Tracking Business Incomes And Investments

This essay is based on the working paper “On the Nature of Entrepreneurship” by Anmol Bhandari, Tobey Kass, Thomas J. May, Ellen R. McGrattan, and Evan Schulz.

In 2019, the Internal Revenue Service reported that total receipts less deductions for businesses in the United States—corporations, partnerships, and proprietorships—totaled three trillion dollars. The Bureau of Economic Analysis (BEA) used these estimates to compile the National Income and Product Accounts, but in doing so added 1.1 trillion dollars as an adjustment to account for underreported income and the illegal filings of tax returns. While some misreporting can be uncovered in IRS audits, tight budgets mean that most misreported income goes untaxed, leaving less revenue for infrastructure, education, and other investments that are crucial for long-run growth and prosperity. Research efforts are underway to crack down on tax evasion by business owners, many of whom run private companies with little to no regulatory oversight. Critical to this research is a better understanding of the nature of entrepreneurship, specifically the returns to and determinants of business ownership, and the types of investments made by business owners. 

Past studies of entrepreneurship that rely primarily on survey evidence portray business owners as individuals choosing self-employment for the nonpecuniary benefits. For example, there may be advantages to having flexible hours and creative control. This good-lifestyle view is perhaps not surprising given that survey data for business owners with high net incomes or large business losses are censored because of privacy concerns. With the top and bottom censored, most of the attention in the economics literature on entrepreneurship has been on median incomes that are relatively low when compared to incomes of individuals in paid employment who have similar characteristics. From the vantage point of the tax administrator, these low incomes might signal noncompliance, but random audits would not net any additional tax revenue if lower pay is simply a reflection of higher amenity value. As the BEA data make clear, the IRS auditors are indeed finding significant noncompliance. 

To address this noncompliance, we need to study the whole distribution of business incomes, including those at the top and the bottom. In our paper, “On the Nature of Entrepreneurship,” my coauthors and I reassess the survey data evidence and associated conclusions about business incomes and owners of S corporations, partnerships, and sole proprietorships (also known collectively as pass-through entities because the business incomes are passed through to the individual tax form). To do this, we first assemble a novel longitudinal database of business owners using the administrative data from the Internal Revenue Service, which have been appropriately disaggregated to ensure no confidential information is disclosed. More specifically, we construct a balanced panel of all living individuals born over the period 1950–1975 (with data in the period 2000–2015). For each individual and tax year, we compute total paid-employment income derived from Form W-2 wages and total self-employment income derived from pass-through business ownership. We gather information about demographics (for example, birth year, gender, marital status, children), professions (for example, industry and occupation), other incomes (for example, asset and spousal incomes), and imputed levels of education and skill sets. We then partition our sample on the basis of employment attachment in order to separately study individuals who switch frequently between self-, paid-, and non-employment and those who are more “attached” to their employment status. Individuals are categorized as attached to a particular employment status if they have fewer than two switches in status during our sixteen-year sample period and no intermediate spells of non-employment. This notion of attachment is especially relevant in the case of owners who make firm-specific investments—for example, those building a customer base or client list—before production can begin. There is no counterpart of this kind of activity in paid employment. As a result, entrepreneurs growing businesses could potentially have different life-cycle income patterns from paid employees or individuals switching between self- and paid employment. 

With these data and concepts of employment status, we estimate life-cycle income profiles for each subgroup in our sample using a flexible specification for the income process. The income process has three components. The first is an individual-specific component that captures one’s latent characteristics. The second is a time effect that captures changes in income specific to the sample, like recessions and booms. The third is an age effect that captures changes in income that occur because individuals might gain experience, learn on the job, or, in the case of business owners, grow the business. By taking differences in income—and hence by studying income growth—we can identify time and age effects once we impose two conditions. The first condition is that age effects are the same across groups of individuals in the same binned cohort (containing at least two birth years). The second condition is that the growth rate of the time effects is the same for all individuals in a group. 

When we compare income profiles of individuals who are categorized as attached self-employed to peers in paid employment with similar demographics and labor market characteristics (other than employment status), we find stark differences. The business owners in our attached self-employed group have significantly steeper life-cycle profiles, with higher and more hump-shaped growth profiles than their attached paid employed peers. As an example, consider first the average income of a group of accountants who start their own firms at age twenty-five and spend their early years building their client lists. Compare this to the average income of a second group of accountants who start as associates at one of the Big 4 Accounting Firms and work with the firm’s existing clients. The first group will likely have relatively low or negative incomes initially. However, if their investments pay off, they will have high business earnings later in the life cycle. They might also have higher incomes as they gain experience on the job, but the same is true of paid-employed peers. The main difference between the accountants in the example is the payment for firm-specific investments that the entrepreneur earns in addition to payments for time in production. 

We find that by age fifty-five, the average income of the attached self-employed group is $210,000 (in 2012 dollars), which is more than twice that of their attached paid-employed peers at $89,000. I should note that the relative ranking of average incomes for the attached groups is in no way a foregone conclusion because we work with a balanced panel and do not precondition assignment to attached status on life-time incomes. If most business income is attributed to low-income individuals who remain in business because of nonpecuniary benefits, we would have found that their paid-employed peers have higher incomes on average. If most business income is attributed to individuals with net income losses who remain in business because they can effectively consume on the job and avoid taxation, we would have found that their paid-employed peers had higher incomes on average. In fact, while there are many of these low- and negative-income individuals in our attached self-employed sample, they do not account for most business income. Most business income is earned by those ranked in the top 25 percent of lifetime income. Their growth profiles are higher and more hump-shaped on average than those of the attached paid-employed group, even though the paid-employed group includes many successful chief executive officers who took their companies public when younger. 

If we were to adjust our estimates to account for misreporting, then average self-employed earnings would be on the order of 60 percent higher than what we report. In other words, at age fifty-five, we would predict an average “true” income of $335,000 rather than the “reported” $210,000 (in 2012 dollars). Adding up these unreported incomes across all of the owners in the sample yields a significant sum and, if collected, could significantly shrink the gap between taxes owed and taxes collected, commonly referred to as the “tax gap.” This is the argument made recently by the IRS. In response, Congress allocated $80 billion over ten years as part of the Inflation Reduction Act. 

But knowing that business owners are a large source of the tax gap is only one small step in the longer-run solution. Without third-party information on most transactions, auditing businesses and their owners will remain a challenge even with an increased enforcement budget. To meet this challenge, researchers inside and outside of the IRS are trying to better understand behavioral responses when there are changes in tax policy and administration. To do this, we need to have better predictions of how taxpayers will behave if policy changes. For example, how many individuals will choose to run their own businesses if audit rates and penalties rise? What legal forms will they choose? Will business investments rise or fall? 

In our work, we have attempted to provide the inputs to improve economic theorizing about business entrepreneurship in order to answer these questions. For example, we have studied the risks that business owners face. When we compare the dispersion of incomes for individuals who are categorized as attached self-employed to attached paid-employed peers with similar demographics and labor market characteristics (other than employment status), we find it is about three times greater for the business owners than employees throughout the life cycle. For both groups, we find that the dispersion in incomes falls over the life cycle. In other words, it is not the case that growth in income levels is simply compensation for higher income variances over time. This is relevant for tax administrators, who might assume that occupational choice is driven by preferences, say, because they think people differ in their risk tolerances or their love of running a business. 

Relatedly, we have studied the differences between individuals who enter self-employment in a current year and those who have similar characteristics but enter later. Differences in these individuals tell us something about their opportunities. For example, using our longitudinal database, we can compare past labor incomes and asset incomes for people who switch from paid employment to self-employment. Most entering in the current year have higher past labor incomes than their non-switching peers, suggesting that self-employment is not the backstop for individuals with poor employment prospects. Most entering have lower past asset incomes than their non-switching peers, suggesting that liquidity constraints are not a first-order deterrent to entrepreneurship. 

Using our longitudinal database, we can also study individuals who start businesses while young, say in their twenties and early thirties, as they may be most affected by changes in policy. Specifically, we track individuals who ran a business for at least five years before age thirty-five and then compare those who continue in self-employment after age thirty-five with those who do not. Interestingly, those who stay in business have growth profiles that are higher and more hump-shaped then those who exit, which is similar to what we found when comparing the attached self-employed to the attached paid-employed. The businesses of those who remain in business do experience losses initially, but eventually their investments pay off. 

Overall, the picture emerging from our analysis of the administrative IRS data is very different than that based on survey data. Surveys track the typical self-employed individual. With the administrative data, we can track the typical dollar earned in self-employment as well as the typical individual. The typical dollar is most relevant for tax administration and for those concerned about long-run prosperity and growth. 

Read the full working paper here.

Ellen McGrattan is a Visiting Fellow at the Hoover Institution and a consultant at the Federal Reserve Bank of Minneapolis, a professor of economics at the University of Minnesota, and director of the Heller-Hurwicz Economics Institute. She is also a research associate at the National Bureau of Economic Research, a Fellow of the Econometric Society, a Fellow of the Society for the Advancement of Economic Theory, a member of the Bureau of Economic Analysis Advisory Committee, a member of the Minnesota Population Center Advisory Board, and President-elect of the Midwest Economics Association.

Research briefings highlight the findings of research featured in the Long-Run Prosperity Working Paper Series and broaden our understanding of what drives long-run economic growth.