1992 to 2000 when Clinton was president. When a Democratic administration delivered more federal spending per tax dollar to a state, the state was more likely to vote Republican in Federal spending at the president s discretion is such a small percentage of the federal budget. A few recent studies have focused on small slivers of the federal budget 6

8 such as grants and procurements, which are not programmed into the budget like Social Security or Medicare. Shor (2006) finds that from 1983 to 2001, federal grants are more likely to flow to areas that are electorally pivotal and that have a higher Electoral Votes per capita. Similarly, Chen (2008) shows that FE grants in the aftermath of Hurricane Katrina were more likely to be awarded to Republican neighborhoods and that Republican voters were more likely to vote for Bush in 2004 after receiving federal. Moderates were also more likely to vote for Bush after receiving aid, but it took about $21,000 to convert a vote to Bush in a moderate precinct compared to $7,000 to convert a vote to Bush in a Republican neighborhood. Democratic neighborhoods did not increase their vote for Bush in response to federal grants. Given the findings in Lacy (2002), however, it may be that any federal grant from any administration Republican or Democrat may increase Republican vote share in a neighborhood, congressional district, or state. All of the aforementioned studies except for Lacy (2002) and Francia and Levine (2006) focus only on the spending side of the ledger without looking at taxes. If Congress and the president increase federal spending to groups or geographic areas, the spending may not be beneficial to voters if it is offset by even higher taxes. It is the ratio of spending to taxes paid, not spending alone, that should matter to voters. The Fiscal Paradox: Net Beneficiaries of Federal Spending Vote Republican Studies of redistribution run into an interesting conundrum. At the individual level, it is easy to figure out how much a person pays in federal taxes, but it is difficult to measure how much they receive in federal spending, directly or indirectly. At the state or 7

9 congressional district level, it is easy to figure out total federal spending (the Census Bureau keeps track of this at the behest of Congress), but it is difficult to figure out how much everyone within a geographic boundary pays in taxes. Calculating a state s federal tax burden and federal spending benefits is possible with data provided by Tax Foundation. For FY 1984 through 2005, the Tax Foundation calculates the federal tax per capita paid from each state (Moody 2001). The Bureau of the Census provides per capita federal spending in each state, broken down by a number of spending types and federal agencies (U.S. Census Bureau Consolidated Federal Funds Report). The Tax Foundation, using Census data, further calculates a federal spending to tax ratio for each state, which represents the amount of federal spending for each dollar received from the state in taxes. 3 3 The Tax Foundation is a non-partisan research organization, though it does advocate tax reduction and simplification. The Foundation s tax numbers are widely-used by researchers. Some scholars, such as Paul Krugman, have criticized the Tax Foundation s numbers as biased. Even if biased, which we do not concede, the form of bias is unimportant for tracking over time changes in voting patterns as long as the measures are consistently biased. To calculate the spending to tax ratio, the Tax Foundation removes budget deficits or surpluses by scaling total revenues received from the states to match total federal expenditures. This adjustment does not significantly change the states relative rankings on the spending to tax ratio, but it does reduce the number of states that appear to receive more in federal spending than they pay in taxes, given the budget of FY Using the Tax Foundation-adjusted spending to tax ratios, 30 states in FY 2000 were net beneficiaries of federal spending. Using unadjusted ratios, only 16 states were. In times of deficit spending, more than 30 states appear to benefit from federal spending. Rescaling a state s spending benefit to remove federal deficits and surpluses makes sense as these are common pool resources (or debts) and cannot be allocated to individual states. 8

10 In each year from1984 through 2005, the most recent year for which Tax Foundation data are available, between 25 and 32 states received more in federal spending than they sent to Washington in taxes, as shown in Figure 1. In most years, 30 states are net beneficiaries of federal spending, which is perhaps a coincidence or an invidious indicator of the politics of redistribution. If the activities of Congress could be defined as taking money from some to give to others, then in order to make any forced redistribution pass the Senate, a Filibuster-proof majority of 60 Senators (or 30 states) would need to benefit. Political-economic models of redistribution tell us that potential beneficiaries of forced redistribution would form a minimum winning coalition just enough votes to ensure passage of their redistribution plan, but no more (Riker 1962). The fact that in recent years about 30 states are net beneficiaries of federal spending while 20 are net contributors fits neatly with such a simple model of forced redistribution. However, the states that benefited from federal spending in 2004 held only 231 Electoral votes, less than the 270 needed for a majority. Interestingly, in 2000, one state, Florida, received one dollar for every dollar it sent to Washington, making it the pivotal state in the line at the pork barrel, just as it was in the 2000 Electoral College. In 2004, Ohio occupied a break-even position in federal spending and was also widely-seen as the key swing state. Figure 2 shows the net contributors to the federal government in 2000 as the shaded states; the net recipients, the unshaded states. The pattern in most years since the 1980s is similar. The Northeast and Midwest bankroll the South and Great Plains. The pattern of shaded and unshaded states also looks remarkably like the pattern on the Electoral College maps of 2000 and In fact, the relationship between the states that 9

11 voted for Bush in 2000 and 2004 and the states that are net beneficiaries of federal spending is strong by any measure (χ 2 =6.04, p<.02). Federal spending is not fixed over time, however. Even though two thirds of the budget of predetermined by programs such as Social Security and Medicare, and much of the rest of the budget is non-discretionary, federal spending ratios fluctuate quite a bit across the states. Figures 3 through 5 depict changes in federal spending ratio over time. Figure 3 shows on the horizontal axis a state s ratio of federal spending per tax dollar in 1984 while the vertical ais shows the same ratio for A 45-degree on the graph passes through the set of states whose ratio stayed the same between 1984 and States above the line increased their ratio of federal spending to tax dollars during the Reagan-Bush budgets from 1984 through States below the line received a lower ratio of spending to tax dollars (either through spending going down or tax burden going up) between 1984 and Figure 4 is the same, but with the horizontal axis starting in 1992 and the vertical axis capturing 1999, spanning the Clinton years. Figure 5 captures the Bush years, with the horizontal axis showing spending ratios in 2000, and the vertical axis showing the ratios as of the last year available, In each graph the blue states are the states that voted Democratic and the red states the ones that voted Republican in the previous election (2004 for Figure 5, 2000 for Figure 4). In all three graphs, most states cluster along the 45 degree line. States do change their ratios a bit over time, but the changes are not large. New Mexico and Mississippi are always among the highest in federal spending ratios; Connectictut and New Jersey among the lowest. If a pattern does emerge across the three graphs, it is is that red states appear more frequently than blue above the 45 degree line, particularly during the

12 1999 period. Red states increased their share of federal spending per tax dollar during most the period and across the larger range 1984 through Figures 6 through 8 show the relationship between the ratio of federal spending per tax dollar and Republican presidential vote, focusing on the 2000 through 2008 elections. The horizontal axis in each figure is the ratio of federal spending in each state to the federal taxes paid by residents of the state in the year before the election for which tax data are available (1999, 2003, and 2005). All states to the right of the vertical center line receive more in federal spending than they pay in taxes; states to the left pay less than they receive. The vertical axis is the Republican margin of victory, with Republican states above the horizontal line at 0 and Democratic states below. The straight line sloping upwards through the points is the linear regression line that best captures the relationship between federal spending ratio and Republican vote. The crooked line is a lowess non-linear regression line, which in all three graphs shows a slight decreasing effect of federal spending ratio on Republican vote share for higher spending ratios. The states shown in blue in all three graphs voted for John Kerry in 2004; in red, for Bush in The one blue state in 2004 that voted for Bush in 2000 was New Hampshire, a net contributor to the federal government. The one red state in 2004 that voted Democratic in 2000 was New Mexico, one of the largest net beneficiaries of federal spending. From 2004 to 2008, several states switched to the Democratic column, one with a high spending ratio (New Mexico), and several with low ratios (Colorado, Nevada, Florida). In all three graphs, a positive relationship appears between federal spending per tax dollar in each state and Republican electoral margin. The relationship is actually 11

13 stronger in 2004 than in 2000 as New Hampshire, a net contributor, switched to the Democratic side of the horizontal line, while New Mexico, and net beneficiary, jumped across the Republican side. Although New Mexico flipped back to Democratic in 2008, and Virginia, a net beneficiary, moved Democratic, the additions of several moderate to low spending ratio states kept the pattern of 2008 similar to The non-parametric lowess line shows a slight non-linear effect of spending ratio on electoral votes, where the effect is greatest -- the line is steepest -- right where a state breaks even in the federal budget. The effect of the federal spending ratio on Republican margin is steepest for the marginal states. It may or may not be coincidence that in 2000 Florida was a break-even state in the federal budget and the focus of all American electoral attention, while in 2004 it was Ohio, and in 2008 a set of states on the vertical line including North Carolina, Indiana, Ohio, Florida, and Georgia. The figures thus far simply show a relationship between federal spending ratio and Electoral votes. Other things may explain Electoral vote that are also correlated with or related to federal spending ratios. To assess further the spending-votes link, I estimate a series of regressions of Republican margin of victory in a state on different measures of a state s benefit from federal spending, focusing on the 2000 election. 4 Table 1, column 1, shows the basic regression. As a state s ratio of federal spending to taxes increases, Republican margin of victory increases. 5 Substantively, the regression shows that Bush s 4 Margin of victory is Bush s percentage of the vote in a state minus Gore s percentage of the vote, ranging from a maximum value of 43.4 in Utah to in Rhode Island, with the District of Columbia omitted. Margin of victory has a value of 0 in Florida and New Mexico. 5 The ratio reported by the Tax Institute is a state s adjusted federal spending per capita divided by its per capita tax burden, ranging from a high of 2.3 in New Mexico to 0.6 in Connecticut. I delete the District of 12

14 margin over Gore increases by 20 percentage points for every dollar in spending per dollar of tax that the state receives. Put another way, Bush s margin goes up by 2 percentage points for every additional dime of federal spending in a state per dollar of taxes paid by that state. Possible Causes of the Fiscal Paradox Lumping all federal spending together may paint a distorted picture of the relationship between spending and votes. It may be that Republican states receive most of their federal return on the dollar in the form of defense spending, especially given the relatively larger military presence in the South and Mountain West than in the Northeast or Great Lakes. To examine this possibility, I define two separate ratios of federal spending per tax dollar: One ratio for the amount spent by the Department of Defense per state, including military salaries and procurements, and another for the non-defense dollars spent per state. Putting both of these variables in the regression model, column 2 of Table 1, pushes the paradox of federal spending even further: There is no relationship between defense spending per state and the state s Electoral College vote. However, the relationship between non-defense spending and the vote becomes even stronger: States that benefit the most from non-defense spending (from retirement and welfare payments to farm subsidies to highway construction) give even higher margins to Republicans. For each additional 10 cents per dollar that the federal government spends in a state, Bush s 2000 margin increases by over 2.9 percentage points. Columbia from the analysis since its spending to tax ratio of 6.5 is clearly an outlier with tremendous leverage in any statistical model. 13

15 The paradoxical relationship between federal taxes and spending levels and votes for Bush may be an artifact of Congress. Republicans in the House and Senate may be funneling federal dollars back to Republican states and congressional districts. Republicans may also spend more federal dollars in states with Republican governors in order to bolster their popularity. An easy test of this hypothesis is to control for the proportion of Republicans in a state s Senate and House delegations, as well as whether a Republican held the state s governorship at the time of the 2000 election. 6 Results from the model appear in Table 2, where federal spending per state is measured in the first data column by the ratio of defense and non-defense spending to tax dollars, and, in the second data column, by the per capita levels of federal taxes paid, federal defense spending, and federal non-defense spending. In both columns it is clear that the proportion of Republicans in a state s House delegation is positively associated with Republican margin of victory; the number of Republicans in a Senate delegation, less so. More importantly, in both models the level of federal spending in a state remains statistically and substantively significant. The ratio of federal spending per dollar of tax revenue from a state is positively related to Republican margin: More federal spending per tax dollar is associated with a higher Republican vote. In the second model, a higher federal tax burden is negatively associated with Republican margin: States with a higher per capita tax burden give Democrats more votes. Spending levels are not statistically significant. 6 I code Independent governors, Senators, and House members as 0.5 (half Republican, half Democrat), but doing so does not affect the results. The results are also unaffected if I measure Republican strength using post-2000 election numbers. 14

16 As another check on the robustness of the relationship between a state s financial gain from the federal government and its vote, I add a control for the proportion of the state s vote earned by Bill Clinton in This is yet another measure of the partisan balance in a state, but before the fiscal paradox appeared fully. Controlling for Clinton s vote, we find again that a state s contribution to the federal government measured by ratio of federal spending to tax dollars or by separate measures of tax burden and spending is closely associated with Republican margin in The connection between federal spending per district and Republican vote is in 2000 is remarkably robust. Controlling for Clinton s margin of victory does wipe out the effect of the Republicans proportion of delegates to the House and Senate from a state. It is a state s proportion of Republican voters, not its proportion of Republican leaders, that determines Bush s margin. Another perspective on the relationship between a state s representation in Congress, its prior presidential vote, and its level of federal spending comes from turning federal spending into the variable to be explained. With the ratio of federal spending to tax dollars as the dependent variable in Table 3, it is clear that the state s proportion of Republicans in the House or Senate, and whether the state s governor is Republican, has no explanatory value. Clinton s vote share in 1996 does, however, explain the state s ratio of spending to taxes, but the relationship is the opposite of what one would expect if Clinton were paying off states that supported him in Instead, higher Clinton vote shares in 1996 are associated with lower ratios of spending to tax dollars in a state in Similarly, in column 3, a higher Clinton vote in 1996 is associated with a higher per capita federal tax burden in For each additional percentage point in Clinton s 15

17 1996 vote share, a state s residents could expect an 83 dollar increase in their federal taxes. Party control of Congress and governorships does not explain why red states vote Republican while blue states pay the bills. The real driving force behind Republican vote shares in presidential elections may be the Culture War, values, or the general ideological and issue positions of a state s voters. This conservatism may be correlated with federal spending in a state. Coming up with measures of state-level political ideology or issue positions across all 50 states using comparable questions is not a straightforward task given the paucity of good survey data across states. Recent surveys by Polimetrix, conducted over the Internet, have large enough samples of respondents across all states to provide a measure of state ideology (Bafumi and Herron, 2009), but these surveys begin in For years prior to 2004, the best data source for cross-state comparisons on ideology is the Senate National Election Study, which presented samples of a few hundred voters the same set of issue questions in 1988, 1990, or 1992, depending on the year of the Senate election. Norrander (2001) uses these data to compare the states on several issues by calculating the median response in each state on a set of a dozen questions ranging from general ideology to abortion to defense spending to aid to minorities. I factor analyze Norrander s data on the median response on each issue scale in each state to derive a two-factor model that summarizes 55 percent of the variance across states. The first dimension is roughly a measure of social conservatism, with high loadings for general ideology and opinions on abortion, public funding of abortion, aid to minorities, funding to fight AS, and spending on the environment. The second factor comprises party identification, funding for public schools, and unemployment insurance. 16

18 It seems reasonable to call the first factor social issues, which explain 36 percent of crossstate variation. The second factor is economic issues, explaining 20 percent of cross state variation. These measures are available only for the period, but state-level opinions on these indicators are likely to be quite stable during the period. California has not become conservative over this period; Texas has not become liberal. Figure 9 plots the states on these two dimensions, with red and blue states colored based on the 2008 election. The pattern of states in Figure 9 is both intutive and striking. A one dimensional model of state ideology would not appear to fit well with the data given the significant variation across states on the two dimenstions. On social issues, Vermont and Rhode Island are the most liberal states while Arkansas and Oklahoma are the most conservative. On economic issue, Hawaii, Maryland, and Louisiana are the most liberal with Alaska, Idaho, and Wyoming the most conservative. These placements make sense substantively. It is also interesting that some states, such as Vermont, are liberal on social issues but not particularly liberal on economics. Other states, particularly Louisiana, West Virginia, Kentucky, and Mississippi, are conservative on social issues but liberal on economic issues. To the extent that these data from around 1990 still reflect politics in 2008, it is noteworthy that the states split between blue and red on social issues, not economic issues. The two red states furthest into the blue on social issues are also the states represented on the 2008 Republican ticket: Arizona and Alaska. The purpose of the factor analysis defining state ideology is to provide control variables for a regression of presidential vote on federal spending ratio. For each election year from 1984 through 2008, I estimate a regression model where the observations are 17

19 the 50 states, the dependent variable is Republican presidential vote margin and the independent variables are the state s score on the social issues dimension, the state s score on economic issues, and federal spending ratio in a state. Figures 10, 11, and 12 graph the results of this regression over time. First, Figure 10 shows the over time effects of state spending ratio, along with a 90 percent confidence interval. The substantive effect of this variable is the increase in Republican presidential margin due to a one unit (one dollar) increase in federal spending per tax dollar. The graph shows that federal spending per tax dollar has had a positive effect on Republican vote across all years, but the effect differs from zero starting only in The effect of federal spending reaches its maximum in 2000 and has drifted slightly downward since, though it is still clearly statistically and substantively significant. Figure 11 shows the effect on Republican presidential margin of a one standard deviation shift to the right in a state s population on social issues. Social issues have a non-zero effect on Republican vote margin throughout the period, and the effect is clearly increasing. Even against the backdrop of the occupation of Iraq and a failing economy, social issues were a stronger predictor of a state s vote in 2008 than in 2004 or Figure 12 presents the same effect for economic issues on the same scale as social issues. The graph peaks in 2000, when state position on economic issues exerted its strongest effect on the vote. Since 2000, economic issues have declined in their effect on the vote, even in the wake of the economic downturn of While economic issues still have a clear non-zero effect on state presidential vote, the effect is not as substantively large as for social issues. Comparing the figures, Figure 10 overstates the magnitude of the effect of federal spending since the unit of measure is in dollars rather 18

20 than standard deviation units. To rescale the Figure to be in standard deviation units, reduce the size of the effect by about half. Social issues have a larger substantive effect than economic issues, whose effect is larger than federal spending ratio up to In 2008, a one standard deviation change in federal spending ratio had a larger effect on Republican margin than a one standard deviation change in economic issues. Taken together, the three variables of federal spending and two issue dimensions explain about 70 to 75 percent of the total variance (as measured by the adjusted R- square) in Republican vote for each year from 1984 through Federal spending as measured as a ratio to tax dollars paid by a state has risen considerably during this period as a positive predictor of Republican vote. It has not reached the same level as issues during this period, as measured by the factor scores. But when one considers that a state s issue positions are less likely to vary over time than federal spending, the impact of federal spending may be an even greater short-term force in elections. It may be that federal spending ratio is simply a proxy for the median income of a state s citizens. Citizens of wealthier states pay more in taxes and may also receive less in federal spending, though very little federal spending is specifically earmarked for lower income states or citizens. One would expect wealthier states to vote Republican given the Republican party s opposition to income redistribution, social spending, and regulation of business. But, income may be a proxy for the limousine liberalism that Republicans suggest drives Democratic vote shares. To test this hypothesis, I added state median income to the regression models reported above (Table 5), and in no case was median income a statistically significant predictor of Republican vote controlling for federal spending ratio and state conservatism on issues. However, median income alone 19

21 in a regression model is significant negative predictor of Republican vote: states with higher median incomes tend to vote Democratic. Yet another possible explanation rests on misinformed voters. It could be that the states, and voters, that benefit most from federal spending do not realize their gains and instead believe that their tax dollars subsidize someone else. Many pundits and journalists seem to believe that liberal, Northeastern and Pacific Coast states pull funds out of the federal government for entitlements, welfare, pork barrel projects, and other spending programs. Yet few pundits speak of the large agricultural and welfare spending in Southern and Great Plains states. It could be that Southerners do not realize they are net beneficiaries of federal spending, nor that residents of the Northeast, Pacific Coast, and Great Lakes realize that they lose money to federal spending. It is difficult to imagine, however, that voters would remain misinformed for long. 7 To test whether misinformation accounts for the federal fiscal paradox, I included in a 2004 election survey a set of questions to assess voter knowledge about federal spending. The survey was conducted by Knowledge Networks using their nationwide randomly-drawn sample. In the three wave panel, questions about federal spending appeared in the first wave, in April, The 1317 respondents answered questions via Web-TV in their own homes. Responses were recorded by using a computer mouse to 7 The paradox of federal spending and Electoral votes finds mention only once in the popular press, at the end of a little-known article: Middle America may not even, alas, be a haven of rugged individualism. Midwestern voters who tell pollsters that they favor small government are in no rush to give up farm subsidies. One little-known irony of political geography is that the more conservative states tend to be among the top feeders at the federal trough. Bush won 25 (sic) of the 31 states that get more in subsidies from the federal government than they pay in taxes. (Young 2001). 20

22 click on responses. The questions appeared at the end of the survey, which asked voters several questions about their political opinions and intended vote in The first question asked: Think about all the taxes you pay to the federal government, such as income taxes and Social Security taxes. Also think about how much money you get from the federal government in Social Security, Medicare or Medicaid, salary or wages if you are employed by the government, farm subsidies, education loans, veterans' benefits, or anything else. Would you say that you get more money from the federal government than you pay in taxes, do you pay more in taxes than you get back, or do you not know? [1] I pay more in taxes than I get back from the government [2] I get back from the government more than I pay in taxes In the states that were net contributors to the federal government in 2003, 23 percent of respondents chose [2]. In the net beneficiary states, 25 percent of respondents chose [2]. There was no statistically significant difference between the states in the proportions of respondents who believe they are net beneficiaries of federal spending (p=.40). A second question asked: Now think about all the people in the state in which you live. Think about all the money they pay in taxes to the federal government. Also think about all the people in your state who may work at military bases, who get Social Security, Medicare or Medicaid, Veterans' benefits, education loans, farm subsidies, salaries and wages, or anything else. Would you say that the people in your state 21

23 as a whole get more from the federal government than they pay in taxes, or do they pay more in taxes than they get back from the federal government? [1] People in my state pay more in taxes than we get back from the government [2] People in my state get back from the government more than we pay in taxes In the net contributor states, 24 percent of respondents chose [2]. In the net beneficiary states, 30 percent chose [2], indicating a statistically significant difference (p=.02) in the proportion of residents in net beneficiary states who believe their state is a net beneficiary. Putting results from the two questions together, residents of net beneficiary states do not believe they are net contributors to the federal government, either as individuals or as states. Residents of net beneficiary states know their states are net beneficiaries of federal spending. They are also as likely as residents of net contributor states to believe that they are personally benefiting from federal spending. Of course, most respondents in both net contributor states and net beneficiary states believe they are net contributors to the government. Testing whether this is actually true is impossible. But in this case perception of benefit is more important than actual benefits. Based on the data from 2004, we can rule out voter misinformation as a cause of the federal fiscal paradox. Another possible explanation is the opposite of the misinformed voters story: Residents of states that receive more in federal spending than they pay in taxes may live closer to the effects of such spending and want to cut it. It could be that voters in the states that benefit from federal spending see such spending negatively. In the Mountain West, some federal spending takes the form of frequently-reviled national forests and water projects. In the South, some voters may believe that they benefit little from federal 22

24 spending while other residents of their state benefit a lot. This explanation is difficult to believe: if true, why would voters in the Northeast, Great Lakes, and Pacific Coast continue to vote for sending their tax dollars to other states? Similarly, the relationship between federal spending and Electoral votes could be an artifact of examining states rather than individual voters. Imagine a state with a high ratio (greater than 1.0) of federal spending to tax dollars. Suppose that less than half of the residents of the state receive more in federal dollars than they pay in taxes, but the half that receives subsidies more than makes up for the half that do not, such that the subsidized population must be receiving some federal dollars from out of state. It could be that the subsidized voters in the state voted for Gore, while the majority the voters who lose money to taxes voted for Bush. In states that lose money from taxes, a majority of voters might still benefit from federal spending while a small and substantially-taxed minority lose money. The over-taxed minority may vote for Bush while the net recipients of federal spending vote for Gore. This explanation presumes that federal spending in a state produces benefits almost exclusively for the people who directly receive federal payments. It ignores the possibility that individuals who are net contributors to the federal government may receive indirect benefits from spending, such as payments (in wages or purchases) from individuals who are direct recipients of federal dollars. The explanation also leaves questions unanswered: Why do states with higher federal tax burdens vote for Gore, independent of per capita federal spending in the state, while states with lower tax burdens vote for Bush? Why wouldn t states that are net contributors to the federal government favor less federal spending and more state spending, so that they can keep their tax dollars in state? 23

25 Discussion The relationship between federal spending and Electoral Votes is not an artifact. It is robust to controlling for income, social issues, political characteristics of the states, and a number of other factors that could create a spurious correlation. States the get the most money from the federal government relative to their taxes paid are more likely to vote Republican in presidential elections, and increasingly so. At first glance, this statement makes little sense. Why would residents of states that get the most in federal spending vote for candidates who want to cut that spending? But the relationship between federal spending and electoral votes is more complex than simple self interest. Many reasonable and intuitive explanations for the federal fiscal paradox do not withstand logical or empirical scrutiny. Two potential explanations seem to remain. New Deal programs entrenched federal spending programs that voters at the presidential level do not want to change. It makes sense that voters guard their favorite forms of federal spending and want to keep it while electing presidents that pledge to keep the size of government from growing more. Voters in states that get the most in federal spending may also support low state spending, creating a more conservative political environment in the types of candidates that win local and state offices and move on to national office where they then protect the federal programs the benefit the state. Is the federal fiscal paradox electorally stable over the long term? Probably not. Why would voters in Connecticut and New York and California and Ohio continue to vote for federal programs that send their dollars to other states instead of gutting those programs at the federal level and reinstituting them at the state level where the dollars 24

26 stay within the states s borders? One implication of the federal fiscal paradox may be that Democrats in the Northeast, Great Lakes, and Pacific Coast may become advocates for state s rights and smaller federal government. Republicans in the South and Great Plains may protect or even increase federal spending programs as necessary for national improvement, much as the American System of Henry Clay, while nationalizing issues such as gay marriage to create a Republican form of good government conservatism that some pundits already proclaim George W. Bush has done. One possible direction that the federal fiscal paradox may push the parties is toward a modern version of the 1820s Democrats, who argued for limited government and states rights, and the Whigs, who pushed for more expansive federal government. Whatever the long-term outcome, the federal fiscal paradox in its current form is not likely to persist. 25

27 References Anderson, Gary, and Robert Tollison Congressional Influence and Patterns of New Deal Spending, Journal of Law and Economics 34(April): Arrington, Leonard Western Agriculture and the New Deal. Agricultural History 44 (October): Bickers, Kenneth N., and Robert M. Stein Congressional Elections and the Pork Barrel. Journal of Politics 56(2): Brooks, David Are We Really One Country? A Report from Red and Blue America. The Atlantic Monthly. December: Chen, Jowei When do Government Benefits Influence Voters Behavior: The Effect of FE Disaster Awards on US Presidential Votes. Working paper. Couch, Jim F., and William F. Shughart II The Political Economy of the New Deal. Northampton, : Edward Elgar. Fiorina, Morris P. with Samuel J. Abrams and Jeremy C. Pope Culture War? The Myth of a Polarized America. New York: Longman Frank, Thomas What s the Matter with Kansas: How Conservatives Won the Heart of America. New York: Holt. Peter L. Francia and Renan Levine "Feast or Famine at the Federal Luau? Understanding Net Federal Spending under Bush" The Forum 4.2 Levitt. Steven D., and James M. Snyder, Jr Political Parties and the Distribution Of Federal Outlays. American Journal of Political Science 39: Mayer, Kenneth Electoral Cycles in Federal Government Prime Contract Awards: 26

28 State-Level Evidence from the 1988 and 1992 Presidential Elections. American Journal of Political Science 39: Moody, Scott Federal Tax Burdens and Expenditures by State. Tax Foundation Special Report 105 (July). Norrander, Barbara Measuring State Public Opinion with the Senate National Election Study. State Politics and Policy Quarterly 1 (March, 2002): Reading, Don New Deal Activity and the States, 1933 to Journal of Economic History 33 (December): Riker, William H The Theory of Political Coalitions. New Haven: Yale. Rundquist, Barry, and David Griffith An Interrupted Time Series Test of the Distributive Theory of Military Policy-Making. Western Political Quarterly 24: Shor, Boris Presidential Power and Distributive Politics: Federal Grants Expenditures in the 50 States. Working paper Stein, Robert M The Allocation of Federal Monies: A Synthesis of Demand-Side and Supply-Side Explanations. American Political Science Review 75: Stein, Robert M., and Kenneth N. Bickers The Electoral Dynamics of the Federal Pork Barrel. American Journal of Political Science 40(4): U.S Census Bureau Consolidated Federal Funds Report, Washington, DC: U.S. Department of Commerce. April Wallis, John J Employment, Politics, and Economic Recovery During the Great Depression. Review of Economics and Statistics 49(August):516 27