The Literature on Legislative Salaries

The few academic studies that examine legislative compensation all look at American state legislatures. Although meagre in number, these works offer some interesting results. Economists McCormick and Tollison conceive of legislatures as unions and find that wages in the mid-1970s were considerably higher in bodies where members could set their own pay than in bodies where wages were set externally, typically by a constitution (1978).¹

Analyzing legislative compensation through the prism of legislative professionalization, political scientists Moncrief (1988) and Squire and Hamm (2005) come to slightly different conclusions from those reached by the economists. Moncrief reports that state per capita income is the only substantively and statistically significant variable explaining legislative compensation at four time points between the 1950s and the 1980s (1988). Squire and Hamm arrive at a similar conclusion (2005: 88–92). Examining five different time points between 1910 and 1999, they find that total state income accounts for differences in legislative wages cross-sectionally at each time point and it also explains the cross-sectional change in wages from one time point to the next. Both Moncrief (1988) and Squire and Hamm (2005) devise arguments that link the level of policy demands made by a state's electorate, as suggested by a state's per capita income or total state income, to legislator wages, arguing higher demands are associated with higher pay.

It is this latter line of reasoning that I pursue. As Squire and Hamm observe, “It is very easy to understand how 35 million Californians can more easily finance their 120 member state legislature at a generous level than 500,000 Wyomians can support their 90 legislators. Large populations generate more income that can be used to finance the legislature, and the costs are spread across more people” (2005: 86). From an economics perspective membership sizes of American state legislatures do not really vary much (Stigler, 1976)—and the same is true of the national and other subnational chambers studied here. But the costs of financing these legislatures are spread across national or state economies of vastly different sizes. Thus, as national or state wealth increases the per capita cost of financing a legislature declines, which in turn may produce a strong relationship between nation or state wealth and legislative compensation.

In addition to the capacity to finance legislative compensation, total wealth might tap another dimension as well. Generally speaking, where national or state wealth is greater, populations are larger and, as a result, usually more socially and economically diverse. This diversity generates more interests and potentially more conflict, leading to greater demands on government. It seems reasonable to expect that increased demands or workloads should produce support for higher levels of legislative compensation to attract better qualified people to government service. Consequently, total national or state wealth may in part capture the level of demands on the legislature as well as the nation or state's capacity to pay for legislative service.

Legislator Salaries in National Assemblies

To conduct a first test of the state wealth hypothesis I collected 2005 legislative salary data from 35 national assemblies, 20 in countries deemed advanced economies by International Monetary Funds standards, and 15 in countries considered emerging market economies.²

Explaining National Legislator Salaries

What explains national differences in legislative salaries? The argument advanced above offers a simple theory: wealthier nations pay more than poorer nations. But if, as I argue, wealth matters, the question then becomes whether that means total national wealth or per capita wealth. I argue on theoretical grounds that it should be total wealth and not per capita wealth that explains differences in legislative pay. I take this position because given that legislatures are of roughly equal membership sizes, a nation or state with considerable total wealth, even if it is only of moderate per capita wealth, will still be able to generate more revenue with which to support its legislators than will a country or state with less total wealth, even if it is wealthier on a per capita basis.

A series of OLS (ordinary least squares) equations testing this theory are presented in Table 1. In each equation, legislative salary is the dependent variable. Equation 1 examines all 35 countries, with only gross domestic product and per capita gross domestic product entered as independent variables.³

National Assembly Legislator Salary, Controlling for Gross Domestic Product (OLS regressions)

Equation 2 examines the influence of country population size on legislative salaries. Population is an obvious alternative to gross domestic product as a measure of country size. Because population is not highly correlated with either gross domestic product or per capita gross domestic product in this sample of countries, it can be entered in an equation to assess its impact on legislative salaries.⁵

There is, however, another notable distinction among the countries in the study that might well influence legislative salaries: some have what the International Monetary Fund (IMF) considers to be advanced economies while the others have emerging market economies. Equation 3 includes a simple dummy variable for an advanced economy along with the gross domestic product and per capita domestic product variables. The coefficient for the advanced economy variable is statistically significant and very large. Controlling for both measures of national wealth, the difference in legislative salary between an advanced economy country and an emerging market economy country is $90,128. Importantly, the coefficient for gross domestic product, the variable of theoretical interest, remains both statistically significant and substantively powerful. The coefficient for per capita wealth, however, takes the wrong sign and fails to achieve statistical significance. Overall, equation 3 performs quite well, with an adjusted R² of .770.

The results of equation 3 raise the possibility that explanations for legislative salaries in countries with advanced economies may differ from explanations for countries with emerging market economies, suggesting the need to run two separate equations (Pindyck and Rubinfeld, 1991: 223). Equation 4 reruns equation 1 on the subset of countries with advanced economies. As predicted, legislative salaries in these countries appear tied to total gross domestic product. The coefficient for that variable is statistically significant and large, with a predicted difference of $87,080 between the largest and smallest advanced economies. The variable for per capita wealth takes the incorrect sign and is not statistically significant.⁷

In contrast to equation 4, the results of equation 5 reveal that neither gross domestic product nor per capita gross domestic product influences legislative salaries in emerging market economy countries. Both coefficients take the wrong sign and do not approach statistical significance. Moreover, the equation explains virtually none of the variance.⁸

Equations 6 and 7 present a different test of the basic theory examined here. Equation 6 is run on data from the 23 OECD countries contained in the larger data set, while equation 7 is run on the full data set. Both equations examine whether legislative salaries are linked to the financial size of the government.⁹

The results of equations 6 and 7 throw cold water on this hypothesis. The coefficient for total government revenues as a percentage of gross domestic product takes the predicted sign, but it is substantively small and very far from achieving statistical significance. The same is true for the government size coefficient. Both signal that the level of demands on the government does not drive legislative salaries. Only the coefficient for gross domestic product is statistically significant in both equations, and once again, it is substantively impressive in each.¹⁰

Overall, having attacked my simple state wealth theory from a number of different angles I find that, with the exception of the subset of developing economy countries, gross domestic product is strongly related to the size of the salaries paid to national legislators. For the most part, per capita wealth does not appear to play a role. Given the limited data set on which the theory has been tested, however, additional tests are required to confirm it.

Subnational Legislator Wages in Four Federal Systems

The second set of data I use to test the state size hypothesis is constructed from compensation paid to subnational legislators in four federal systems. The annual salaries paid to state or provincial legislators in Australia, Canada, Germany and the United States in 2005 are presented in Table 2. (Subnational assembly data sources are given in appendix B.) For each country the table provides the minimum and maximum salary offered by a state or provincial legislature and the median salary. Salaries are given both in the local currency and US dollars.

2005 State Assembly Salaries in Four Federal Systems

The highest median salary and the smallest gap between the highest salary and lowest salary are found in Australia. Parliamentary salaries in six of the eight Australian state and territorial legislatures are pegged to that of the federal parliament. In some cases the state or territorial salary is only slightly less then the federal parliament wage, being set lower by only 500 AUD in both New South Wales and Queensland. In other cases the difference is larger, the most extreme being Tasmania where wages are fixed at 85.19 per cent of the wage for the national parliament, creating a gap of 20,193 AUD in 2005.

Salaries are also comparatively high in the German Landtage. Indeed, the Landtag in North Rhine Westphalia offers the highest wage among the state or provincial legislatures in the four countries in the study: 114,000 EUR or $136,344, (or more than the salary paid in all but three of the 35 nations studied above). This sum is, however, potentially deceptive because it reflects a recent decision to lump together a number of allowances given the Landtag parliamentarians, some of which were intended to cover constituency service expenses and the like; thus it is not clear that this entire sum is to be used as salary by North Rhine Westphalia parliamentarians. The range of salaries is greater in Germany than in Australia, in part because of the considerably lower wages paid to members of the three city-state Landtage (Berlin, Bremen and Hamburg).

Members of provincial and territorial legislative assemblies in Canada are less well paid than their counterparts in Australia and Germany. The gap between the highest salary legislative assembly (Ontario, $69,848) and the lowest salary legislative assembly (Prince Edward Island, $38,821) is bigger than the gap among Australian state parliaments but less than the gap among German Lantage. There is little to distinguish the salaries paid members of the Canadian territorial assemblies from those paid to members of provincial assemblies; MLAs from the Northwest Territories are among the best paid, those from Nunavut are in the mid-range, and Yukon assembly members are near the bottom of the salary rankings.

Salaries in American state legislatures are markedly different from those paid to their counterparts in the other three countries. Most notably, the median salary in American state legislatures is much lower, representing only between 20 per cent of the salary earned by Australian state and territorial legislators and 31 per cent of the salary earned by members of the Canadian provincial and territorial legislative assemblies. And the range between the highest salary and the lowest salary is far greater than elsewhere. In 2005 state legislators in California were paid $99,000. In contrast, state legislators in New Hampshire earned only a $100 annually, a sum set in the state constitution and left unchanged since 1889. The difference between the two salaries is a staggering $98,900.

Explaining Subnational Legislative Salaries

Does state or province wealth account for the observed differences in legislative pay across the subnational bodies in each country? In Table 3 gross state or provincial product and per capita gross state product are correlated with state legislative pay in each of the four countries. Correlations are initially used because the number of cases in three of the countries is small. The relationships revealed by the correlations are consistent across the four countries. Large positive correlations are found between salaries and gross state product. These correlations easily surpass traditional levels of statistical significance in three of the four countries, the lone exception being Australia which only provides eight cases for analysis. But even the Australian correlation is large, in the expected direction, and close to achieving statistical significance.

Correlations of State Assembly Annual Salary with Gross State Product and Per Capita Gross State Product

In contrast, the correlations between salaries and per capita gross state product are much smaller and, in the case of Germany, the correlation is in the wrong direction. Moreover, the correlations are far from reaching traditional levels of statistical significance in Australia and Germany. Only in the American case are they statistically significant.

Clearly, the results in Table 3 support the hypothesis that state wealth, as measured by gross product, is associated with legislative salaries. The existing literature, however, raises several competing explanations which need to be tested. Only American case provides a sufficient number of cases to conduct fully specified analyses along these lines. There are, however, some simple multiple regression equations that can first be used to probe the strength of the state wealth-legislative salary relationship in the Canadian and German cases.

Table 4 provides the results of two OLS equations with Canadian provincial and territorial legislative assembly salaries as the dependent variable. In the first equation only gross product and per capita gross product are entered as independent variables.¹¹

Provincial Assembly Salary, Controlling for Gross Provincial Product, Per Capita Gross Provincial Product, and Provincial Status (OLS regressions)

I use similar equations to examine the German case. Table 5 presents the results of two OLS equations with Landtag salary as the dependent variable. In the first equation, only gross product and gross product per capita are entered as independent variables.¹²

Landtag Salary, Controlling for Gross Länder Product, Per Capita Gross Länder Product, and Länder Status (OLS regressions)

In the second equation two dummy variables are added. One dummy variable is for city-state status and the other dummy variable is for former German Democratic Republic (GDR) status. The rationales for the two dummy variables are straightforward. As noted above, the Landtage in Berlin, Bremen and Hamburg are paid considerably less than the other Landtage. Moreover, their functional status might arguably be considered somewhere between a city council and a state parliament. The dummy variable for city-state status checks whether the apparent state size-legislative salary relationship in Germany is really a product of their special status. Similarly, it seems reasonable to control for the former GDR states on the assumption that they might pay less well than their longer established western counterparts because of their dissimilar political histories.

The results of equation 2 confirm the importance of the state size hypothesis in the German case. Among the independent variables only the coefficient for gross product reaches traditional levels of statistical significance. Moreover, the coefficient's substantive impact is large and in the predicted direction. The wage difference between North Rhine Westphalia, the Land with largest gross product, and Bremen, the Land with the smallest gross product, works out to 48,343 EUR. Although the coefficients for the other independent variables are not statistically significant, it is worth noting that the coefficient for city-state status just misses meeting traditional levels of significance, takes the predicted negative sign, and suggests that the three city states pay lower wages by 17,365 EUR. Finally, the adjusted R² in each equation shows that these simple models account for an impressive amount of the variance.

American state legislatures, of course, provide 50 cases to analyze, enough for a more rigorous test of the state wealth-legislative salary hypothesis. As noted earlier, Moncrief (1988), Sollars (1990; 1994), and Squire and Hamm (2005) report a positive relationship between legislative pay and the size of state economies measured in various ways, findings that are compatible with the state wealth-legislative salaries hypothesis. But alternative explanations can be offered. McCormick and Tollison find that legislative control of wages leads to higher salaries (1978), while Sollars finds that pay increases with legislative session length (1990; 1994). State partisanship might also be related to legislative salaries. Democrats are arguably more supportive of higher government spending than are Republicans which may translate into backing for higher legislative wages. Finally, region must be controlled to account for any statistical residue of the argument that the “South is different” (Fiorina, 1997: 156–57). The expectation is that, everything else being held equal, the states in the South will pay lower wages to their state legislators than states in the rest of the country, largely for cultural and historical reasons.

OLS equations with state legislative salary in the United States as the dependent variable are presented in Table 6. Again, the variable of interest is gross product. Also entered are per capita gross state product, a dummy variable for whether a state legislature controls setting its own pay, the number of days the legislature meets on an annual basis, the vote percentage for the Democratic presidential candidate in the state in the 2004 election, and a dummy variable for the eleven states of the Confederacy.¹³

American State Legislative Salary, Controlling for Gross State Product, Per Capita State Product, Partisan Leaning, Region, and Legislative Characteristics (OLS regressions)

Equation 1 again gives results of the simple, two-variable regression model. The coefficient for gross state product is large and statistically significant; the coefficient for per capita gross state product is in the right direction but fails to achieve statistical significance. The results of the more fully specified equation 2 further reinforce the state wealth-legislative salary effect. Even with the host of control variables entered in the equation the coefficient for gross state product is statistically and substantively significant. Indeed, the predicted wage gap between the state with the largest gross product (California) and the state with the smallest gross product (Vermont) is huge: $67,145. Although several other coefficients also are statistically significant, each is of less substantive importance. The predicted difference in legislative pay between the state with longest legislative session and the state with the shortest legislative session computes to $27,366. Similarly the most Democratic state pays $20,322 more than the least Democratic state. And states in the South pay $12,873 less than the other states. Consistent with Sollars' findings (1994), the coefficient for legislative control of setting its pay is statistically insignificant and even takes the wrong sign. Finally, once again per capita gross state product falls far short of statistical significance.

Conclusions

The analyses presented here strongly suggest that differences in both national legislative salaries and subnational legislative salaries are explained in large part by differences in total state wealth. Simply stated, nations or states with larger gross products pay their legislators more. The strength of this finding is impressive. It is found cross-nationally across a range of different size economies, as well as subnationally within each of four different federal systems. Thus a state wealth-legislative compensation effect is confirmed.

Perhaps this finding should not be surprising because arguably there is an analogy between state wealth and legislative salaries and a well-established relationship in labour economics between firm size and wages. First uncovered by Moore (1911), the positive association between firm size and wages has been confirmed many times (Brown and Medoff, 1989; Hollister, 2004; Mellow, 1982), and has held in studies of countries other than the United States, notably in Australia (Izan et al., 1998; Wooden and Bora, 1999); Canada (Morissette, 1993: Zhou, 2000), and Germany (Schmidt and Zimmerman, 1991), as well as in the United Kingdom (Cosh, 1975) and Belgium (Lallemand et al., 2005). Although firm size in these studies is usually measured by the number of employees, which would suggest population as the appropriate analog, several (Cosh, 1975; Izan et al., 1998; Lambert et al., 1991; Kostiuk, 1990; Zhou, 2000) use total assets or sales revenue instead, measures which correspond more closely to gross product measures. Moreover, firms and nations have been equated by Dahl (2001: 250) and Lake and Baum (2001: 590). Consequently, it seems reasonable to consider legislative salaries from the perspective that nations or states are motivated to offer legislators salaries for many of the same reasons that firms offer wages to employees.

Does the labour economics literature suggest any additional reasons beyond those tested here to explain why such relationships appear? Surprisingly, it provides very few. As one exhaustive attempt to pin down encompassing explanations concluded, “In all cases, there still remains a large, significant and unexplained employer size-wage premium” (Troske, 1999: 25). One hypothesis that has received little empirical backing links higher wages to efforts by larger firms to lower employee performance monitoring costs. Somewhat stronger empirical evidence has been uncovered in support of the related proposition that larger firms hire better skilled workers accounting for their higher wages (Brown and Medoff, 1989; Troske, 1999). Unfortunately, neither of these notions translates well into the legislative setting.

Another explanation has surfaced in labour economics but been left untested. Bayard and Troske mention the possibility the firm-size wage effect results from the fact “that large employers simply have more output over which to amortize large sunk-cost investment” (1999: 103). This explanation, of course, fits with the theory offered here. Given that legislatures do not really vary much in membership size, wealthier states are better able to finance their legislators than are less wealthy states. Although I speculated that there may be more to the wealth-salary relationship than the simple idea that wealth makes it easier to pay legislator more, the analyses presented here provide little support for the notion that increased demands on the political system require governments to pay their legislators higher salaries as a means to attracting better qualified people, at least at the national level. In the end, the findings produced here document the simple but powerful finding that bigger nations or states pay more.