Abstract
In The Misuse, Misrepresentation, and Politicization of Statistics in American Society, Robert Parker argues that some of the statistics cited most in our public discourse are routinely misused by commentators and government officials. Parker specifically takes issue with the way four key statistics are deployed: official unemployment rates, life expectancy at birth, Uniform Crime Reports (UCR) crime rates, and U.S. Census population counts. Parker argues that these metrics do not capture the social phenomena they claim to measure and that the primary beneficiaries of this misuse of statistics are neoliberalism and its proponents, for whom such statistics provide unwarranted support.
I briefly summarize each of Parker’s main points and then look more closely at each. I find that Parker’s points have kernels of vital truth, but these truths are buried within mountains of hyperbole.
Parker criticizes the following four key statistics and their problems:
The official unemployment rate, which vastly overestimates the economy’s capacity to employ workers by only counting as unemployed those who have actively looked for work in the previous four weeks, neglecting the underemployed as well as those who stopped looking for a job.
The life expectancy at birth, which is misleadingly used to tout our society’s capacity to prolong life and argue for increases in the age of eligibility for full Social Security benefits, when much of those gains in life expectancy is due to reductions in infant mortality that have no bearing on the longevity of older people.
The Uniform Crime Report (UCR) rates, which are used to track crime rates over time, even though they are a highly selective indicator that omits white-collar crime and cybercrime. These omissions are glaring, according to Parker, because the economic cost of white-collar crime far exceeds that of street crime as captured by the UCR.
The U.S. Census population counts, which are used to allocate federal funding for Medicaid, SNAP, and other benefits, even though post-enumeration surveys by the Census Bureau have shown that these figures undercount minorities.
Each of these points is reasonable on its own and worthy of discussion. Take the official unemployment rate, for example. In recent years, even as unemployment rates have fallen to record lows, the percentage of working-age people who are employed has fallen considerably. There are increasing numbers of Americans classified as out of the labor force due to disability and substantial numbers who have left the labor force due to the outsourcing or automation of jobs and the inability to find work of comparable quality or renumeration. The official unemployment rate fails to capture these important trends.
And Parker is no doubt correct that we overestimate advances in longevity in later life by focusing exclusively on gains in life expectancy from birth. As a zero-inflated distribution, life expectancy from birth is often most clearly analyzed using two distinct models: a binary model predicting zero versus nonzero, and a count model predicting larger versus smaller nonzero numbers. Like many averages of skewed distributions, life expectancy from birth can only provide answers to the most general questions about health and longevity and obscures as much as it illuminates when we look at more specific questions.
Yet undermining Parker’s persuasiveness on these issues is a tendency to take these arguments to an extreme and a failure to consider alternative points of view. Any persuasive case should include a recitation of counterarguments along with rebuttals. Parker doesn’t include much in the way of counterargument and rebuttal and as such fails to take his opponents with sufficient seriousness. Instead, his book takes the tone of a jeremiad, repeating the same points and blaming neoliberalism for nearly everything.
Consider the official unemployment rate. Parker argues that it should include workers who have stopped looking for work, and that the only reason to exclude them is to make capitalism look good—that is, to provide unwarranted support for the status quo. However, Parker is neglecting the complexity of measuring concepts like unemployment and labor force participation. Measuring unemployment is tricky because the reasons why any particular person of prime working age is not working are a mixture of personal choice and external constraint. Why should we assume that constraints predominate? A person may be out of the labor force because there are no available openings in their field of expertise. Yet there are many others who have overcome that situation and are now in the labor force performing other jobs. In a dynamic, flexible labor force, we cannot assume lack of agency.
Furthermore, a more inclusive measure of unemployment has advantages but also disadvantages. The most inclusive measure—the percentage of people of prime working age who are not working—is not unambiguously a sign of an economy’s capacity to employ labor. An economy of high wages in which a family can be supported by a single income would likely have a low labor force participation rate. So would an economy in which workers take breaks from work to pursue hobbies. No single measure will capture this complex reality. Parker is correct that the official unemployment rate, on its own, is inadequate. But so is any other measure, on its own. This is why the Bureau of Labor Statistics presents a panel of monthly statistics, as Parker acknowledges.
Parker claims that the official unemployment measure is a social construction for political purposes rather than an objective reality. He is certainly correct that most commentators who use unemployment statistics tend to have a rosier view of our economic system than he does and are thus less likely to use statistics that paint a darker picture. But the much higher unemployment estimates that Parker prefers, from a website called Shadow Government Statistics, are also social constructions for political purposes, as they were created to reveal flaws in the system. Parker does not explain how these alternative statistics are computed, so we cannot compare them to the current standards, and he does not show how they are more objective than the statistics currently used.
Parker notes here and elsewhere that commonly used statistics are not an objective reality, but he thinks they are often seen as such. Maybe this is a point people need to be reminded of, but there is also a hint of straw man argumentation here.
Sophisticated users of statistics have long believed in the aphorism attributed to George Box that “all models are false, but some are useful.” Few analysts see official employment rates as ontologically similar to any specific person they know to be unemployed. There is at least some understanding that statistics are estimated quantities about which there is substantial uncertainty.
Parker’s case against the use of life expectancy at birth is also useful yet overly tendentious. He is right that life expectancy at birth gives an overly optimistic picture of improvement in later-life longevity and cannot by itself be used to justify increasing the Social Security eligibility age. But a recent proposal to increase the eligibility age for Social Security by three years is not inconsistent with increases in later-life longevity. Life expectancy in the United States at age 65 has increased by three years since 1980 (CDC 2024), and the Social Security retirement age was last changed in 1984.
Parker emphasizes the regressivity built into Social Security. Higher earners live longer and will thus earn more in benefits than lower earners. This is certainly true, but the question is how much. And economic research on this question has found that differential mortality by SES plays at most a minor role on the impact of Social Security on inequality (Harris and Sabelhaus 2005).
In Parker’s claims about crime statistics, I find truth amid exaggeration. He also tends to blame a neoliberal agenda as the sole cause when there are simpler explanations.
For example, Parker is right that the Uniform Crime Reports get too much attention, given their inability to capture the “dark figure of crime.” But are capitalism and neoliberalism the root of this problem, as he argues? When we look at aggregated police reports for crime data, aren’t we just looking for the keys next to the lamppost because that’s where the light is? Crime is often hard to measure because it is underreported, so our first instinct is to look at what has been reported. But we know that unreported crime is usually estimated with the National Crime Victimization Survey. If that survey had not been running since 1973, Parker’s critique would be more trenchant.
Parker also argues that our focus on the UCR makes us ignore white collar crime. This is true, but Parker also claims (following Edwin Sutherland) that the economic costs of white-collar crime are much higher than the economic costs of street crime. He further argues that the focus we put on street crime is the result of a culture and society that serve their capitalist masters.
I see several problems with these arguments. First, a common definition of white-collar crime is the use or abuse of one’s position for economic gain. So, of course if we define a type of crime by economic gain, then it’s likely to have a higher economic cost than another type of crime that isn’t defined that way.
More importantly, the estimate of the economic cost of street crime may include only the value of stolen property, and that is often only a fraction of the total cost. So much of the cost of street crime is left out of such estimates, such as medical costs of treating injuries, the pain and suffering from those injuries, the value of personal security lost, and above all, the value of lives lost from homicide. Economists often estimate the cost of a crime by how much a person would pay to avoid it. Most people would pay a lot to avoid being the victim of a violent attack, or worse. This is why protection rackets can work when the state is weak. Business owners choose to pay large sums when there is a real threat of violence. People usually value their health as much as or more than their money. So, any cost estimate of street crime that doesn’t account for this is missing a lot. If such costs were added to the total cost of street crime, it’s unclear whether they would still be lower than the costs of white-collar crime.
Parker suggests that the status of the offenders influences the focus on street crime over white collar crime in society. But he overlooks some important caveats. Most white-collar criminals are not elites, but people who vary in status but have positions that allow them to steal. White-collar crime is thus committed by retail clerks, truck drivers, and warehouse workers as much as by C-suite executives.
Parker ignores another factor that could make white-collar crime less severe than street crime. The direct victims of white-collar crime are usually better off than the victims of street crime. Many victims of white-collar crime are businesses that have factored the costs into their business model. The restaurant owner knows and plans for some food being eaten by their employees. Stores budget for stolen goods, and credit card companies budget for fraud. Consumers ultimately pay for these crimes through higher prices. But this way, the costs of much white-collar crime are spread across many people. And people are likely to tolerate diffuse costs more than violent or property crime committed against individuals, no matter how capitalistic a society is.
Another reason why street crime gets more attention than white-collar crime is that the latter is harder to measure. Estimates of the costs of white-collar crime vary widely and make point estimates less useful. As with the unemployment rate, Parker underestimates the complexity and ambiguity of measurement.
Also, victim surveys are not enough to measure white-collar crime because many people don’t know they have been victims. And many of the most profitable cases of white-collar crime depend on technicalities. For example, at investment firms, what is the difference between using insider information and research? Many investments firms’ actions in this area are close to the legal limit. But where is that line? When the difference between legal and illegal actions depends on complex and unclear rules, there will be less public interest and less public anger. It is much harder for people to care about things that need to be explained to them.
Finally, why should we expect white-collar crime to be less serious in a pro-capitalist society than in a socialist one? According to the typical laissez-faire economist, the state’s role is to protect property rights. White-collar crime is about violating property rights. One might then expect that a society that favors capitalism would take property crime more seriously than violent crime. According to neoliberal ideology and the logic of market forces that only focus on profit, human lives and well-being would matter less than property. But this is not how we act in the United States. People generally see violent crime as more harmful. The murderer and the rapist cause more desire for revenge than the tax cheat.
Likewise, if capitalist ideology was driving our crime focus, we might expect crimes against large corporations to be perceived as especially problematic. But we often see the opposite. Theft from large corporations and from rich people is usually seen as less serious than theft from individuals. For example, many more people would steal music from record labels than would steal a CD from an individual. It is easier for people to rationalize theft when the victim is a corporation rather than an individual. In fact, many people view such theft is harmless.
Regarding the Census, again Parker makes some good points but exaggerates his case. He points out that post-enumeration sampling showed significant errors in 14 states in the 2020 count. These errors are more worrying because Blacks and Hispanics were undercounted by 3.3 percent and 4.99 percent, respectively. Parker is right—politicians should think about adding statistical adjustments to official counts. But he goes too far in asserting that undercounts of minorities necessarily mean they will not get their fair share of Census-based federal funding.
The question is important. As of 2016, about $500 billion in federal funding to states was linked to Census counts. The biggest budget item in this spending is Medicaid, which is based on a formula that gives more funds to states with lower per capita incomes. If a state’s population is undercounted, its per capita income will be overestimated, and it will get less financial support than it should.
But how much will 2020 Census errors affect minorities differently? We can’t answer this question based only on data on different undercounts by race and ethnicity nationwide. Since federal funding goes through states, the answer depends partly on how undercounting is related to minority population share across states.
When I run linear regressions at the state level of the 2020 Census error rate (as estimated by the post-enumeration survey) on percent Black and on percent Hispanic, I find that undercounting is a bigger problem for the former than for the latter. I weighted the data by state population so that states count according to their population. The (weighted) correlation between a state’s 2020 Census error rate and percent Hispanic in that state is low (-0.04) and not statistically significant. The correlation between the error rate and state percent Black is high (-0.36), showing that states with larger percent Black populations have more undercounting. This relationship remains when controlling for state GDP per capita.
My (preliminary) analysis suggests that Parker is right to highlight the undercounting of minorities on the Census, but the reality is more complicated than a simple critique might suggest. Blacks seem to be the main victims of the recent Census undercounts, but more research is needed to show how funds are allocated differently by race. This analysis is tricky because the post-enumeration survey is not big enough to give accurate undercounts by race or ethnicity within states. Nonetheless, a detailed, thorough analysis is needed. The extant analyses on the question focus on specific states or metro areas. To show a different impact by race or ethnicity nationwide, one needs to combine data across states and deal with how the Medicaid funding formula is affected.
Despite its exaggeration, Parker’s book is worth reading, as it raises important questions about metrics that we often assume are accurate—questions that have not been studied enough elsewhere. Parker is mostly right in his main claim that the link between the most common metrics and the concepts they measure is weaker than people think. This is a point worth exploring whether the cause is a hidden plan to support capitalism, as Parker thinks, or simply a result of oversimplification of complex phenomena.