Slate has published a piece by Zachary Karabell entitled “Stop Obsessing Over exorbitant CEO Pay” in which the author appears unaware that he has reported, and then ignored, evidence that scholars he cites favorably are “resoundingly convinced” proves the opposite. Karabell’s author’s page shows that he has recently joined Slate and promotes theoclassical dogmas about economics. He is a Wall Streeter of the kind that thinks it is an honor to be a “regular” on CNBC. He also touts being a favorite of the Davos plutocrats. In roughly a month with Slate he has managed to be an apologist for high unemployment, inequality, and high frequency trading (HFT) scams.
Karabell begins by acknowledging that we need, urgently, to correct the perverse incentives of modern executive compensation.
“Business school professors who study the effect of excessive executive compensation are resoundingly convinced that too much comp hurts the overall performance of companies. Fifty years ago the ratio of average CEO comp to average salaries was 24-to-1; now it is 204-to-1. Many business scholars believe tying so much of CEO comp to stock and the performance of a company’s shares incentivizes CEOs to make quarterly earnings look good whether or not it benefits the company’s long-term health.”
Karabell concedes that even though the business school scholars are “resounding convinced” on the basis of their econometric studies that executive compensation harms corporate performance CEOs have responded to the research by exacerbating the perverse incentives and personal payoffs. At this juncture the obvious question, which Karabell never asks, is that since he concedes that executive compensation increases inequality and concedes that it harms business productivity it follows logically that inequality driven by CEO compensation harms other workers. Business productivity is a necessary condition to workers increasing their incomes, so Karabell has established at least one reason why increased inequality is harming the middle and working classes. Karabell is oblivious to the logical implications of the business school research.
If Karabell were familiar with the broader literature on executive and professional compensation he would have made a far stronger case on why it is critical to remove the perverse incentives of both forms of compensation. The accounting and finance literature confirm the criminology literature – executive and professional compensation encourage and aid control fraud. I have explained these points in scores of articles, so I will only summarize the key conclusions here.
- Executive compensation is the principal means by which sophisticated “looting” (Akerlof & Romer 1993) occurs.
- Looting through seemingly normal executive compensation makes prosecution far more difficult.
- When executive compensation is much larger the incentive to engage in control fraud increases dramatically. Accounting control fraud provides a “sure thing” of exceptional wealth and prestige for the CEO.
- Executive compensation below the CEO level is a superb device for enlisting officers to aid the control fraud while simultaneously providing deniability for the CEO.
- The CEO abuses modern professional compensation to suborn the “controls” who are supposed to prevent insider frauds but instead aid and abet the CEO’s fraud.
Control frauds cause greater financial losses than all other forms of property crime – combined. This means that modern executive and professional compensation are, in conjunction with the three “de’s” (deregulation, desupervision, and de facto decriminalization) the primary factors that produce the criminogenic environments that drive our recurrent, intensifying financial crises as well as many control frauds that maim and kill thousands of people. Each of these factors compounds the (much smaller) perverse incentives of executive compensation that Karabell concedes have been demonstrated to the satisfaction of business school scholars using econometric testing. The non-econometric forms of research used by criminologists, regulators, law, and some accounting scholars also offer the opportunity to determine causality directly.
The rest of Karabell’s article consists of another logical error. His logic chain is: some scholars believe inequality is very harmful and note that it is correlated with a range of undesirable variable but do not believe that econometric tests can prove causality. That is, of course, a correct statement on their part because, by definition, econometric tests cannot demonstrate causality. We can disprove a hypothesis, but we cannot prove a hypothesis through such econometric tests. Notice that Karabell treated the business scholar’s econometric tests as gospel (“resoundingly convinced”) even though such tests cannot prove that changing the independent variable caused the change in the dependent variable. “Resoundingly convinced” is an accurate description of the state of business scholar’s confidence on this point, but it not a scientific statement.
The inherent limitations of making statistical inferences about potential causality on the basis of econometric hypothesis testing illustrate the advantages of using multiple research methodologies, particularly ones that can examine causality. Our “autopsies” of failed S&Ls and our civil, enforcement, and criminal cases in which we had to demonstrate the existence of the fraud and the fraud mechanisms required us to prove causality against skilled legal opponents with enormous resources and insider expertise that allowed them to propose alternative forms of causation. Econometric studies are particularly likely to fail horrifically when they rely on accounting data that the CEO is systematically inflating through accounting fraud.
If we look at the entire body of research on the perverse incentives of executive and professional compensation we have a compelling case for concluding that that it is vital and urgent to change modern executive and professional compensation. We cannot use “autopsies” or court cases to establish causality about issues like whether inequality reduces the compensation of middle class and working class workers. We must rely on logic and the limits of statistical inference. I explained the logic example above – if Karabell concedes that CEO pay reduces business productivity then it must harm working and middle class compensation.
But Karabell has done something far worse than failing to understand the inherent limits of econometric inferences. He concedes that inequality at levels we now witness in the U.S. is correlated with a range of adverse conditions, including wage stagnation. He then takes a totally banal statement – correlation does not prove causality – and fails to understand even the most basic aspects of statistical hypothesis testing. Here is an example of Karabell’s failure to understand even the concepts he purports to be discussing.
“Progressive economist Jared Bernstein has also found that we can’t prove the assumption that inequality leads to slower growth, given available evidence. It may be true, Bernstein wrote, but we do not have enough concrete proof.”
Adding “concrete” before the word “proof” is a dead giveaway that Karabell is lost. Econometric tests can disprove “that inequality leads to slower growth.” They sometimes fail to disprove a false hypothesis because sufficient “evidence” (data) are not “available” to test the hypothesis. Econometric tests inherently cannot “prove” any “assumption that [X] leads to slower growth.” It doesn’t matter what the “X” is – we cannot prove through econometric tests that the (putative) independent variable causes predictable changes in the (putative) dependent variable. We can find a correlation. We can add “control” variables to try to tease out the impact of alternative independent variables. Neither process allows us to “prove” causality.
Karabell reverses the scientific method and misses its logic. He claims that because econometric tests inherently cannot “prove” that inequality leads to slower growth it follows logically that “income inequality isn’t as harmful as we think.” That is a nonsensical statement. He concedes that if inequality reduces growth it is very harmful. Karabell seems to believe that because an econometric study that shows a negative correlation between inequality and economic growth inherently cannot “prove” causality it must not be causal. We inherently cannot prove through econometric tests that racism is bad for economic growth. This does not, logically, imply that we have proven that racism “isn’t as harmful as we think.”