Bicameral legislatures and fiscal policy.

Crain, W. Mark

John Charles Bradbury (*)

W. Mark Crain (+)

Early and modern scholars both presume that bicameral chambers limit the exploitation of minorities by the ruling majority similar to supermajority voting rules. We explain theoretically why bicameralism is a unique and desirable institution for protecting minority interests. The empirical analysis examines the structure of bicameralism in the American States. Using detailed data to proxy voter preferences, we find the degree of constituent homogeneity across chambers to be an important determinant of government expenditures for several budget components. Decreased constituent homogeneity tends to reduce redistributive spending and increase spending on public goods.

1. Introduction

Of all the topics relating to the theory of representative government, none has been the subject of more discussion, especially on the Continent, than what is known as the question of the Two Chambers. (Mill 1950, p. 436)

John Stuart Mill's observation in the mid-19th century is clearly outdated; the two-chamber question receives scant attention in modern political economy. The absence of continued debate might simply reflect a general satisfaction with the dual-chambered structure as a desirable democratic institution. Most developed democracies and all but one of the American states have adopted a bicameral legislative structure. In this paper we stress the many variations in bicameralism, even among the American states, and the effects these variations have on policy making and fiscal outcomes. Put simply, bicameralism is not an all-or-none institution. Instead, bicameralism is a variable institution that alters the identity of the median legislator within each chamber, and determines the similarity of the dominant coalitions between the two chambers. In this perspective, the impact of bicameralism on policy comes principally from the different bases of representation across the two chambers. Moreover, the analysis stresses that the degree of bicameralism changes regularly as a consequence of the redistricting process.

The remainder of the paper is organized as follows. Section 2 reviews the main conceptual arguments underlying the function of a bicameral organization. Section 3 examines implications for fiscal policy using data from the American states. We develop several measures of "bicameralism" using detailed constituency data on economic and demographic characteristics at the level of state legislative districts. We then use these measures to examine the effect of interdistrict constituent homogeneity on total state government spending as well as on the major components of state budgets. Section 4 summarizes the main findings and offers concluding observations.

2. Toward a Positive Theory of Bicameralism

Bicameralism has not always been the norm in the United States; instead, it evolved from unicameral systems used in the pre-Revolutionary War era. (1) Initially, most colonies had singlechamber legislatures whose members represented different coalitions of citizens. These coalitions ultimately grew and then separated into distinct chambers for passing laws. By the time of the founding of the United States in 1789, only Pennsylvania and Georgia still operated under unicameral systems. At the national level, the United States' first legislature under the Articles of Confederation was unicameral, a contentious element with the Articles. Drawing on this experience, the Framers feared that assigning legislative power only to states or to the population as a whole would result in the tyranny of majority by small-population states or large-population states. The eventual compromise over constituent representation was deemed so important to the Constitutional Convention that it was dubbed the "Great Compromise." The adoption of a bicameral system under the new Constitution, with two legislative bodies differently composed, was considered a progressive step to alleviate many of the problems with the former system.

Several modern scholars who have analyzed bicameralism using formal techniques generally reach a common conclusion, one that reinforces the Framers' intuitive logic. In Madison's terms "the improbability of sinister combinations will be in proportion to the dissimilarity of the two bodies" (Hamilton, Jay, and Madison 1993, p. 247). (2) We exposit two channels through which bicameralism predictably protects minority interests against policies that they would not favor. These channels, labeled Constituent Homogeneity and Legislative Stability, provide the basis for the implications regarding state fiscal policy examined empirically in the next section.

Constituent Homogeneity

Seminal works in modern political economy posit that the major effect of bicameralism stems from different bases of representation in the two chambers. Tullock (1959) first noted that a second legislative chamber dampens the inherent problem of tyranny of the majority much like the adoption of inclusive voting rules (for example, a rule that requires a two-thirds majority to approve legislation). The requirement to satisfy legislative preferences in two differently composed chambers decreases the likelihood that a minority's interests will be ignored in the final agreement. (3)

If the composition of chambers leads to differing median policy preferences, median differences should predictably affect legislative outcomes. Chambers will find less room for agreement as their demographic compositions become less similar. Consider the extreme case in which the districts in the two chambers hold no constituents in common. As Gilligan, Marshall, and Weingast (1989, P. 42) note, in that case "when different interests dominate different houses, each interest, in effect, holds a veto over legislation." However, as the difference shrinks between the median district in the upper chamber and lower chamber, it becomes increasingly easy for the two chambers to reach consensus on policy. More similarity across the two chambers facilitates intercameral trade, because as constituencies across chambers become more similar, bicameral chambers will be more likely to agree on which constituents to tax or to subsidize.

The homogeneity of constituencies raises potentially observable implications. First, two chambers with quite different constituencies should experience less spending on redistributive projects than two chambers with similar preferences. Second, spending on public goods should increase as constituent diversity across chambers increases. As redistribution becomes difficult, and therefore relatively less likely to occur, constituents will be more willing to bear higher tax burdens because revenues will be devoted to expenditures on public goods.

Legislative Stability

A bicameral structure may also reduce cyclical majority problems by excluding many non-Condorcet-winning majorities from the legislative choice Set. If the legislature enacts a law that can be defeated by the formation of a new coalition, a change in legislative preferences, or legislative turnover, policy outcomes will change frequently. This can impose costs on the economy through the instability of laws and by encouraging legislators to act strategically in response to cycles. (4) A dual-chamber legislature induces legislative stability similar to a supermajority rule, yet it has the advantage of allowing majority agreement on single-dimension policy issues that do not cause policy cycles. For instance, spending decisions regarding public goods, on which there is likely to be stable majority agreement, will pass under a simple majority rule. However, under a supermajority requirement these projects might not pass even though they would not increase budget swings.

Riker (1992) touts this second virtue of bicameralism: the use of dual legislative chambers to reduce the feasible set of policy outcomes, which in turn produces stability in legislative outcomes. If preferences are multidimensional and unstable, fiscal policy may cycle with changes in the majority coalition over time. The preferences of legislatures are inherently unstable because of frequent elections that result in continuous turnover. As a result, Riker argues that any majority coalition will seek a higher level of government activities than it would prefer if it were to continue in power. Riker's argument runs as follows. Coalition members know that in future periods they may be net losers in legislative transfers; thus coalitions will behave strategically to exploit their temporary majorities. Riker postulates that all current majorities will seek inefficiently high levels of government activity to offset future losses. Though a supermajority requirement in a unicameral chamber could achieve identical s uccess in producing Condorcet winners on multidimensional issues, one-dimensional issues that would be Pareto optimal under simple majority rule will fail. Bicameralism uniquely reduces the passage of non-Condorcet winners on multidimensional issues while allowing majority agreement on single-dimensional issues. (5)

Dixit, Grossman, and Gul (2000) provide a related argument for why a bicameral legislature is superior to a unicameral chamber operating under a supermajority rule. In the model, two interests vie for the power to make political decisions. The party in power may use its status to exploit the minority or adopt general rules. Because this is a repeated game, parties may decide to enter into agreements so that the ruling party will not exploit the minority. The current party knows that in the future the minority party may come to power and punish the current party for wealth-reducing policies. Under a simple majority rule, party dominance is more likely to change than under supermajority rule where the defecting party cannot be punished until the exploited party gains the necessary representation. A party changeover in power is less likely under a supermajority rule, because the minority power must obtain a larger vote share than it would under simple majority rule. Thus, the ruling party will be more willing to renege than it would be under a simple majority rule, because of the reduced chance of retaliation or the retaliation is far in the future.

Thus, we postulate that bicameralism offers a second advantage in protecting minority interests relative to a unicameral chamber with a supermajority rule. Bicameral chambers composed of opposing interests can continually enforce legislative agreements rather than relying on the threat of retaliation to enforce legislative bargains over time, which is the only enforcement option in unicameral legislatures. If different interests control the chambers then the parties do not have to enforce legislative agreements through retaliation over time. A veto player limits the likelihood of minority exploitation in each period.

Bicameral differences decrease redistributive spending that results from legislative instability. By raising the decision-making costs through adding a veto player, redistributive spending will decline similar to requiring a supermajority in a unicameral chamber. In the context of measuring bicameralism as a continuous variable, as the chamber preferences become more dissimilar, the check on redistribution rises, resulting in a predictable decline in tyranny of majority spending. However, a bicameral system uniquely avoids problems imposed by a unicameral chamber with a supermajority rule. First, stable Condorcet-winning policies that are favored by a majority in both chambers will succeed, whereas they would fail in the alternate institution framework. Second, if coalitions restrain redistributive spending because of rotating majorities, a supermajority rule may remove this constraint.

Bicameral differences restrict the ability of coalitions to manipulate fiscal policy for redistributive purposes. This limitation has a secondary effect on government spending on public goods. First, tax revenues that may have been used for redistributive spending can now be used for public goods. Second, citizens may be willing to accept higher taxes because the revenue cannot be used for redistribution. Thus, bicameral differences should reduce government expenditures on redistributive programs, but increase government spending on nonredistributive programs.

3. Empirical Model

Data and Specification Issues

The theoretical case that bicameralism acts as a check on majority tyranny is quite solid, yet lacks empirical scrutiny. The two arguments developed above imply that bicameralism should constrain fiscal agreements that redistribute income from one constituent group to another. In this section we use demographic measures of chamber constituencies to examine this implication empirically.

The analysis treats bicameralism not as a discrete structure, but rather as a continuous institutional arrangement based on the degree of constituent homogeneity across chambers. This ranges from identically composed chambers to totally different bases of representation across the chambers. In this perspective, as the chambers' demographic characteristics become more diverse the legislature becomes increasingly "bicameral." The main implication we address relates to the impact of varying degrees of bicameralism on state fiscal performance. Specifically, increasing degrees of bicameralism should act as a break on redistributive fiscal policies. This implication is consistent with classic and modem theory: As the two chambers become increasingly similar, or less bicameral in our framework, policy choices approach the unicameral outcome. This follows both from the direct effect of different constituencies drawing from a common tax base, and from the indirect effect of reducing the probability of achieving non-Co ndorcet outcomes.

To implement the empirical analysis we first construct four alternative measures of bicameralism on the basis of proxies of the median voter of each chamber. (6) Each measure uses a specific economic or demographic constituent characteristic, including average household income, the percentage of constituents with incomes greater than $50,000, the percentage of constituents receiving Social Security benefits, and the percentage of constituents employed in the manufacturing sector. Symbolically, the bicameral measure is [absolute value of (([X.sub.L] - [X.sub.U])/[X.sub.L]], where [X.sub.L] and [X.sub.U] denote the demographic characteristic of the median legislator in the lower chamber and upper chamber. (7) For example, consider the bicameralism measure based on average household income. We find the median legislative district in each chamber with respect to household income. We then compute the percentage difference between these two median incomes. Increases in this value thus reflect greater disparity between the two chamber s. We note that prior studies have used constituent income to proxy of legislative preferences of the median voter (e.g., Inman 1978 and Pommerehne 1978). The latter three measures proxy constituencies that hold common interests in government transfers. For example, the percentage of constituents with incomes greater than $50,000 proxies the potential demand for transfers from high- to low-income families. (8) The other two measures reflect the degree of interchamber similarity with respect to age-related and job-related policy preferences.

Equation 1 specifies the empirical model we use to test the hypothesis.

[G.sub.it] = [THETA][H.sub.it] + [PHI][X.sub.it] + [[tau].sub.t] + [[eta].sub.i] + [[member of].sub.it] (1)

where the subscripts i and t denote an observation for an individual state in a particular year. The dependent variable, [G.sub.it] is total state government spending (denominated in real per capita terms) from 1994 to 1997. (9) We estimate Equation 1 using weighted least squares with cross-section weights to correct for detected heteroscedasticity. Separate models examine three main categories of state spending: public welfare, education, and highways. We also include estimates for total state spending.

[H.sub.it], in Equation 1 represents a vector of variables that proxy constituency difference across the two legislative chambers in the American states. It includes the bicameral measures described above, the demographic variables for each chamber from which we compute the bicameralism variable, and a dummy variable equal to one where the legislative districts for the two chambers are identical. We initially focus on the interchamber income metric. Then we extend the analysis to examine other demographic measures of constituencies in chambers that are jointly influenced by transfers and rent-seeking. Chambers with constituents of similar demographic median representation should have higher levels of spending than chambers with more different constituent preferences. We include the demographic characteristics of the separate chambers to control for abnormally high differences that could influence the bicameralism variable. (10) Five states select representatives for both chambers from the same electoral distr icts, which means that the bicameral variable (the percent difference of the chamber medians) in these cases equals zero. (11) We handle the problem in several ways to ensure robustness. First, in preliminary models we excluded the five states from the sample altogether. These results did not differ materially from those reported in the Tables 1-7 below in which they are included. Second, we include a "Same Districts" dummy variable equal to one for the five states that share legislative districts for both chambers to check for a discrete effect. These results are listed in Table 3. This method has the advantage of comparing these states that should be the least bicameral with other states within the same regression. Third, we examine specifications that include a squared bicameral term to control for nonlinearity in the relation between bicameralism and government expenditures. A Wald coefficient test indicates that the inclusion of the squared term is necessary; thus, we report only the results that include the squared bicameral term.

[X.sub.it] is a vector of six control variables: state median household income, log of state population, population growth, the percentage of the population between the ages of 17 and 64, the unemployment rate, and a measure of state constituent ideology. [[tau].sub.t] is a vector of fixed year effects,[[eta].sub.i] is a vector of state-specific time trends that captures any time trends in government spending specific to individual states, and [[member of].sub.it], is the error term. State median household income proxies the median voter of the state, and therefore the district configuration variables capture the difference between statewide and district income. We expect the demand for government spending to rise with income in accordance with Wagner's Law; however, the support for a direct relation between income and the demand for government spending is mixed. (12) The unemployment rate proxies potential claims for unemployment insurance and related welfare programs. Population and population growth are incl uded to control for economies of scale in government expenditures. As population rises, additional government spending per capita should decline. The percentage of the population between the ages of 17 and 64 proxies the share of the population that is neither old nor young. That is, old and young constituents tend to be the largest consumers of government programs, and we anticipate a negative correlation between this variable and state spending. In addition, a larger work force means there are fewer Social Security recipients in the state population; therefore, a larger working age population should be associated with lower government spending because of the lower demand for redistributive transfers. We use the citizen ideology index of Berry et al. (1998) to proxy the constituent ideology of the state. Index scores are positively correlated with the citizen taste for government spending in the state. Table 1 lists the summary statistics for all of the variables.

4. Results

Tables 2 and 3 list the results for the income measure of bicameralism for the major budget components and for total expenditures. (13) First we regress expenditures on a baseline model that excludes the constituent homogeneity variables, year effects, and state-specific time trends. The baseline model (not reported) explains close to 50% of the variation in public welfare and highway spending; however, it only explains 5% of the variation in education expenditures. In the total expenditures model, the control variables explain almost 60% of the variation.

For the public welfare component (Table 2), bicameralism is negatively related to expenditures. At the average, a 1% increase in bicameralism leads to a 0.027% decline ($0.19 per capita) in welfare spending. Considering that the theoretical work on bicameralism postulates a reduction in redistributive activity, this effect for the public welfare component is not surprising. Public welfare spending tends to reflect a pure redistribution, whereas the other spending components may provide some efficiency-enhancing public goods.

In support of this interpretation, the results indicate that bicameralism is positively related to education expenditures, highway expenditures, and total expenditures. In addition, states with the same legislative districts in both chambers (meaning they have the minimal bicameral check) experience high levels of total government expenditures and in these budget components. Using the mean level of government spending, a 1% increase in bicameralism leads to a 0.03% increase ($1.02 per capita) in total government spending, a 0.018% increase in highway spending ($0.05 per capita), and a 0.046% increase ($0.49 per capita) in education spending. Table 3 reports the regression results with the Same District dummy to isolate the effect of these states that have the same median voter in each chamber. The results for bicameralism do not change when we control for this factor. The states with the same electoral districts for both chambers are associated with a rather large increase in total expenditures of $160 per ca pita, a $14 increase in public welfare spending per capita, a $134 increase in education spending per capita, and a $49 increase in highway spending per capita.

Though general spending may be redistributive, we cannot easily identify this type of redistribution. Where we observe a specific redistributive program and can identify specific demographic interests aligned with these programs, on average bicameralism should reduce this type of political activity. That we do not observe similar behavior with regard to other types of spending programs indicates that general spending is more likely to be efficient than spending associated with redistribution. Recall that the theory predicts that bicameralism should make spending more efficient -- not simply limit the spending level -- by limiting agreement to the set of policies agreed upon by the median constituents of both chambers. This suggests that increased bicameralism may increase spending in some areas of the state budget but not others. As legislators become less able to seek transfers, they may devote more spending to efficient public goods where they agree. Matsusaka (1992) finds that states in which citizens have the ability to check their representatives (through voter initiative institutions), representatives pay more attention to efficiency compared with states without voter-initiative institutions. The results in Tables 2 through 5 suggest that bicameralism exerts a similar effect.

To examine the redistributive hypothesis in further detail, we use a second round of regression models that use measures to proxy specific transfer-seeking coalitions. (14) We use three different measures of bicameral differences driven by desire to allocate transfers to median legislative constituencies. Tables 4 through 7 list the effect of bicameral differences with respect to transfer-seeking coalitions. We do not report the results for specifications that include the Same Districts dummy because it is almost always insignificant. The results provide strong support for the contention that bicameral constituency differences limit redistributive spending. Bicameralism is negatively and significantly related to total government expenditures, and in all the spending components except for two specifications of highway spending.

Consistent with the median income measure of bicameralism, bicameralism as proxied by constituent groups has the strongest impact on public welfare spending. At the average level of spending, a 1% increase in bicameralism leads to between a 0.05 and 0.08% (approximately $0.50 per capita) decrease in public welfare expenditures. For education expenditures, which compose almost a third of total government expenditures, for each bicameral proxy a 1% increase in bicameralism leads to a 0.03% (approximately $0.33 per capita) decline in spending. The results for highway spending are mixed, ranging from a positive effect of 0.04% ($0.11 per capita) to a negative 0.06% ($0.16 per capita) change in expenditures. In total, a 1% increase in bicameral difference leads to a decline of between 0.005% ($0.19 per capita) and 0.03% ($1.04 per capita) in total spending.

5. Concluding Comments

Classical political economists such as Montesquieu and Madison considered bicameralism to be a critical legislative institution. With a few exceptions, modern scholarship tends to ignore the potential consequences of this institution. Tsebelis and Money (1997) point out that studies of political outcomes that ignore bicameral interaction neglect an important component and are therefore biased. In this paper we attempt to fill this void in the study of political institutions by examining the role of bicameralism in determining state fiscal policy. We construct several measures of bicameralism from demographic characteristics of American state legislatures, and on the basis of these measures we find evidence to support the claim that bicameral interaction has a predictable effect on fiscal policy. Different representative constituencies in bicameral chambers limit the use of fiscal policy for wealth transfers. Additionally, by reducing the ability of representatives to pursue redistributive transfers, legislato rs may direct spending to more efficient uses. The empirical results indicate that, on average, increasing the bicameral difference in terms of redistributive coalitions between chambers reduces government expenditures.

In sum, bicameral chambers serve the intended purpose of the Founding Fathers to limit government to the protection of the "general welfare." This has particularly strong policy implications for new and developing countries that seek to design constitutions that restrict the government from engaging in harmful redistributive activities.

(*.) Department of Economics, The University of the South, Sewanee, TN 37383, USA; corresponding author.

(+.) Center for Study of Public Choice, George Mason University, Fairfax, VA 22030, USA; E-mail mailto:[email protected].

We thank Noel Campbell, Joe Johnson, Thomas Stratmann, Todd Zywicki, and two anonymous referees for helpful comments. All remaining errors are our responsibility. The Center for Study for Public Choice provided financial support for this study.

Received October 2000; accepted May 2001.

(1.) Montesquieu (1997) first put forth the idea to use separate legislative chambers to limit the ability of one group to use its majority status to dominate the minority. Tsebelis and Rasch (1995) and Tsebelis and Money (1997) provide extensive discussions of the history and logic of bicameralism internationally and in the United States.

(2.) For examples, see Hammond and Miller (1987); Brennan and Hamlin (1992); Levmore (1992); Riker (1992); Persson, Roland, and Tabellini (1997); Tsebelis and Money (1997); and Diermeier and Myerson (1999).

(3.) For additional discussion on the effect of constituent homogeneity on policy outcomes see Buchanan and Tullock (1962), Stigler (1976), Crain (1979), Hayek (1979), McCormick and Tollison (1981), Tsebelis and Money (1997), and Bradbury and Crain (in press).

(4.) Madison comments on this issue when discussing bicameralism, stating, "[t]he internal effects of mutable policy ... poisons the blessings of liberty." Additionally, he notes that mutable legislation enhances the political influence of mischievous interest groups: "Another effect of public instability is the unreasonable advantage it gives to the sagacious, the enterprising and the moneyed few, over the industrious and the uninformed mass of people" (Hamilton, Jay, and Madison 1993, p. 249).

(5.) Also, see Besley and Coate (1998) for an alternative model of political compromise in a dynamic setting.

(6.) This is based on the standard assumption that the preferences of the median voter determine the equilibrium outcome in democracy. See Bradbury and Crain (2001) for a discussion of the median legislator as the median voter in representative democracy.

(7.) We obtain the data from Lilley, DeFranco, and Diefenderfer (1998). The sample includes 48 states for a total of 192 observations over the time period. We exclude Nebraska because of its unicameral legislature. Also, we exclude Alaska because of its heavy reliance on severance taxes.

(8.) See Meltzer and Richard (1981) and Krussel and Rios-Rull (1999) for models of transfers driven by income differences in the population.

(9.) We use the post-1990s redistricting period for which state district level data are available.

(10.) Including demographic variables for both chambers in the same regression does cause multicollinearity problems. However, because the object of analysis is the similarity between chambers, remedies such as removing one or both of the variables may induce bias in the model. For the majority of specifications the standard errors remain within the realm of statistical significance, indicating that there is little bias in the current specifications. Taking either variable out may cause more problems than the remedy of separate specifications.

(11.) The five states are Arizona, Idaho, New Jersey, North Dakota, and Washington. In these states, both chambers of the legislature are composed from the same set of legislative districts, and therefore the chamber medians are identical.

(12.) See Ram (1987) for a survey of empirical tests of Wagner's Law.

(13.) We do not report the results from the reduced sample because of the similarity of the results. Results for the reduced sample are available from the authors upon request.

(14.) We purposely use the term "transfer-seeking" instead of "rent-seeking." Even if no interest groups expend effort to influence representatives, reelection-maximizing representatives will prefer to allocate transfers to their constituents.

References

Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. Measuring Citizen and government ideology in the American States, 1960-93. American Journal of Political Science 42:327-48.

Besley, Timothy J., and Stephen Coate. 1998. Sources of inefficiency in a representative democracy: A dynamic analysis. American Economic Review 88:139-56.

Bradbury, John Charles, and W. Mark Crain. 2001. Who is the median voter in representative democracy? Unpublished paper, George Mason University.

Bradbury, John Charles, and W. Mark Cram. In press. Legislative organization and government spending: Cross-country evidence. Journal of Public Economics.

Brennan, Geoffrey, and Alan Hamlin. 1992. Bicameralism and majoritarian equilibrium. Public Choice 74:169-79.

Buchanan, James M., and Gordon Tullock. 1962. The calculus of consent: Logical foundations of constitutional democracy, Ann Arbor: University of Michigan Press.

Crain, W. Mark. 1979. Cost and output in the legislative firm. Journal of Legal Studies 8:607-21.

Diermeier, Daniel, and Roger B. Myerson. 1999. Bicameralism and its consequences for the internal organization of legislatures. American Economic Review 89:1182-96.

Dixit, Avinash K., Gene M. Grossman, and Faruk Gul. 2000. The dynamics of political compromise. Journal of Political Economy 108:531-68.

Gilligan, Thomas W., William J. Marshall, and Barry R. Weingast. 1989. Regulation and the theory of legislative choice: The Interstate Commerce Act of 1887. Journal of Law and Economics 32:35-61.

Hamilton, Alexander, John Jay, and James Madison. 1993. The federalist LXII. In The debate on the Constitution by Bernard Bailyn. New York: Library Classics of the United States, pp. 244-50.

Hammond, Thomas H., and Gary J. Miller. 1987. The core of the constitution. American, Political Science Review 81:1155-74.

Hayek, Friedrich A. 1979. Law, legislation and liberty; volume 3: The political order of a free people. Chicago: University of Chicago Press.

Inman, Robert P. 1978. Testing political economy's 'as if' proposition: Is the median voter really decisive? Public Choice 33:45-65.

Krussell, Per, and Jose-Victor Rios-Rull. 1999, On the size of US government: Political economy in the neoclassical growth model. American Economic Review 89:1156-81.

Levmore, Saul. 1992. Bicameralism: When are two decisions better than one? international Review of Law and Economics 12:145-62.

Lilley, William, Laurence J. DeFranco, and William M. Diefenderfer. 1998. Almanac of state legislatures. Washington, DC: Congressional Quarterly, Inc.

Matsusaka, John G. 1992. Economics of direct legislation. Quarterly Journal of Economics 107:541-71.

McCormick, Robert E., and Robert D. Tollison. 1981. Politics, legislation, and the economy. Boston: Martinus Nijhoff Publishing.

Meltzer, Allan H., and Scott F. Richard. 1981. A rational theory of the size of government. Journal of Political Economy 89:914-27.

Mill, John Stuart. 1950. Utilitarianism, liberty and representative government. New York: Polyglot Press.

Montesquieu, Charles de Secondat. 1997. The spirit of the laws. Cambridge, UK: Cambridge University Press.

Persson, Torsten, Gerard Roland, and Guido Tabellini. 1997. Separation of powers and political accountability. Quarterly Journal of Economics 112:1163-203.

Pommerehne, Werner W. 1978. Institutional approaches to public expenditure: Empirical evidence from Swiss municipalities. Journal of Public Economics 9:255-80.

Ram, Rati. 1987. Wagner's hypothesis in time-series and cross-section perspectives. Review of Economics and Statistics 69:194-204.

Riker, William H. 1992. The justification of Bicameralism. International Political Science Review 13:101-16.

Stigler, George J. 1976. The sizes of legislatures. Journal of Legal Studies 5:17-34.

Tsebelis, George, and Jeanette Money. 1997. Bicameralism. Cambridge, UK: Cambridge University Press.

Tsebelis, George, and Bjornerik Rasch. 1995. Patterns of bicameralism. In Parliaments and majority rule in Western Europe, edited by Herbert Doering. New York: St. Martin's, pp. 365-90.

Tullock, Gordon. 1959. Problems of majority voting. Journal of Political Economy 67:571-9.

Table 1

Summary Statistics

 Mean St. Dev Max. Min.

Spending components
 Total expenditures $3,363 $636 $5,374 $2,368
 Public welfare $701 $200 $1,521 $390
 Education $1,047 $207 $1,515 $511
 Highway $271 $85 $632 $142

Controls
 Median household income $35,737 $5,145 $48,736 $25,614
 In (Population) 8.14 0.99 10.38 6.16
 Population growth 0.01 0.01 0.053 -0.007
 Unemployment rate 5.17 1.14 8.9 2.5
 Population (ages 18-64) 61.11 1.57 64.5 56.4
 Berry ideology 45.64 13.97 84.04 7.67

Upper chamber medians
 Income $38,142 $6,914 $56,761 $27,495
 Income > 50 K (%) 20.13 7.66 41.00 11.00
 Social security (%) 27.02 3.10 35.00 21.00
 Manufacturing employment (%) 23.96 5.66 36.00 11.00

Lower chamber medians
 Income $37,693 $6,786 $55,003 $27,321
 Income > 50 K (%) 19.77 7.61 40.00 10.00
 Social Security (%) 27.00 3.28 35.00 19.00
 Manufacturing employment (%) 24.02 5.68 36.00 11.00

Bicameralism measures
 Bicameralism (income) 0.018 0.019 0.085 0.000
 Bicameralism (income > 50 K) 0.043 0.052 0.250 0.000
 Bicameralism (Social Secrity) 0.017 0.026 0.105 0.000
 Bicameralism (manufacturing) 0.023 0.029 0.125 0.000
Table 2

Effect of Bicameralism on Expenditures

 Public
 Welfare Education
 2 2

Median house- 4.09E-03 -6.67E-03
 hold income (1.64) (*) (2.31) (**)
In(Population) 38.18 -94.28
 (5.20) (***) (7.96) (***)
Population -6397.62 287.44
 growth (11.84) (***) (0.27)
Unemployment 47.97 1.56
 rate (11.69) (***) (0.21)
Population -4.50 -1.78
 (ages 18-64) (1.04) (0.34)
Berry citizen 6.08 -2.99
 ideology (10.25) (***) (3.77) (***)
Upper median 0.02 -0.08
 district income (1.19) (5.05) (***)
Lower median -0.02 0.08
 district income (1.22) (5.37) (***)
Bicameralism -3730.03 4140.76
 (income) (3.37) (***) (2.81) (***)
Bicameralism 73011.73 -40831.80
 (income) squared (3.99) (***) (1.33)
Constant 76.60 1979.21
 (0.33) (6.59) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.75 0.59
F-Statistic 1340.6 499.2
Observations 192 192


 Highway Total Exp.
 3 4

Median house- -1.52E-03 1.15E-03
 hold income (1.75) (*) (0.17)
In(Population) -57.35 -238.55
 (14.82) (***) (10.12) (***)
Population 84.89 -11635.70
 growth (0.27) (8.03) (***)
Unemployment -0.01 134.66
 rate (4.64E-03) (8.77)
Population -5.56 -61.17
 (ages 18-64) (2.81) (***) (5.00) (***)
Berry citizen 0.45 15.48
 ideology (1.55) (8.48) (***)
Upper median 8.87E-03 -0.08
 district income (2.22) (**) (2.30) (**)
Lower median -8.24E-03 0.12
 district income (2.00) (**) (3.48) (***)
Bicameralism 1084.90 5976.21
 (income) (2.79) (***) (1.97) (**)
Bicameralism -22030.54 -12515.48
 (income) squared (3.58) (***) (0.29)
Constant 1073.83 5822.37
 (9.83) (***) (8.67) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.69 0.79
F-Statistic 283.7 2023.2
Observations 192 192

Absolute t-statistics in parentheses.

(***)Significant at the 1% level.

(**)Significant at the 5% level.

(*)Significant at the 10% level.
Table 3

Effect of Bicameralism on Expenditures: Robustness Table

 Public
 Welfare Education
 2 2

Median house- 3.92E-03 -5.85E-03
 hold income (1.60) (2.24) (**)
In(Population) 39.35 -89.54
 (5.33) (***) (7.49) (***)
Population -6447.98 -823.88
 growth (11.77) (***) (0.77)
Unemployment 47.90 3.47
 rate (11.94) (***) (0.50)
Population -3.82 1.33
 (ages 18-64) (0.88) (0.22)
Berry citizen 6.15 -3.54
 ideology (10.59) (***) (4.45) (***)
Upper median 0.02 -0.08
 district income (1.21) (5.40) (***)
Lower median -0.02 0.09
 district income (1.25) (5.50) (***)
Bicameralism -3492.31 7303.58
 (income) (3.01) (***) (5.20) (***)
Bicameralism 69927.11 -73582.24
 (income) squared (3.67) (***) (2.63) (***)
Same districts 13.70 134.72
 dummy (0.64) (4.50) (***)
Constant 32.91 1805.47
 (0.14) (5.16) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.74 0.60
F-Statistic 1893.7 1278.3
Observations 192 192


 Highway Total Exp.
 3 4

Median house- -l.23E-03 -3.11E-03
 hold income (1.38) (0.47)
In(Population) -53.59 -225.15
 (13.31) (***) (9.67) (***)
Population 27.97 -10634.70
 growth (0.09) (7.75) (***)
Unemployment -3.25 131.91
 rate (1.02) (8.50) (***)
Population -2.54 -66.17
 (ages 18-64) (1.21) (5.79) (***)
Berry citizen 0.72 15.61
 ideology (2.32) (**) (8.21) (***)
Upper median 8.58E-03 -0.07
 district income (2.09) (**) (2.13) (**)
Lower median -8.85E-03 0.12
 district income (2.08) (**) (3.34) (***)
Bicameralism 1551.91 9896.41
 (income) (3.99) (***) (3.13) (***)
Bicameralism -27388.98 -57614.88
 (income) squared (4.72) (***) (1.35)
Same districts 48.62 160.08
 dummy (3.93) (***) (2.13) (**)
Constant 877.87 6043.11
 (7.50) (***) (9.64) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.70 0.79
F-Statistic 264.6 2515.6
Observations 192 192

Absolute t-statistics in parenthese.

(***)Significant at the 1% level.

(**)Significant at the 5% level.
Table 4

Effect of Different Demographic Measures of Bicameralism on Public
Welfare Expenditures

 YHH > 50 K1 %S.S.
 1 2

Median household income 7.93E-04 4.54E-03
 (0.33) (2.43) (**)
In(Population) 40.69 35.97
 (6.82) (***) (4.45) (***)
Population growth -5716.55 -5666.88
 (9.56) (***) (8.32) (***)
Unemployment rate 37.18 46.41
 (8.09) (***) (8.16) (***)
Population (ages 18-64) 4.26 8.41
 (0.95) (1.62)
Berry citizen ideology 6.15 6.27
 (12.04) (***) (8.26) (***)
Upper median district 39.08 -6.32
 (4.47) (***) (0.72)
Lower median district -38.15 3.44
 (4.18) (***) (0.42)
Bicameralism -1124.51 -4405.95
 (4.00) (***) (6.61) (***)
Bicameralism squared 4138.10 30697.75
 (1.58) (3.77) (***)
Constant -349.37 -631.68
 (1.27) (1.86) (*)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.57 0.77
F-statistic 1337 136.9
Observations 192 192

 %Mnfg.
 3

Median household income 4.28E-03
 (3.05) (***)
In(Population) 50.97
 (6.72) (***)
Population growth -5680.78
 (1l.87) (***)
Unemployment rate 16.35
 (2.88) (***)
Population (ages 18-64) 12.11
 (2.86) (***)
Berry citizen ideology 4.72
 (8.82) (***)
Upper median district 34.98
 (4.83) (***)
Lower median district -31.97
 (4.62) (***)
Bicameralism -4912.91
 (9.82) (***)
Bicameralism squared 52131.51
 (9.44) (***)
Constant -885.32
 (3.83) (***)
State specific time trend Yes
Fixed-year effects Yes
Adjusted R-squared 0.74
F-statistic 435.5
Observations 192

Absolute t-statistics in parentheses.

(***)Significant at the 1% level.

(**)Significant at the 5% level.

(*)Significant at the 10% level.
Table 5

Effect of Different Demographic Measures of Bicameralism on Education
Expenditures

 YHH > 50 K %S. S.
 1 2

Median household income 5.21E-03 (3.39) (***)
 (2.57) (***) -68.03
In(Population) -62.38 (7.22) (***)
 (6.87) (***) -1381.52
Population growth -2351.52
 (2.82) (***) (1.41)
Unemployment rate -2.72 4.08
 (0.44) (0.52)
Population (ages 18-64) 4.29 -15.85
 (0.96) (3.19) (***)
Berry citizen ideology -0.58 1.46
 (0.97) (1.99) (**)
Upper median district -52.73 -35.14
 (6.91) (***) (3.66) (***)
Lower median district 47.43 10.03
 (6.25) (***) (1.09)
Bicameralism -2004.98 -2202.83
 (6.11) (***) (3.48) (***)
Bicameralism squared 14495.01 17748.83
 (4.37) (***) (2.31) (**)
Constant 1284.70 3396.54
 (5.18) (***) (9.64) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.56 0.57
F-statistic 2810.5 517.9
Observations 192 192

 % Mnfg.
 3

Median household income -3.20E-03
 (2.36) (**)
In(Population) -44.63
 (5.51) (***)
Population growth -540.11
 (0.68)
Unemployment rate -4.02
 (0.74)
Population (ages 18-64) 12.25
 (2.77) (***)
Berry citizen ideology -0.89
 (1.82) (*)
Upper median district 10.94
 (2.41) (**)
Lower median district -21.00
 (4.33) (***)
Bicameralism -3313.34
 (6.62) (***)
Bicameralism squared 31438.23
 (5.13) (***)
Constant 1082.43
 (4.85) (***)
State specific time trend Yes
Fixed-year effects Yes
Adjusted R-squared 0.57
F-statistic 3217.4
Observations 192

Absolute t-statistics in parentheses.

(***)Significant at the 1% level.

(**)Significant at the 5% level.

(*)Significant at the 10% level.
Table 6

Effect of Different Demographic Measures of Bicameralism on Highway
Expenditures

 YHH > 50 K %S.S.
 1 2

Median household income -1.28E-03 -8.64E-04
 (1.48) (1.16)
In(Population) -52.20 -52.56
 (15.71) (***) (14.81) (***)
Population growth 48.04 -197.69
 (0.20) (0.69)
Unemployment rate -2.53 -1.01
 (0.88) (0.38)
Population (ages 18-64) -6.57 -3.93
 (3.21) (***) (1.92) (*)
Berry citizen ideology 0.59 0.44
 (.17) (**) (1.46)
Upper median district -7.38 9.16
 (3.26) (2.60) (***)
Lower median district 8.17 -8.60
 (3.53) (***) (2.68) (***)
Bicameralism 80.82 546.29
 (0.59) (2.19) (**)
Bicameralism squared 2054.00 -6579.03
 (1.80) (*) (2.56) (***)
Constant 1098.09 932.50
 (8.71) (***) (6.27) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.68 0.67
F-statistic 286 233.7
Observations 192 192

 %Mnfg.
 3

Median household income -9.99E-04
 (1.67) (*)
In(Population) -47.21
 (15.31) (***)
Population growth -344.27
 (1.53)
Unemployment rate -5.32
 (2.57) (***)
Population (ages 18-64) 8.70
 (4.81) (***)
Berry citizen ideology -5.00E-03
 (0.01)
Upper median district -20.35
 (5.47) (***)
Lower median district 14.55
 (4.02) (***)
Bicameralism -1335.98
 (7.15) (***)
Bicameralism squared 12364.08
 (5.46) (***)
Constant 341.59
 (3.87) (***)
State specific time trend Yes
Fixed-year effects Yes
Adjusted R-squared 0.78
F-statistic 325.5
Observations 192

Absolute t-statistics in parentheses.

(***)Significant at the 1% level.

(**)Significant at the 5% level.

(*)Significant at the 10% level.
Table 7

Effect of Different Demographic Measures of Bicameralism on Total
Expenditures

 YHH > 50 K %S.S.
 1 2

Median household income 8.63E-03 0.04
 (1.68) (*) (8.06) (***)
ln(Population) -209.69 -234.92
 (10.07) (***) (11.03) (***)
Population growth -10944.51 -15318.17
 (7.06) (***) (7.90) (***)
Unemployment rate 111.66 162.25
 (7.87) (***) (9.59) (***)
Population (ages 18-64) -56.75 -8.67
 (5.71) (***) (0.79)
Berry citizen ideology 16.60 19.72
 (10.29) (***) (11.07) (***)
Upper median district 29.97 -133.05
 (1.32) (4.80) (***)
Lower median district 4.11 86.98
 (0.18) (3.36) (***)
Bicameralism -1705.19 -6608.93
 (2.38) (**) (4.34) (***)
Bicameralism squared 14422.68 13764.11
 (2.70) (***) (0.73)
Constant 6200.85 4077.53
 (10.84) (***) (5.20) (***)
State specific time trend Yes Yes
Fixed-year effects Yes Yes
Adjusted R-squared 0.79 0.80
F-statistic 5311.3 2586.4
Observations 192 192

 %Mnfg.
 3

Median household income 0.04
 (9.39) (***)
ln(Population) -221.98
 (7.44) (***)
Population growth -16194.71
 (8.28) (***)
Unemployment rate 192.60
 (10.80) (***)
Population (ages 18-64) 24.87
 (1.53)
Berry citizen ideology 14.17
 (6.62) (***)
Upper median district -38.49
 (1.31)
Lower median district 17.13
 (0.62)
Bicameralism -3777.24
 (2.34) (**)
Bicameralism squared -2166.26
 (0.12)
Constant 1028.08
 (1.36)
State specific time trend Yes
Fixed-year effects Yes
Adjusted R-squared 0.79
F-statistic 544.4
Observations 192

Absolute t-statistics in parentheses.

(***)Significant at the 1% level.

(**)Significant at the 5% level.

(*)Significant at the 10% level.