Later in this chapter, a series of exercises will help you test the theory of representative government. But before you begin to perform those exercises you might want to arm yourself with the knowledge of how others have studied legislative behavior.
The study of legislative behavior has proceeded on four major fronts: party, interest groups, region or area, and policy issues.
Since Lowell's 1902 work, scholars have intensified their interest in systematically studying "The Influence of Party Upon Legislation." Most research has confirmed Turner's (1951) conclusion that party is the dominant factor in explaining voting behavior in Congress and others have confirmed similar results in many state legislatures (Sorauf 1962; Hevasi 1975; Gove and Carlsen 1976; Kingdon 1977; Kirkpatrick 1978). Do these generalizations hold true today in the California Legislature?
Further researchers examine legislative votes for intra-party cohesion and inter-party differences. They compare cohesion and conflict between two chambers of the same legislature, between one legislature in one place with another in another place or at a different time, and between one policy area and another. Jewell (1982) concludes that "Party discipline is likely to be stronger in states where each party represents a relatively homogeneous set of interests and where there are clear differences between the two sets" (6).
Can you test how California legislators actually behaved in such situations? How much partisan cohesion and conflict do you detect? Is each party fairly unified internally while significantly different from the other party? How much party conflict should there be?
When there are differences between legislators' party positions and their constituents' interests, which way do they vote? In your opinion which way should they vote? Does it make a difference whether the legislator was elected by a lot or a little? Some researchers hypothesized that legislators who were elected by close margins and wish to be reelected are more likely to respond to their constituents' concerns, while those elected by a large amount can afford to support their party even if it means voting against their constituents interests (Pesonen, 1963; LeBlanc 1969). Fiorina rejects this notion (1974).
In one way partisan cleavages may be viewed as a manifestation of party as a coalition of interest groups. In another sense partisan differences may be compared to divisions based on other factors such as group interests, region, ideology, and policies. To what degree are the ratings of legislators by various interest groups related to the party affiliation of the legislators? Is party as strong a factor in explaining voting cleavages today as it was in the past? If not, is the reduction of party division in legislative voting associated with other factors related to the decline of party such as a weakening of strong party identification in the electorate and less party involvement in conducting campaigns and raising funds?
Since the results of the first studies of elections employing survey research (Lazarsfeld et al. 1944; Campbell et al. 1952; Campbell and Kahn 1952; Berelson et al. 1954), political scientists have hypothesized that voters behave politically as they are socially. Thus demographic characteristics of voters are often used as explanatory variables. The group basis of politics, noted early by Bentley (1908) and Odegard (1928) was rediscovered and emphasized in the 1950s (Truman 1951; Latham 1952).
Can you use the demographics and group ratings in this module to measure what influence interest groups might have on legislators? What is the difference between latent interests in the constituency and those that are manifest in the demands of organized groups? Kingdon (1977) believed that constituency interest was the first cue congressmen looked for, although they often considered other matters as well. Does the voting pattern of legislators from districts with more homogeneity of characteristics, such as ethnicity and income, differ from the behavior of legislators whose districts are more diverse? For example, if you were a legislator, would you feel you had more independence in casting your vote if your district was heterogeneous, where there was no dominant interest, latent or manifest? If so, what other cues would you take? Do heterogeneous groups result in less representation but better laws? Is this what Madison hoped would happen?
Region or Area
During the 1970s many studies of Congress and state legislatures indicated that region or territory may generate ideological divisions that translate into legislative votes. Schneider (1979) claims the ideology (liberal, conservative, and progressive) of Congress members explains most of their voting behavior. This was especially noticeable in the U.S. Congress when Southern Democrats voted with Republicans on many matters. Normative scholars are concerned with the potential conflict between local interests and statewide or national interests. In California what areas have different interests in California? Why? Do you want your legislators to vote your interests, or should they take into account what might be best for the state or the nation as a whole?
Collie (1985) points out that policy dimensions may be more pronounced in American legislatures, since following party lines is not required to maintain the regime. But in countries with parliamentary systems highly disciplined parties are necessary to maintain stable government and thereby contribute to strong party voting in the legislature. Further various studies have shown that cohesion varies from one policy to another. Lowi (1972) has concluded that the subject of a policy affects the alignment of votes on that policy, or "policies determine politics." Clausen (1973) has defined five policy areas (government management, social welfare, international involvement, civil liberties, and agricultural assistance) that he claims persist over time. Sinclair's work (1982) indicates that policy agendas change over time and voting alignments change on the same policies. What issues do you believe your legislator should consider most important? What issues are most salient to the various interests related to the demographic characteristics of the legislative districts and to the various interest groups whose ratings of legislative performance are provided in the module's data?
Now that you have some background about the substance of representation and legislative behavior you may proceed to develop your own research questions or test those posed in the following exercises.
Accompanying this module is a file called reprsp.por. This file is an SPSS system file containing data for the California legislature. A codebook describing these data also accompanies the module. (Note: any analysis you do should be carried out separately for the senate and the assembly, and the results compared.)
The underlying model for these exercises is a sort of chain of causality. The political ideology of an assembly or senate district (specifically, how liberal or conservative it is) is hypothesized to be influenced by the social, economic, and demographic characteristics of the district. The partisan inclinations of a district (Democrat or Republican) is thought to result in part from both the characteristics of the district and its political ideology. The sort of person chosen for the legislature by a district (especially, his or her party) is seen as a function of its characteristics, its political ideology, and its partisan leanings. Finally, the way a member votes on roll calls is seen as influenced by all of these other factors.
- District ideology
In the codebook, you will see that there is information on how each district voted on each of five propositions that appeared on the November 1996 ballot. There is also a variable, DISTLIB, that combines these five variables into an overall estimate of how liberal or conservative the district is. Calculate the correlations between DISTLIB, or any of the individual propositions, and district characteristics (urbanism, ethnicity, age distribution, type of household, education, income, home ownership -- i.e., variables URBAN through OWNOCC in your codebook). Examine the strongest correlations by producing scatterplots and least squares equations, with DISTLIB or individual propositions as dependent variables. Using multiple least squares, find out which variables are, other things being equal, the best predictors of district liberalism/conservatism.
- District partisanship
Repeat the above analysis, but use REGISTER (Democratic registration as percent of two-party [Democratic plus Republican] registration) as your dependent variable. Now add DISTLIB to the mix of independent variables. Which predicts registration patterns better, the social, economic, and demographic characteristics of the district, or the political ideology of the district. Does taking both factors into account give you a more complete explanation than you could get with just one?
- Election outcomes
This time, use LEGVOTE (Democratic percent of two-party vote in legislative races) as the dependent variable. What independent variable (district characteristic, ideology, or partisanship) best explains these election results? You probably guessed partisanship (i.e., REGISTER). Using multiple least squares, find out if other variables contribute anything directly to election results, or operate only indirectly through partisanship.
- Member characteristics
Using the mean and eta2, compare Democrats and Republicans in terms of district characteristics, ideology, and partisanship. To what extent do they represent different kinds of constituencies?
Repeat the analysis, but this time comparing male and female legislators and legislators of different ethnicities.
- Member voting behavior
ROLLCALL measures the voting behavior of representatives with an index that combines the ratings of five different interest groups, three conservative (the California Chamber of Commerce, Gun Owners of California, and the California Taxpayersí Association) and two liberal (the California League of Conservation Voters and the California Public Interest Research Group, also known as CALPIRG). All five ratings are highly correlated, positively or negatively. For example, a member getting a high rating from one conservative group almost invariably received high ratings from the other conservative groups and low ratings from the liberal groups. The combined measure is an index of conservatism: it has been scaled so that, in each chamber, the member with the most conservative score is coded 100, and the member with the most liberal score is coded 0.
Calculate the mean and standard deviation for this variable for the entire assembly or senate. Repeat, but with results broken down by party, and calculate eta2. (Here and elsewhere in this section, exclude the one independent, Senator Quentin Kopp, from the analysis so as not to distort the results.) You will see that, all by itself, party affiliation accounts for the bulk of the variation in membersí voting scores. In fact, a close examination will reveal that, in each chamber, the most conservative Democrat is more liberal than the most liberal Republican. The relationship between party and voting is so strong that there would be little point in comparing the impact of other variables with that of membersí party affiliation. Party would clearly dominate the results.
We can, however, reformulate the question. There is some variation within each party; some members within each party are somewhat more liberal or conservative than others. Again excluding the lone independent, calculate the least squares equation between ROLLCALL (dependent variable) and PARTY (independent), then compute the difference between each memberís actual score and the score predicted by his or her party. This will give you a new variable, the residual roll call voting score.
Are there any variables in your data that help, singly or in combination, explain the variance in this newly calculated variable? That is, what accounts for why some members are more liberal or conservative than you would predict based solely on their party?