Policy Agendas and Economic Inequality in American Politics

The Political Effects of Inequality

Large economic disparities have been linked to a number of economic and social problems, including worse population health (Wilkinson and Pickett, 2009), higher crime rates (Kaplan et al., 1996), heightened political polarization (Rosenthal et al., 2006), slower economic growth (Cynamon and Fazzari, 2014), and depressed political engagement among all but the richest citizens (Solt, 2008). When it comes to spending for political purposes, inequalities give a comparative advantage to economic elites and the recent erosion of campaign finance laws has greatly exacerbated this advantage (Winters and Page, 2009). More than ever, presidential and congressional hopefuls rely on a small number of rich political donors to finance their campaigns (Bartels, 2008; Bonica et al., 2013). In a report for the Sunlight Foundation, Drutman (2011) writes, “In a world of increasingly expensive campaigns, The One Percent of the One Percent effectively play the role of political gatekeepers. Prospective candidates need to be able to tap into these networks if they want to be taken seriously.” Furthermore, the advantage of the wealthy is compounded by the fact that the economically disadvantaged are notoriously difficult to mobilize politically, so the political capital of lower income groups does not increase proportionally to their numbers (Piven and Cloward, 1978). Indeed, Solt (2008) showed that the newly poor are less likely to participate politically than more economically advantaged groups.

Wealthy interests are willing and able to spend huge amounts of money on politics, but what are they buying? The answer has proven surprisingly elusive. Baumgartner et al. (2009) randomly sampled close to 100 public policy issues where lobbyists were active, interviewing key players about tactics, the make-up of coalitions, and financing. They found that more often than not, the lobbyists with the most financial backing lost the legislative fight. Their study joins a series of articles suggesting that the link between political spending, on either lobbying or campaigns, and concrete policy benefits is highly conditional (Ansolabehere et al., 2003; Witko, 2006, 2013). A meta-analysis on the subject by Roscoe and Jenkins (2005) found that of more than 350 tests of a link between campaign contributions and roll-call voting, only around a third identified a statistically meaningful relationship. ¹ It appears that only in rare circumstances can political spending be identified as a key factor in bill introductions or passages.

Still, these mixed results do not mean that economic resources are unimportant. A likely scenario is that the effects of money on policymaking are more nuanced than a simple quid pro quo. Gilens (2014), for example, suggests that while wealthy interests may not win (or even align) on every legislative fight, they still receive much better representation than other socioeconomic groups. One powerful way in which money is thought to influence the political process is by narrowing the scope of policy conflicts. Wealthy interests already occupy advantageous socioeconomic positions and therefore simply protecting the status quo can be a de facto political victory. American political institutions are designed to be slow moving and meticulous, so there is a built-in bias in favor of the status quo. Research by Enns et al. (2014a) suggests that this bias has contributed to economic stratification; the more the inequality, the harder it is for a society to reverse course, as the political remedies needed to correct high levels of inequality often require major policy change.

Protecting the status quo means keeping new approaches to old issues off the political agenda (Hacker and Pierson, 2005), a phenomenon that undermines the government’s ability to solve big problems (Drutman and Teles, 2015). For example, researchers have noted that while public concern over inequality has grown dramatically in recent years, Congress has done very little about the issue. Redistributive issues are clearly an area where wealthy interests prefer the status quo and the general interpretation has been that the government’s failure to address rising inequality is an indicator of the negative-agenda power exercised by wealthy groups (Franko et al., 2013; Kelly and Enns, 2010; Mettler, 2011; Scruggs and Hayes, 2017).

Finally, we know that wealthy interests have different policy preferences than other Americans. The largest divergence is on economic issues. Middle-class and poor Americans are more likely to favor social programs aimed at reducing inequality, and those at the top are much more likely to be concerned about the deficit than average Americans (Page et al., 2013). Furthermore, wealthy Americans are generally more concerned with inflation, while lower income groups worry more about unemployment.

Beyond economic issues, those in the upper class are far from a representative demographic sample of the United States. In 2014, Forbes’ annual list of the 400 richest Americans included only 1 African American and 34 women (Dolan and Kroll, 2014). The demographics of the richest 1% skew heavily toward older, white males. Of course, there is variance in the political ideologies of those at the top. Many of the richest Americans may have policy preferences that are largely congruent with those of middle-class Americans, women, or minority groups. But the uneven concentration of wealth (and political influence) among certain demographics groups still runs counter to pluralistic ideals. As the literature on descriptive representation has noted, ideological sameness is a poor substitute for real diversity (Baker and Cook, 2005; Juenke and Preuhs, 2012; Griffin and Newman, 2008; Whitby and Krause, 2001).

Hypotheses

Schattschneider’s argument (1960) was that political power could be understood as the ability to control the scope of conflict, and that economic elites would attempt to constrain this scope to protect their vested interests. As we cannot directly observe the “scope of conflict,” it is best thought of as a latent variable, which, I argue, manifests itself in the diversity of the congressional agenda. When many unique issues are on the agenda, this implies a far-reaching policymaking enterprise in which a variety of ideas and problem solutions compete for attention. Conversely, a more myopic agenda suggests the opposite: that conflict expansion has not taken place and that policymaking is proceeding along a few narrow channels where legislators perceive a legitimate role for government intervention. All this suggests that when agendas are less diverse governmental capacity for detecting and solving multifaceted problems is diminished.

As inequality grows, elected officials become reliant on a smaller number of ultra-rich donors to fund their campaigns. By contributing disproportionately to candidates who “think small” when it comes to social programming, the wealthy can cultivate a policymaking environment that discourages governmental activism. These efforts may be concentrated in some policy areas (such as those dealing with redistribution or government regulation) more than others, but the overall effect should be to reduce the problem-solving capacity of government. In this way, the wealthy limit the scope of policy conflict to protect a status quo that is advantageous to them. Furthermore, higher levels of inequality reduce the propensity of middle- and lower income groups for participating in politics. These groups might benefit from conflict expansion that disrupts the status quo, so their reduced presence in the political arena only amplifies the voice and influence of the wealthy.

If Hypothesis 1 is correct, then this implies that economic inequality will dynamically affect the diversity of congressional agendas. Rising inequality would lead to more members of Congress who behave myopically, either because legislators who engage with a broader range of social issues are unable to raise the money necessary to compete and are voted out of office or because legislators adapt their behavior to attract wealthy donors. In either case, the cumulative effect would be an overall reduction in agenda diversity. Of course, this dynamic assumes that as inequality rises, members of Congress become more reliant on large donors to fund their campaigns, but this assumption is well-supported by recent scholarship.

Hypothesis 2 suggests the possibility of a feedback loop. If agendas become less diverse as inequality rises, then there is less room on the agenda for policymakers to attend to any given issue, and thus it becomes less likely that economic inequalities will be addressed. This points to inequality as a problem that can be self-perpetuating, making it all the more alarming. Of course, the dynamics I have proposed suggest that higher inequality should decrease the likelihood of any given policy issue receiving attention; economic inequality is just one of many problems that may be neglected. It is also worth noting that these hypotheses are not specific to the US context. They depend on just four factors: (1) that economic elites have unusually strong attachments to the status quo, (2) that these elites have a disproportionately heavy influence on the policy process, (3) that they use this influence to protect the status quo, and (4) that this influence varies depending on the relative economic advantage that elites enjoy. These conditions might well be met in a number of political systems, so the forthcoming results should be understood as an entry point to a research agenda that could be expanded comparatively.

Study 1: Campaign Financing and Legislator Behavior

The first study is designed as a test of Hypothesis 1. Campaign finance data are available from the CRP, which tracks the donation and spending of money in US elections on the website OpenSecrets.org. CRP aggregates the total amount of money congressional candidates raised from individual contributors, PACs, self-financing, and an “other” category. For the 2010 and 2012 cycles, CRP breaks donations from individual contributors down into small and large donor categories. Anyone who contributes $200 or less to a candidate within a cycle is classified as a “small donor” and those who give more are “large donors.”

I use the percentage of a legislator’s total campaign contributions within a cycle that come from small donors as the independent variable in two linear regression models, one that predicts the total number of bills introduced by that legislator and the other the diversity of those bills. Contributions from small donors range from 0% (for legislators who did not raise any money from small donors) to 63%, with a mean of 8% and a median of 5%. The expectation is that legislators who receive more financing from small donors will be more politically active, introducing more bills across a greater range of topics than their counterparts who rely more heavily on other sources to finance their campaigns. (The majority of money not raised from small donors comes from large donors and PACs.)

Of course, many Americans who are not economic elites are capable of donating more than $200 to their favorite congressional candidate, but, as CRP notes, very few people actually do. In the 2014 election cycle, only 0.23% of the total US population gave more than $200; however, donations of more than $200 accounted for 67% of the total sum of individual contributions (Center for Responsive Politics (CRP), 2014). Deep-pocketed donors are an increasingly important component of the campaign finance landscape, serving as political gatekeepers for prospective candidates (Drutman, 2011). Candidates who can buck the trend and raise competitive amounts of money from small donors may feel less pressure to conform to a “do less” agenda.

Data on congressional bills are available online from the Policy Agendas Project. The database codes each bill by subject matter into one of 19 possible “major topic” categories (e.g. agriculture, defense, or social welfare). Using these data, I collapse bills by the legislators who introduced them to produce a total bill count for each member of Congress. I also calculate the congress-to-congress diversity of bill introductions for each legislator using a measure of entropy known as Shannon’s H (Boydstun et al., 2014; Shannon, 1948; Shannon and Weaver, 1971). Entropy measures the spread of objects (or events) across different categories. Higher entropy indicates greater diversity and a noisier signal. Lower entropy indicates less diversity and a clearer signal. If in a given Congress every bill a legislator introduced was on the same topic, healthcare for example, then that legislator would have a very low entropy score for that Congress. If we were to ask, what did this legislator pay attention to, then the data would provide a clear signal: the legislator focused on healthcare. Conversely, if a legislator introduced 10 bills on each of the 19 policy topics, then the entropy score for that legislator would be high, reflecting a diverse spread of attention. Values of Shannon’s H can range from 0, where every object is in only one category, to 1, the most entropic possible system. The measure of legislator bill entropy that I calculate varies between 0 and 0.89.

Bill totals and entropy are the two dependent variables, which I match with the corresponding campaign finance data from OpenSecrets.org. (Finance data from the 2010 cycle are paired with bills’ data for the 112th Congress and the 2012 cycle is matched with the 113th Congress.) Altogether, the models include 1046 observations, which represent member–congress dyads (campaign finance data are unavailable in some instances, so there are fewer than 1070 observations). Controls include indicator variables for the chamber (coded 1 for the Senate), if the member occupies a party leadership position, if the member is a committee chair, and if the member is part of the chamber majority. The model also uses Poole and Rosenthal’s (1985) DW-nominate scores as a control for partisanship and includes random effects for the state legislators’ represent and the congressional session (either the 112th or 113th). ² Table 1 shows the results.

OPEN IN VIEWER

Table 1.

The relationship between contributions from “small” donors and legislator behavior.

Variable	Number of bills	Bill entropy
Proportion small donors	1.76* (0.50)	0.01* (0.00)
Senate	16.91* (1.19)	0.13* (0.01)
Leadership	0.96 (1.59)	−0.00 (0.01)
Committee chair	5.55* (1.78)	0.00 (0.02)
DW-nominate	−10.24* (1.10)	−0.03* (0.01)
Majority party	10.16* (1.19)	0.00 (0.01)
Constant	8.33* (0.95)	0.52 (0.01)
N	1046	1046
R2 within	0.311	0.116
R2 between	0.323	0.219
R2 overall	0.305	0.126

Note: Both models use state- and congress-level random effects. A 1-unit shift in the “proportion of small donors” variable is coded as a 10% change.

*

p ≤ 0.05.

The model predicts that legislators whose campaigns are financed by a higher proportion of small donors are more legislatively active than their counterparts; they introduce more bills and these bills span a greater range of policy topics. Conversely, members who are more heavily dependent on other sources—PAC contributions, large individual donors, or self-financing—do less legislating. The median number of bills introduced by members during the 112th and 113th Congresses is 13 with a range from 0 to 212 bills, so the effect of small donors is substantial. The effect of small donors on bill entropy is also statistically significant and in the expected direction. Legislators who raise more money from small donors engage with a broader range of policy topics, holding other covariates at their means.

Parameters for the other independent variables present few surprises: legislators who are senators, committee chairs, or members of the majority party introduce more bills, while the coefficient for the DW-nominate variable indicates that right-leaning legislators tend to introduce fewer bills. Senators are more entropic in their bill introductions than House members and right-leaning legislators are less entropic, but these are the only control variables from the entropy model with statistically meaningful parameters.

In summary, the models are supportive of Hypothesis 1. There appears to be a relationship between how candidates raise money and how they behave once in office. Previous scholarship has found that as inequality grows, members of Congress become increasingly reliant on large donors and PACs to fund their campaigns. It follows that given the relationship described in Table 1, there should be fewer members pursuing wide-ranging policy goals during periods of acute economic stratification. The cumulative effect would be an overall narrowing of the congressional agenda. Study 2 tests for this possibility.

Study 2: Economic Inequality and the Congressional Agenda

Study 2 uses data on congressional hearings, bills introductions, and public laws from the Policy Agendas Project to measure the entropy of the congressional agenda. These data include every hearing that was held from 1946 to 2010, every bill introduced from 1947 to 2012, and every law passed from 1948 to 2011, amounting to 89,382 total hearings, 430,815 bills, and 16,941 laws. (Note, this is the same bills’ dataset that is used in Study 1, but over a longer period of time and for this study the data are not collapsed by legislator.) Each policy activity is topically coded based on subject matter to one of the 19 possible major topic codes.

Figure 1 shows annual Shannon’s H values for each policy activity. Since the late 1940s, no activity has a value of less than 0.65, indicating that the congressional agenda is always relatively diverse. The entropy of congressional hearings and bill introductions are highly correlated, while the range of topics addressed in public laws appears marginally less diverse in almost every year. All three series trend together, however. Starting in the late 1950s, the diversity of the congressional agenda begins to increase dramatically. This diversification peaks in the 1980s and then begins a gradual, but steady decline.

OPEN IN VIEWER

Figure 1.

Entropy of the congressional agenda.

As each policy activity shows similar trends, I use principal-components factor analysis to compute a diversity index. The first factor accounts for 80% of the variance. Factor loadings for hearings, bills, and laws are 0.924, 0.942, and 0.814, respectively. Figure 1 tracks the diversity index with a solid black line and while the entropy values can range from 0 to 1, the index takes values between −2.8 and 1. During the period of study, the congressional agenda was the most concentrated in 1952 and the most diverse in 1989. ³ Entropy scores for each policy activity and the diversity index are used as dependent variables in subsequent analyses, allowing for the possibility that the effects of inequality may be more acute at some stages of the policy process than others. I expect that inequality will affect all three areas, but that if anything effects should be larger for laws than hearings or bills. Public laws are the final result of the policy process and their passage requires coordination across multiple legislators, making them more vulnerable to negative-agenda control as economic elites need only moderate the behavior of a few legislators to prevent a law from passing.

Figure 2 compares the factored diversity index with the percentage of wealth owned by people in the top 1% of the wealth distribution, a common measure of economic inequality. In 1948, the top 1% owned 30% of the nation’s wealth. This percentage declines throughout the 1960s and 1970s to a low of 23% in 1978. Then, in the early 1980s, wealth inequality begins to increase dramatically, a trend that continues through the end of the time series and is expected to continue for the foreseeable future (Piketty, 2014). For much of the time series, inequality appears inversely related to the diversity index. The congressional agenda undergoes great diversification from the 1950s through the 1970s, when wealth inequality is declining. But in the late 1980s, the diversity of the congressional agenda peaks and begins a steady decline, coinciding with the dramatic rise of inequality. ⁴

OPEN IN VIEWER

Figure 2.

Wealth share of top 1% and the diversity of the congressional agenda.

Figure 1 introduces four measures of agenda diversity (the dependent variables) and Figure 2 tracks wealth inequality (the key independent variable). Having established these operational parameters, I use single-equation ECMs to estimate the short- and long-term effects of wealth inequality on agenda diversity. ⁵ This model takes the following form

Δ Y t = β 0 + β 1 Y t − 1 + β 2 Δ X t + β 3 X t − 1 + ε t

The β₁ coefficient is the error-correction parameter, which is the rate at which a change in Y will decay over time. Each independent variable is associated with two parameters (β₂ and β₃). The coefficient for the differenced term, β₂, estimates the short-term effect (also called the Granger effect) of X on Y. This effect is realized immediately and then decays in subsequent time periods at the rate indicated by β₁. The second parameter, β₃, represents the long-run effect between X and Y. It estimates how much a change in X will upset the equilibrium level of Y. The long-run effect describes the size of the effect in only one time period, which is important but does not convey the full extent of the relationship between X and Y. The full effect can be calculated by estimating the long-run multiplier, which is simply the ratio between β₁ and β₃ describing the total effect of X on Y distributed over multiple time periods. ⁶ In summary, the single-equation ECM allows analysts to estimate two types of effects, those that occur immediately and those that manifest themselves more gradually (Ura and Ellis, 2012).

The model used here includes five independent variables measuring polarization in the House of Representatives, the average annual DW-nominate score for members of Congress, unified government, election years, and public policy mood. These are in addition to the variable for wealth inequality, as shown in Figure 2. Before proceeding to the results, I briefly review the measurement of each variable and the reasons for its inclusion. ⁷

Since the late 1970s, polarization in Congress has been increasing, reaching levels in the 112th Congress (2011–2013) not seen since the early 1900s. Heightened partisan conflict may affect the diversity of the congressional agenda in various ways. On one hand, polarization can be a major roadblock to successfully passing bills into law. Indeed, the historic unproductiveness of the 112th Congress is widely attributed to the inability of the two parties to find common ground on almost any issue. But partisan conflict can also bring multiple perspectives to the table and force policymakers to consider (if not act on) a wider range of policy options, as legislators compete to give voice to very different ideological prerogatives. A polarized Congress may therefore have a more diverse agenda, while also being less likely to actually pass legislation. Each model includes a variable that measures House polarization using Poole and Rosenthal’s DW-nominate scores. Specifically, the measure is the difference between the mean DW-nominate score for the Democratic and Republican coalitions. From 1948 through 2010, these scores vary between 0.41 and 1.03, with lower values indicating less polarization. ⁸

Another variable measures the liberalism of Congress over time by taking the annual mean DW-nominate score of legislators from both chambers. This variable ranges from −0.12 to 0.14, with higher values indicating a more conservative Congress. A central tenant of conservative ideology is limited government. Reagan’s presidency began as backlash to the expansionary period of the post-war era, and the rise of neoliberalism in the 1990s showed that calls for retrenchment can sometimes have bipartisan appeal. It makes sense then that the diversity of the congressional agenda may decline when legislators are, on average, more conservative.

The model also includes a variable for unified government, which is coded 1 if the same party controls the White House and Congress. Unified government affords the party in power an unusual opportunity to shape the government’s agenda and this might lead to greater productivity in terms of laws being passed, as one party can act unilaterally. On the other hand, periods of unified government may reduce the diversity of the agenda, as the party in power may be better situated to exclude rival viewpoints from the policymaking process. So the inclusion of a control for unified government follows the same logic that justifies a control for House polarization: Inter-party conflict may promote agenda diversity, even if it makes the passage of legislation less likely.

There is also a dichotomous control for election years. Policymaking follows a well-known pattern where most bills are introduced in the first congressional session and most laws are passed in the second. The diversity of the congressional agenda may likewise follow the electoral calendar, with a greater range of policy options on the table in the first session and a more focused agenda in the second.

Finally, the model includes a measure of Stimson’s (1991) public policy mood. This measure is created by aggregating across national survey results to isolate the public’s latent attitudes about the role of government. Stimson argues that these attitudes can be reduced to a single dimension—more or less government—with higher values indicating a public that is in the mood for more government interventions and lower values a public that favors a diminished role for government. I control for mood in the model as it seems plausible that declines in agenda diversity could result from legislators adjusting their behavior in response to a public that wants less from government. Table 2 presents the results of models for hearings, bills, laws, and the diversity index, with coefficients and standard errors in parenthesis. ⁹

OPEN IN VIEWER

Table 2.

Predicting the diversity of the congressional agenda.

	Hearings	Bills	Laws	Index
Long-run multipliers
Top 1% wealth sharet−1	−0.13* (0.01)	−0.18* (0.03)	−0.14* (0.02)	−2.80* (0.21)
House polarizationt−1	0.04* (0.00)	0.05* (0.00)	0.03* (0.00)	0.74* (0.03)
Unified governmentt−1	−0.01* (0.00)	−0.05* (0.01)	0.01 (0.00)	−0.23* (0.09)
Election yeart−1	–	–	–	–
Mean DW-nominatet−1	−0.56* (0.05)	−0.48* (0.14)	−0.48* (0.09)	−9.04* (0.89)
Public policy moodt−1	−0.00 (0.00)	0.00 (0.00)	0.00* (0.00)	0.05* (0.02)
Long-run effects
Top 1% wealth sharet−1	−0.04* (0.01)	−0.17* (0.03)	−0.11* (0.02)	−1.88* (0.37)
House polarizationt−1	0.01* (0.00)	0.04* (0.00)	0.02* (0.00)	0.50* (0.08)
Unified governmentt−1	−0.00 (0.00)	−0.04* (0.01)	0.00 (0.00)	−0.15 (0.09)
Election yeart−1	–	–	–	–
Mean DW-nominatet−1	−0.18* (0.07)	−0.44* (0.16)	−0.36* (0.11)	−6.06* (1.37)
Public policy moodt−1	−0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.03 (0.02)
Short-run effects
ΔTop 1% wealth sharet	−0.01 (0.03)	−0.18 (0.06)	0.04 (0.05)	−1.06 (0.54)
ΔHouse polarizationt	0.01 (0.01)	0.05 (0.03)	−0.04 (0.02)	−0.01 (0.28)
ΔUnified governmentt	−0.01* (0.00)	−0.01 (0.01)	−0.01 (0.01)	−0.27* (0.10)
ΔElection yeart	−0.00 (0.00)	0.00 (0.00)	−0.00 (0.00)	−0.06 (0.05)
ΔMean DW-nominatet	−0.29* (0.11)	−0.17 (0.22)	−0.08 (0.18)	−5.36* (1.93)
ΔPublic policy moodt	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.03 (0.03)
Error correction and constant
Error correction	−0.32* (0.10)	−0.92* (0.15)	−0.75* (0.12)	−0.67* (0.11)
Constant	0.34* (0.11)	0.92* (0.24)	0.58* (0.15)	0.42 (1.02)
Observations	63	65	63	62
Adjusted R2	0.169	0.405	0.401	0.385
Dickey–Fuller test (β)	−10.14*	−6.10*	−9.08*	−9.61*

Note: The Dickey–Fuller test assesses the null hypothesis of a stochastic trend in the model’s residuals. A significant parameter allows the rejection of the null hypothesis, indicating that there is a linear combination of the variables that forms a long-run equilibrium relationship (i.e. the series cointegrate). The models all show statistically significant parameters for the Breusch–Pagan test, indicating that there is some possibility for the standard errors to be biased by heteroskedasticity. Each model therefore uses Prais–Winston (1954) estimation, which is more robust to issues of residual autocorrelation and heteroskedasticity. A 1-unit shift in the wealth of the top 1% is coded as a 10% change.

*

p-Value ≤ 0.05.

The results across all four models are similar. In each case, the parameters associated with short-run effects are small and not statistically meaningful, with the notable exception of the coefficients for mean DW-nominate and unified government, which are statistically significant and negative for the hearings and index models. On the other hand, the parameters for the long-run multipliers are almost all statistically significant, so the government agenda appears to move only slowly in response to shifting sociopolitical factors. This makes sense; government agendas are large, complex, and not easily changed. Rather than abruptly shifting directions, agendas slowly adapt to new environmental circumstances. Crucially, in each model, the long-run multiplier associated with wealth is negative and statistically significant. This indicates that wealth inequality affects the equilibrium of agenda diversity such that a 10% increase in the wealth owned by the top 1% corresponds to a decrease in the entropy of legislative hearings of 0.13, 0.18 for bills, 0.14 for public laws, and 2.80 for the diversity index. Entropy for the three policy series varies between 0.65 and 1, so the effects described by the model are substantial. (The factored index takes values between −2.5 and 1, explaining the larger coefficient estimates for this model.) Based on the estimated parameter, the increase in inequality observed from the 1980s to present of about 15% would be associated with a dramatic decrease in the diversity of issues on the congressional agenda, controlling for other relevant factors.

These results point toward a dynamic relationship between economic inequality and the diversity of congressional agendas. As inequality rises, diversity is diminished. It makes sense that this effect should manifest itself over the long run. Wealth inequality can be understood to slowly affect the overall policymaking environment, rather than providing a stimulus to which policymakers might immediately respond (if they are even immediately aware of shifts of inequality). One conjecture that the model does not support is that the effects of inequality should be larger at later stages of the legislative process. Instead, it appears that inequality affects agenda diversity at every stage: hearings, bills, and laws. In fact, the coefficient for public laws is smaller than for bills, suggesting that wealthy donors can successfully moderate the behavior of legislators even relatively early in the policy process.

Wealth inequality is not the only factor that sets a targeting rate for agenda diversity. ¹⁰ As expected, increases in House polarization increase the long-run diversity of the agenda and this effect is apparent in all four models. Conversely, unified government is associated with decreasing agenda diversity (where the parameter is statistically significant). Together, these results suggest that partisan conflict can be a powerful force for diversification. When one party controls both Congress and the White House, it has greater procedural authority to suppress rival viewpoints. When polarization is low, such suppression is less necessary. But if control of government is divided or polarization is high, inter-party conflict forces policymakers to consider a wider range of policy options. The effects of congressional liberalism are also consistent with expectations. When average DW-nominate is higher, indicating a more conservative Congress, the agenda is less diverse at every stage of the policy process. This is a particularly important factor to control for, as it is possible that a rightward shift in ideology could be a major cause of the decline in agenda diversity observed in Figure 1. The model shows that this is indeed an important part of the story, but that economic inequality has its own, independent effect. Finally, the parameters associated with public policy mood are only statistically significant in the laws and index models. However, as expected, the effects are positive, indicating that agendas become more diverse when the public is in the mood for more government.

It is important, of course, to be aware of the limitations of the model. A great many factors contribute to setting the congressional agenda, so isolating the effect of inequality against the backdrop of rising polarization, shifting ideologies, and emerging institutional norms is empirically challenging. The results suggest a robust statistical association between inequality and agenda diversity, one that is not easily explained away by factors such as partisanship or polarization, but further testing is necessary to identify the exact nature of the relationship.

Attention to Social Welfare

The government’s agenda becomes less diverse as wealth inequality rises, suggesting a diminished capacity for seeking out and responding to social problems. One area where we might expect policymakers to spend less time during periods of high inequality is on redistributive issues. This is an area where wealthy interests are much more likely to favor the status quo than other Americans. Thus, it makes sense that if the wealthy use political spending to narrow the scope of policymaking, then these effects should be especially pronounced for issues dealing with the redistribution of wealth.

One of the 19 major topic codes used by the Policy Agendas Project is for social welfare. Any policy activity that focuses on food stamps, assistance for low-income families, social security, or other social assistance programs receives the social welfare topic code. Welfare is fundamentally redistributive, so this is an area of policymaking that should be particularly sensitive to rising inequality. Figure 3 shows the annual percentage of total hearings, bills, and laws that address social welfare. All three series are highly correlated. Attention to this issue grows during the 1960s and peaks in the late 1970s, when about 4% of all public laws addressed welfare. This attention dissipates rapidly, however. By the 1990s, in most years less than 1% of total hearings, bills, or laws focus on welfare.

OPEN IN VIEWER

Figure 3.

Congressional attention to social welfare.

Recall from Figure 2 that inequality is relatively low during the 1960s and 1970s, but begins to grow rapidly in the 1980s. Welfare is a potential solution to economic stratification, so it is surprising that Congress pays less attention to social welfare as economic inequality increases. However, this type of dynamic is exactly what we would expect if wealthy interests are successful in undermining the government’s ability to address social problems. The relative decline in attention to social welfare helps explain the general decline in agenda diversity predicted by the model from Table 2.

This interpretation gains support from a series of ECMs. Because all three series are highly correlated and related to the same underlying phenomenon, I use principal-components factor analysis to compute a common index of welfare attention. ¹¹ This index serves as the dependent variable, along with the percentage of total hearings, bills, and laws on social welfare, in subsequent regressions. Independent variables are the same as in the previous model with the addition of a variable measuring the annual percentage of people below the poverty line. One might expect that attention to social welfare will grow in tandem with poverty, as welfare becomes increasingly important (and costly) when many people are below the poverty line. Also, the model substitutes the generic public policy mood measure for one that focuses specifically on welfare policies. ¹² Table 3 shows the results of all three models.

OPEN IN VIEWER

Table 3.

Predicting congressional attention to social welfare.

	Hearings	Bills	Laws	Factor
Long-run multipliers
Top 1% wealth sharet−1	−3.38 (2.00)	−1.79* (0.67)	−1.39 (1.17)	−1.70* (0.58)
Poverty ratet−1	−0.52 (0.26)	−0.05 (0.08)	−0.01 (0.14)	−0.06 (0.06)
House polarizationt−1	−0.19 (0.95)	0.68* (0.32)	0.70 (0.58)	0.38 (0.28)
Unified governmentt−1	−1.02 (0.92)	−1.04* (0.28)	−1.02* (0.48)	−0.43 (0.25)
Election yeart−1	–	–	–	–
Mean DW-nominatet−1	1.15 (17.64)	−19.27* (6.07)	−9.37 (9.55)	−5.10 (4.87)
Welfare moodt−1	−28.43* (9.67)	−9.63* (3.16)	−1.25 (5.16)	−5.64* (2.48)
Long-run effects
Top 1% wealth sharet−1	−2.59 (2.15)	−0.92 (0.79)	−1.11 (1.25)	−1.38 (0.73)
Poverty ratet−1	−0.40 (0.26)	−0.02 (0.08)	−0.00 (0.14)	−0.05 (0.07)
House polarizationt−1	−0.14 (0.94)	0.34 (0.34)	0.56 (0.59)	0.31 (0.29)
Unified governmentt−1	−0.78 (0.93)	−0.53 (0.29)	−0.82 (0.47)	−0.35 (0.24)
Election yeart−1	–	–	–	–
Mean DW-nominatet−1	0.88 (17.63)	−9.86 (6.30)	−7.52 (9.59)	−4.14 (4.94)
Welfare moodt−1	−21.83 (10.83)	−4.93 (3.58)	−1.00 (5.19)	−4.58 (2.75)
Short-run effects
ΔTop 1% wealth sharet	−1.25 (5.18)	−2.81 (1.88)	−7.94 (4.22)	−4.03 (2.15)
ΔPoverty ratet	0.50 (0.52)	−0.03 (0.17)	0.30 (0.28)	0.10 (0.13)
ΔHouse polarizationt	−4.05 (3.19)	−1.40 (1.20)	−7.62* (2.26)	−2.98* (1.08)
ΔUnified governmentt	−1.40 (0.95)	−0.16 (0.37)	−1.19 (0.68)	−0.56 (0.31)
ΔElection yeart	−1.44* (0.59)	−0.61* (0.28)	−1.37* (0.45)	−0.60* (0.21)
ΔMean DW-nominatet	−4.11 (23.82)	1.41 (7.69)	29.32* (12.64)	3.88 (6.83)
ΔWelfare moodt	−29.95 (20.59)	0.94 (7.96)	3.16 (13.59)	−1.29 (6.47)
Error correction and constant
Error correction	−0.76* (0.16)	−0.51* (0.15)	−0.80* (0.16)	−0.81* (0.19)
Constant	30.34* (9.80)	5.69 (3.34)	2.64 (4.16)	5.91* (2.67)
Observations	46	48	41	40
Adjusted R2	0.414	0.161	0.483	0.435
Dickey–Fuller test (β)	−8.68*	−11.85*	−9.56*	−9.72*

Note: Standard errors are shown in parentheses. Each model uses Prais–Winston estimation. A 1-unit shift in the wealth of the top 1% is coded as a 10% change.

*

p ≤ 0.05.

Expectations are that as the top 1% accumulates a larger share of the nation’s wealth, there should be a migration of attention away from social welfare. Looking at the regression for congressional hearings, the long-run multiplier for the top 1% variable is negative as expected, but not statistically significant. The same is true for the laws model. For bills and the index measure, however, wealth shows a statistically significant effect. For every 10% increase in wealth owned by the top 1%, the percentage of total bills on the topic of social welfare decreases by about 1.80%; the index measure decreases by 1.70. Figure 3 shows that at most between 3% and 4% of total bills are on social welfare, so the effects described by the model are sizeable. (Once again, there is no evidence of a short-term relationship between economic inequality and government agendas; the effects are only manifest over the long term.) It is also noteworthy that the number of people in poverty appears to have no effect on attention to social welfare. Furthermore, legislators do not appear responsive to public mood when it comes to welfare; as the public favors greater interventions on this issue, legislators hold dramatically fewer hearings and introduce fewer bills on the subject. ¹³ In all, Figure 3 and the ECMs suggest that as wealth inequality grows and the congressional agenda constricts, policymakers become less inclined to discuss social welfare. ¹⁴ Notably, the notion that lawmakers are more likely to focus on welfare as economic stratification becomes more pronounced does not receive any support.

Conclusion

Previous studies that investigated the effect of economic stratification on public policy have focused predominantly on the positions taken by members of Congress. This article shifts the focus from positions to the policymaking environment. Specifically, the variable of interest is attention and how it is distributed across issues. While this approach is novel to recent studies of inequality, it derives from familiar sources. In particular, Schattschneider’s admonition in The Semisovereign People that political power can be understood as the ability to control the policymaking space is what inspires the current measurement strategy. In effect, the diversity of the congressional agenda can be understood as approximating the policymaking space; when the agenda is narrow and myopic, then fewer issues can be considered. If wealthy interests seek to constrain this space, and if their ability to do so varies depending on degrees of economic stratification, then this implies a dynamic relationship between inequality and agenda diversity. This article presents preliminary evidence of just such a relationship. Members of Congress whose campaigns are financed by large donors and PACs introduce fewer bills on a narrower range of topics (Study 1), and the cumulative effect is to reduce the overall diversity of the congressional agenda (Study 2).

Given that economic inequality is likely to grow, an increased focus on this issue in the social sciences is appropriate. Further studies investigating the negative-agenda power of wealthy interests would be especially welcome, as this is a major but understudied way that economic elites influence politics. This article has presented evidence suggestive of a robust statistical association between economic inequality and congressional agendas. To build on this analysis, scholars might look for discontinuities where campaign contributions increased dramatically from one electoral cycle to the next. Changes brought forth by revisions to campaign finance laws, by either Congress or the Supreme Court, may present opportunities in this regard. If legislative agendas shift abruptly in the immediate aftermath of a sudden influx of political spending, then that would be evidence in favor of the dynamics that I have postulated. Another route for further study would be to code bills based on their redistributive or regulatory effects. The expectation would be that as inequality rises, fewer bills seeking downward redistributions of wealth would make it through the policy process.