Billionaires and Stealth Politics
Page 18
Finally, we expect that wealthy Republicans are more likely to engage
in boundary control than wealthy Democrats. The boundary control strategy most naturally fits with support for Republican candidates, whose conservative political ideologies align well with billionaires’ self interested pursuit of small government on the national level.63 It is probably no coincidence that most of the industries with regulatory and market structures in which boundary control is attractive— including high polluting industries like energy and chemicals, and labor intensive and union averse manufacturing firms— have tended to be aligned with Republicans.64 Since
it would be awkward at best for wealthy Democrats to engage in political
strategies that require extensive public and private support for conservative Republicans, billionaires’ partisanship should matter.
What the Evidence Says
In order to test our expectations about boundary control, we expanded
our data set on the one hundred wealthiest US billionaires to include measures of the regulatory structure and market structure faced by the main industry in which each billionaire was currently active. We coded each
billionaire’s principal industry on four point scales indicating the intensity with which it was regulated (from light to very heavy), the sites of its regulation (from almost entirely national to almost entirely subnational), the extent of barriers to entry (from low to very high), and the net costs it imposed on the public in the areas in which it was located.65 Since the boundary control theory predicts that the strategy is adopted only when
regulation is both heavy and subnational, we combined the intensity scale and the site of regulation scale to form an overall regulation index. We
also summed the scores on each variable for each industry, in order to
develop a total industry score for each industry.66
114
chapter five
Sixteen of our one hundred billionaires were taking no active role in
business during the period of study. Because the point of the boundary
control strategy is to obtain business specific financial benefits that are irrelevant to those not active in any business, the inactives were removed from this analysis. Of the remaining eighty four billionaires, forty three clearly were active only in a single industry, so we used the industry wide scores for that industry to code each of them. For the forty one billionaires who were active in two or more industries, we selected the industry that had the highest total score and coded each individual based on that
industry.67 To ascertain billionaires’ party attachments, we used federal campaign contribution data to develop a measure of the party with which
each billionaire was most aligned.68
We measured billionaires’ use or nonuse of boundary control— our chief
dependent variable— by examining national and subnational level contribution patterns. At the national level, boundary controlling contributors are (by definition) more likely to donate to outside groups— especially
small government groups— or general party funds than to individual candidates. On the subnational level, they are more likely to make donations directly to candidates who are members of dominant state level parties
than to outside groups or minority party candidates. At the state or local level, direct contributions to candidates or to individual candidates’ PACs are necessary as a means to “buy” their support for state level policy favors.
We therefore coded billionaires as engaging in boundary control strategies if they met three criteria. First, they had to be active in a state with a dominant state level party, which we defined as any state that had had unified government (led by the same party) for six or more of the twelve
years from 2000 to 2012.69 Second, at least two thirds of a billionaire’s national level contributions (in terms of total dollars) had to go to outside groups associated with small government causes or to Republican Party
general funds. Finally, at least two thirds of their contributions within a state (again in terms of total dollars) had to go directly to candidates themselves (or to the PACs of those candidates) who are members of the
dominant state level party.70
One billionaire had to be dropped from the analysis because he was
active in a state in which there was no online campaign finance database
that allowed for searches by contributor, leaving eighty three billionaires to be analyzed. Of those eighty three billionaires, eight (nearly 10 percent of those studied, or 8 percent of our entire set of one hundred billionaires) met the criteria for engaging in boundary control strategies.71
reshaping state and local politics
115
table 5.2 Firth Logistic Regression Model Predicting Use of Boundary Control 95% CI
B (SE)
p Value
Lower
Upper
(Intercept)
– 17.105 (5.876)
<0.001
– 33.210
– 7.439
Regulation Index
2.632 (1.645)**
0.038
0.117
7.593
Barriers to Entry
– 1.681 (2.018)
0.369
– 7.969
2.829
Concentrated Costs
1.017 (1.677)
0.494
– 1.622
10.266
Party
2.627 (1.632)**
0.037
0.128
7.587
** p < .05.
Note: Likelihood ratio test = 26.025 on 4 DF; p = .00003; n = 83. Wald test = 11.622 on 4 DF; p = .0204.
All boundary controllers came from Republican dominated states.72
Although both groups were highly active politically, boundary controllers contributed much more money to politics than non– boundary controllers— three times as much (an average annual sum of just over $1.5 million, compared with about $415,000). They also contributed more money to
state politics than non– boundary controllers, particularly if one excludes ballot initiatives (to which a few individuals with intense preferences on social issues like same sex marriage have donated vast sums). Excluding ballot initiative contributions, boundary controllers on average made state level contributions of $109,425, nearly double the average annual
amount contributed by others ($68,768). At the state level, in terms of
total dollars, a much greater proportion of boundary controllers’ contributions (85 percent of them) than non– boundary controllers’ contributions (57 percent) went directly to candidates.73 On the national level, a much greater proportion of boundary controllers’ dollars (82 percent)
than non– boundary controllers’ money (54 percent) went to outside
groups or parties. The contrast was even greater (82 percent vs. 37 percent) in terms of the proportion of national level contributions that went to conservative groups or the Republican Party.
Because of the limited number of billionaires studied, the rather few
cases of boundary control, and the fact that the independent variables
together caused quasi complete separation in a standard logistic regression model, we estimated the relationship between the variables in the data set using a Firth penalized likelihood logistic regression.74 The Firth approach helps with the problems of separation and rare outcomes.
The main results of our analysis are presented in table 5.2. The model includes four independent variables: Regulation, Barriers to Entry, Net Public
116
chapter five
table 5.3 Estimated Effects of Changes on Boundary- Control Independent Variables Average Treatment
95% BCa CI
Effect
Standard Error
Lower
Upper
Regulation Index
0.637
0.195
0.125
0.851
Party
0.123
0.044
0.054
0.233
Note: “Average treatment effect” is the predicted change in probability of the outcome variable (i.e., boundary control) when moving from the minimum observed value on each independent variable to the maximum value.
Standard errors were calculated using bootstrapping. Bias corrected and accelerated (BCa) confidence intervals (Efron 1987) are used to account for potential skewness in the bootstrap distribution.
Costs, and Party. The Regulation and Party coefficients, which are very similar in terms of magnitude, are both statistically significant by the usual criteria. But the coefficients for Barriers to Entry and Net Public Costs are not statistically distinguishable from zero. So our analysis con firmed two of our expectations (hypotheses 1 and 4 in table 5.1) but did not provide support for two others (hypotheses 2 and 3). We earlier explained some possible reasons for those null findings.
These results show a definite relationship between both regulatory
structure and partisanship and the use of boundary control. These results are theoretically intelligible and statistically clear. Thus— while they do not prove a causal relationship— the quantitative findings suggest that boundary control is a distinctive strategy of Republican billionaires who do business in highly state regulated industries.
In order to assess how big an impact regulatory structure and billionaires’ partisanship have on the likelihood of billionaires pursuing a boundary control strategy, we estimated average independent treatment
effects for each: the predicted change in the probability of using boundary control when moving from the minimum to the maximum observed values on the regulation index variable75 or from Democrat to Republican
in party affiliation. The results are presented in table 5.3.
The estimated effect of an increase in state level regulation is a very substantial 0.637. That is, moving from the (observed) lower end of the regulatory structure variable to the higher end is estimated to increase the likelihood of the use of boundary control by fully 64 percent. The estimated effect of party (that is, of a billionaire shifting from Democrat to Republican) is a smaller but still meaningful 0.123, signifying a 12 percent increase in the likelihood of using boundary control. Thus a billionaire’s Republican Party affiliation is estimated to have a small but significant independent impact on the probability that the billionaire will pursue a boundary control strategy.
reshaping state and local politics
117
But our estimates indicate that the regulatory structure of the billionaire’s main business (a large amount of state level regulation) very strongly tends to lead to efforts at boundary control.
Case Studies to Improve Causal Inference
As almost everyone knows, however, correlation is not causation. In order to make more confident causal inferences about which factors lead to the use of boundary control strategies, it is important to grapple with certain problems that quantitative analyses of observational data cannot
by themselves adequately resolve. These problems include the possibility
of omitted confounding variables (factors that, when left out, can distort or “confound” findings) and the possibility of measurement errors in independent variables of interest (which can lead to underestimates of the effects of those variables).
Carefully designed case studies can enrich and deepen quantitative
analyses by dealing with these problems, which are inherent in regression
type approaches to observational data.76 Case studies can also illuminate causal pathways, suggest temporal sequences, and contribute in other ways to the analysis of social and political processes.
To advance our analysis of boundary control strategies, we selected for
a closer look cases that were extreme (either extremely high or extremely low) on key independent variables. Such cases are ideally suited to identify omitted confounding variables, as well as to uncover any measurement error that may exist in independent variables of interest.77 It also turns out— not by accident— that such case selection often produces case
studies that are interesting in themselves, illustrate the workings of the theory, and generate new ideas about causal processes.
Measurement error in this context could consist of simple coding errors. Or it could reflect more profound problems in the way we measured variables of interest. (If, for example, some billionaires were deemed to be engaging in boundary control based on the way we operationalized
that concept, but were found through closer examination to take actions
that in some way indicated otherwise.) Measurement errors might also be
revealed by the discovery of campaign contributions that escaped detection during the creation of the data set and alter our understanding of a billionaire’s political strategies.
We identified extreme high and extreme low cases by summing individual billionaires’ scores on all the independent variables in the model (that is, all the factors we expected to lead to efforts at boundary control)
118
chapter five
and locating who had the maximum and minimum total values in our data
set. Harold Simmons was unique as the extreme high case, most strongly predicted to use boundary control. From among billionaires who tied as
the extreme low case (least likely to use boundary control) we randomly
selected George Soros.78
We also selected a billionaire— Robert Rowling— who was typical (in terms of independent variables) of the eight who used boundary control.79
Study of a typical case is useful for at least two reasons. First, it helps make an otherwise abstract theory more concrete and real. That is, it illustrates how a theory works. Second, and perhaps more importantly for us, studying one of our typical cases can help identify any politically important effects of using a boundary control strategy. Do boundary controllers actually receive direct benefits on the subnational level as a result of their contributions? Our quantitative model helps predict who does or does not try to use boundary control. But we also want to know whether that strategy actually pays off. Robert Rowling was randomly selected as a typical case from among the eight billionaires scored as utilizing boundary control.
In each of the three case studies below, we first identify the billionaire’s financial interests, discuss them in terms of our independent variables
(e.g., the regulatory and market structure of his businesses, and his party affiliation) and then detail his national and subnational level political activities, especially campaign contributions. We then assess whether or
not our qualitative assessment aligns with the findings of the quantitative analysis, by searching for each of the potential sources of measurement
error discussed above. Finally, for cases in which boundary control was
used, we search for evidence on whether or not that strategy was successful in terms of accruing direct, subnational level financial benefits for the would be boundary controllers.
harold simmons. The late Harold Simmons was selected as the extreme high case on the independent variables. (He scored higher on the sum of the four independent variables than any other billionaire studied.) So we strongly expected that he would engage in efforts at boundary control. And we coded him as having indeed done so.
Simmons— born in Golden, Texas, in 1931, to two school teacher parents— spent his early childhood without indoor plumbing or electricity. He attended the University of Texas and was employed as a bank examiner
along the way to becoming a wealthy industrialist active in metals, chemicals, waste management, and industrial manufacturing. Mainly through his
reshaping state and local politics
119
privately held company, Cont
ran, Simmons had by 2013 accumulated a net
worth of $10 billion.80 Simmons’s highest scoring company on the independent variables was his waste management firm, Waste Control Specialists.
The waste management industry, which includes the disposal of highly toxic commercial waste, is very heavily regulated, mostly on the state and local levels. The waste management industry also has high barriers to entry and involves high localized public costs (pollution, unattractive sights and smells, and the risk of environmental and public health catastrophe if toxic waste is mishandled) with few compensating local public benefits.81
Simmons was a major campaign contributor. His involvement in national politics— he once darkly proclaimed that Barack Obama was a
“socialist”— has received a fair amount of attention from both academia
and the mass media.82 Simmons’s political involvement on the state level, however, has been less noticed.
On the national level, Simmons advocated for a small federal government and noninterference with the states. He supported conservative Republicans and attacked Democrats. Most notably, in 2004 Simmons contributed a substantial amount to the infamous “Swift Boat Veterans for Truth,”
a group that then proceeded to run vicious attack ads about the military
career of Democratic presidential candidate John Kerry, who had served
honorably in Vietnam and received a purple heart. Simmons was also responsible for the Bill Ayers– Barack Obama attack ads of 2008, which attempted to link Obama to domestic terrorism.83 (Ayers, a rebel in his youth, was by 2008 a peaceable and mild mannered Chicago area scholar. Ayers
and his acquaintances may have been puzzled by the idea of demonizing
Obama for association with Ayers, but the ads made big waves nationally.) In the 2012 elections Simmons gave just shy of $27 million to conservative super PACs, most of which focused broadly on supporting conservatives in tight races rather than on aligning themselves with individual candidates.84 Simmons also contributed directly to some candidates or
their individual PACs, but these donations paled in comparison to the
donations he made to groups with a broader focus. In fact, he did not give more than $10,000 to any single candidate or any single candidate’s PAC.85