When an Economic Model Fails to Capture Reality

November 1, 2018

Last week, Senator Kamala Harris (D-CA) introduced a sweeping, bold economic policy idea: the LIFT the Middle Class Act. The LIFT (Livable Incomes for Families) Act would essentially be a dramatic expansion of the Earned Income Tax Credit (EITC), making it much larger and available to many more Americans. A few days later, the conservative Tax Foundation released an analysis of Senator Harris’s proposal that purported to show what economic affect the LIFT Act would have. In a completely counterintuitive finding, the Tax Foundation claimed that putting up to $6,000 into the pockets of nearly every middle-class household would reduce employment by the equivalent of over 800,000 jobs. That finding, of course, is utterly ridiculous.


It is a ridiculous finding for many reasons, the most obvious and clear-cut of which is that there is an enormous body of research that clearly demonstrates that the similarly structured EITC substantially increases work hours and earnings overall—primarily because its dominant effect is to encourage people to join the labor force. One study found, for example, that the pro-work effect of the EITC roughly doubles the number of low-wage single women that the EITC lifts out of poverty. In the real world, tax credits like the one proposed by Senator Harris support and reward work.

But in the Tax Foundation’s model, this well-documented effect is overwhelmed by its finding that middle-income people will choose to work slightly less as the LIFT credit phases out (the theoretical result of slightly higher marginal effective tax rates). This finding runs counter to two decades of rigorous empirical study, where very few people even have the ability to calibrate their work hours as finely as the model assumes they do, and just as few even respond to minute changes in their tax rate that are happening underneath the surface.

The Tax Foundation compounds these problems by reporting their model’s results in a way that is also divorced from the real experiences of real people in the real economy. In this case, by turning this theoretical voluntary reduction in work hours into “lost jobs.”

The Tax Foundation reports that the LIFT Act would reduce “full-time equivalent jobs” by over 800,000. But full-time equivalent jobs is a seriously misleading metric, bordering on unusable. It’s not actually a measure of “jobs” in the way most people think of them. If 40 people decide to work 1 hour less each week, that adds up to one “job lost.” Quite obviously, that is a ridiculous way to think about jobs, especially if those 40 people are materially better off in every real way despite choosing to reduce their work hours slightly.

The problem is not only that the Tax Foundation is an avowedly conservative and corporate-funded institution that calibrates its model to come up with answers that are unfavorable to a Democratic policy proposal, while always supporting corporate tax cuts. The underlying problem here is that models like the one on which the Tax Foundation based its analysis are often fundamentally at odds with reality in numerous ways.

For one thing, many of these models simply do not and cannot accurately account for actual trends in our economy. In order to estimate how a given policy might change the overall economy, conventional macroeconomic models have to make a whole load of assumptions about the relationships between various elements of the economy, as well as the behavior of different “economic agents” like households, businesses, and even the Federal Reserve in response to changes. Obviously, the best models try to base these assumptions as much as possible on empirical evidence. The trouble is that a lot of models are still using outdated assumptions that no longer appear to hold up.

Why, for example, is corporate investment so abysmally low, despite high corporate profits and low corporate taxes? That’s not what traditional assumptions about the economy would lead you to guess would happen under these circumstances. Why have wages been so stagnant despite rises in worker productivity over the last several decades? That’s not what conventional assumptions suggest should happen. And why have interest rates declined precipitously despite an overall increase in public debt as a share of GDP? Again, most models predict exactly the opposite.

If these models are based on an understanding of how the economy works that can’t adequately explain what’s actually happening in the economy, that’s a huge problem. And part of the reason that even the best models are having trouble mimicking the real world is that we have not yet figured out good ways of incorporating some of the latest economic insights into the inner workings of a model.

Take income inequality as an example. In 2011, International Monetary Fund economists Andrew Berg and Jonathan Ostry published a landmark study that showed higher income inequality was a barrier to sustained, durable economic growth. They have since followed that study with numerous others confirming and expanding the original findings. Economists are still investigating exactly how and why it is that higher inequality acts as a destabilizing force on the economy. Macroeconomist and modeler extraordinaire Mark Zandi has struggled with this very issue, concluding that macroeconomists should, “not be comfortable that they have a good grip on what inequality means for our economic prospects.” Zandi’s baseline assumption is that inequality has only small effects on growth, but he admits not having “much confidence” in that view. Until some consensus is reached on the various mechanisms and channels through which inequality exerts downward pressure on the economy, most models essentially ignore this finding entirely. In this specific case, that’s a problem, since Senator Harris’s proposal would substantially reduce income inequality, as even the Tax Foundation itself acknowledges.

In recent years, there has also been an explosion of research into the effects of concentrated market power. When the influence of a handful of corporations in a given industry grows to outsized proportions, various basic assumptions about how “free markets” work begin to break down. For example, recent research from economists Ioana Marinescu, Jose Azar, and Marshall Steinbaum shows that the market for labor is anything but “free,” with so-called “monopsony power” being extremely pervasive. A model that implicitly assumes workers enter into a fair negotiation with their employers on a level playing field is a model that is likely to produce an unrealistic outcome.

We also know that what happens to children early in their lives has huge implications for our economy as a whole. Investments in quality childcare and early learning pay enormous dividends, while toxic stress and financial instability act as anchors on the economy. This is especially relevant, of course, for Senator Harris’s LIFT Act, which would very likely bring the existing benefits of the EITC to many more families. But because those enormous benefits materialize over time, as the affected children grow up, they never show up in traditional models that only look so far into the future. That may be unavoidable—after all, even the best models get pretty fuzzy the further out they try to simulate—but it is important to keep in mind.

One last crucial problem to mention—one that is really not specific to the Tax Foundation: Most consumers of these types of models often focus on changes in gross domestic product, treating GDP as a proxy for real economic outcomes. Fundamentally, a good economy is one in which the vast majority of people are doing better and better: their standard of living is rising, their security and stability is expanding, and their children have a better shot at a better life. A rise in overall GDP does not necessarily speak to any of those things. Especially now, in an economy so stratified by income, by geography, and by race, overall GDP has become a very weak measure of the real economy.

There is obviously a place for macroeconomic models. When well-designed and well-calibrated, these models can give us a sense of what might happen as we make big changes to economic policies. That’s clearly useful. But too many of the conventional models that policymakers and reporters rely on today fail to match up to reality. This may be because they are based on outdated assumptions or they are ideologically driven. It can also be because they have not yet been able to incorporate the latest research on how the economy actually functions into their assumptions. And sadly, even the best models still suffer from misuse because of misleading metrics.

Everything we know about the real-world economy tells us that a proposal like Senator Harris’s would be an economic plus, encouraging work, supporting broad-based and stable demand for goods and services, reducing inequality, giving workers some security, and giving their children a better shot at being healthy, doing well in school, and succeeding in their own lives. We know from the real world that those are all ingredients for a strong economy. When a model’s results suggest the opposite, it’s the model that needs to change, not our understanding of the real world.