Improving Economic Models: The Pursuit of Being Less Wrong (But Never Right)
Simplifications lay at the heart of every economic model. From introductory classes onward, professors present students with models held together by seemingly impractical assumptions that “consumers are rational” or “there is complete market information.” As these young economists further their careers and build their own models, they find that their reliance on assumptions does not change. What changes are the assumptions they make, and how false they are. As Stanford Professor John Cochrane said, “All economic theorems are false of course, in that the assumptions are not literally true. The question is, how false?” (Cochrane 2023). But in a world of economics where complexity abounds and simplifications are needed, one simplification stands out: boiling everyone down into a single figure known as the “representative agent.” For half a century, this mythical agent has been a cornerstone of economic analysis, used to forecast the macroeconomic responses to policy. The most common representative agent model is called “RANK” (“Representative Agent New Keynesian”). However, few individuals resemble this representative agent in the real world. As Peter Coy, an economics columnist for the New York Times, explains, “The representative agent is fully rational and unemotional, forward-looking and with perfect information about all the relevant facts. Just like nobody you know” (Coy 2023). To remedy the problems of using one sole (un)representative agent, a group of macroeconomists are championing a new framework called “HANK”— short for “Heterogeneous Agent New Keynesian.” By composing the model of many diverse agents, HANK models capture both the direct and indirect effects of monetary policy, reveal how the composition of incomes moderates the effects of monetary policy, and emphasize the redistributive qualities of monetary policy. As such, the development of the HANK framework and heterogeneous agent models, more generally, will enhance the predictive power of economic models and force economists to re-evaluate the conventional wisdom of monetary policy.
The Representative Agent and New Keynesian models
Before examining these models, it is crucial to understand that representative and heterogeneous agents extend beyond New Keynesian models. RANK and HANK are simply the New Keynesian flavor of these frameworks, hence the “NK.” Representative agent and New Keynesian models gained traction during the 1970s after economists such as Nobel prize winner Robert Lucas Jr. pioneered a shift in macroeconomic modeling now known as the “micro-foundations of macroeconomics.” Lucas and other economists argued that focusing on individual behavior was essential to understanding and predicting macroeconomic phenomena (Tori and Witynski 2023). From this, the representative agent and New Keynesianism emerged.
But if RANK models raise such evident problems, why are they still used after nearly five decades? In reality, the root of their problems is what makes them useful: they are simple. The simplicity of RANK models allows economists to easily control for specific changes and manipulate variables without overcomplicating the model. Often, these simple models provide better predictions than more sophisticated ones. Thus, their simplicity does not sap these models of value; their simplicity makes them valuable (Cochrane 2023). However, as policy concerns shift, our models must evolve in tandem, incorporating broader policy goals from climate change to geo-political risks to income distributions (Moll, Kaplan, and Violante 2023). As such, navigating a wider, more complex economic domain will require a comprehensive understanding of monetary policy’s effects. RANK models tell one side of this story but only capture the direct effects of monetary policy for one agent. HANK models, on the other hand, paint a more complete picture by measuring both the direct and indirect impacts of monetary policy on millions of heterogeneous agents (Moll 2020). Therefore, while representative agents have proved an indispensable tool, HANK models merit further examination.
HANK vs RANK
At the core of every HANK are many diverse agents who differ in socioeconomic status, spending habits, and other intrinsic characteristics. Therefore, while one representative agent fails to account for wealth and income distributions, having various heterogeneous agents shows how policy changes can affect existing inequalities. In other words, rather than inferring what is good for the “typical” consumer, HANK models reveal policy implications for different individuals and the broader economy. Higher interest rates, for example, do not impact everyone equally—they have heterogeneous effects on the economy: some people (savers) benefit, and others (spenders) suffer (Coy 2023). More importantly, HANK models provide insights into both the individuals affected by monetary policy and a clearer understanding of the effects themselves.
The monetary transmission mechanism is a framework used by central banks to predict the effects of monetary policy on the economy’s aggregate performance while factoring in influences of other economic forces, such as changes in fiscal policy or the global economy (Ireland 2005). Within this framework, changes in monetary policy impact household consumption through two main channels: one direct and the other indirect (Moll 2020). On the one hand, if a central bank were to raise interest rates, the direct channel would capture if individuals were more willing to postpone purchases. This is referred to as a consumer’s elasticity of intertemporal substitution—the higher the elasticity, the more responsive individuals are to changes in interest rates. On the other hand, an increase in rates would also indirectly impact other sources of expenditures through indirect channels. This includes mortgage rates, labor income, and fiscal measures (Moll, Kaplan, and Violante 2023). These distinctions are essential when comparing the differences between the RANK and HANK models.
Since RANK models predict the effects of monetary policy on the “typical” permanent income consumer, their marginal propensity to consume (MPC) is relatively small, which renders the indirect effects negligible (Moll, Kaplan, and Violante 2018). This underscores that RANK models only consider the direct channels of monetary policy–through the elasticity of intertemporal substitution mentioned above—and are unresponsive to transitory income shocks. However, this does not hold in practice. Instead, by incorporating numerous heterogeneous agents, HANK models can more accurately reflect the data and the array of responses taken by different individuals (Auclert 2019). Moll, Kaplan, and Violante (2023) find that the MPCs in HANK models are nearly ten times larger than those found in RANK and other traditional models. Therefore, unlike representative agent models, indirect effects drive monetary transmission in HANK models, while the direct effects may prove more modest (Moll, Kaplan, and Violante 2018).
These differences between both models yield several implications. First, by constructing models that weave diverse agents into the equation, HANK models capture responses that RANK models would otherwise fail to convey. For instance, consumers who live hand-to-mouth are more responsive to labor income shocks (an indirect effect of monetary transmission) than to direct changes from interest rates. Thus, by expanding their models to account for population heterogeneity and, consequently, indirect effects, HANK models provide a more comprehensive understanding of the impact of changes in monetary policy (Coy 2023). However, this leads to the second key implication, which points back to an earlier point of tractability: since RANK models only account for direct effects, they require just one parameter—the elasticity of intertemporal substitution—whereas HANK models demand an exhaustive understanding of market structures and the role of other institutions such as the government. As Moll, Kaplan, and Violante (2023) explain, “[HANK models] need a full picture of the distribution of MPCs, income sources, and the components of household balance sheets.” Thus, HANK models require a vast amount of data. For countries that suffer from a lack of data accessibility and transparency, this intrinsic feature could pose a problem. Third, if traditional models have omitted indirect channels, we must re-evaluate our understanding of the monetary transmission mechanism. In other words, we must reassess how the ensemble of economic forces works together to orchestrate changes in the aggregate economy. This warrants a deeper investigation into the redistributive qualities of monetary policy and how they challenge our conventional insights on the relationship between monetary and fiscal authorities.
The Eye-Opening Insights of HANK
Moll, Kaplan, and Violante (2023) show that changes in monetary policy have differential effects on the population. Contrary to conventional monetary wisdom, a rising tide does not raise all ships. Shifts in the interest rate elicit various responses across the economy; some are more affected by indirect effects and fluctuations than others. When central banks cut rates, savers (those with low MPCs) lose interest on their income, whereas spenders (those with high MPCs) benefit from an increase in disposable income—since falling interest rates devalue existing debts. By contrast, previous models essentially ignored debt. Peter Coy (2023) underscores this feature, noting that in “a pure representative agent model, debt isn’t really a thing: Some people borrow, others lend, but it ends up a wash.” It is crucial to understand that when central banks cut interest rates, they not only redistribute from savers to spenders, but also boost business investment, stimulate the housing market, and increase consumer confidence, among other things. As a result, these additional effects experienced through indirect channels amplify the macroeconomic effects of monetary policy. Conversely, raising interest rates dampens the effects of monetary policy by redistributing from spenders to savers and suppressing economic activity. The fact that HANK models distinguish between savers and spenders is important because it allows economists to better forecast how adjustments in the interest rate redistribute between these two groups and how these changes reverberate throughout the economy. In other words, HANK models predict how income distributions influence the effectiveness of monetary policy, forcing economists to recognize the link between redistribution and stabilization (Violante 2021). These predictions can ultimately clarify the “winners and losers” of monetary policy, showing that when the tide of monetary policy rises, not all ships rise with it: “some ships are lifted higher, others are sunk” (Moll, Kaplan, and Violante 2023).
Given that redistribution cannot be separated from stabilization, HANK models require economists to put monetary policy in dialogue with its fiscal counterparts. While economists have historically viewed redistribution as a fiscal responsibility, HANK models reveal that monetary policy operates as another redistribution channel, as mentioned above. However, this does not mean monetary policy should not be the preferred tool for redistribution; as Moll, Kaplan, and Violante (2023) emphasize, “monetary policy is a blunt tool for redistribution.” Rather, what is crucial is understanding how monetary policy’s redistributive properties establish an inextricable link with fiscal policy and its goals. Indeed, just as monetary policy impacts fiscal policy, fiscal policy also plays a central role in determining the effectiveness of monetary policy.
Similar to how income distributions influence the effects of monetary policy, HANK models also clarify how fiscal policy moderates the impact of monetary policy. As such, HANK models exhibit how placing the two in conversation is necessary in order to paint a clearer picture of how one’s actions magnify or mute the other’s. Households are not the only ones who respond to tweaks in interest rates. The government must react as well. Moll, Kaplan, and Violante (2023) note, “When the central bank raises interest rates, the treasury’s borrowing costs increase, and the increase must be funded by raising taxes or lowering expenditures, now or in the future, or through future inflation. In HANK models, the details of how and when the government makes up this fiscal shortfall, and which households bear the burden, have a tremendous influence on the overall effects of interest rate hikes.” Simply put, the government is a fundamental actor in the transmission of monetary policy; however, unlike previous frameworks, HANK models can more accurately predict the effects of monetary policy by considering the government’s response. These government responses to monetary policy are known as “fiscal footprints,” which critically shape the effects of monetary shocks. In this light, HANK models emphasize that monetary and fiscal authorities must work cooperatively to achieve their economic goals (Moll, Kaplan, and Violante 2023).
The Future of HANK
In brief, the development of HANK and heterogeneous agent models could revolutionize the field of economics. By incorporating many diverse agents, they capture both the direct and indirect effects of monetary policy, unveil how income distributions and fiscal policy impact the effects of monetary policy, and highlight the redistributive qualities of monetary policy. These features not only improve the predictive power of economic forecasts but also force economists to reassess their beliefs about monetary policy. Specifically, they can no longer divorce stabilization from redistribution or separate monetary policy from fiscal policy (Moll, Kaplan, and Violante 2023).
Nevertheless, while heterogeneous agent models have existed for nearly three decades and New Keynesian for even longer, HANK models are a nascent contribution to the field. As such, they still have to overcome and adapt to their criticisms. John Cochrane, a professor at Stanford and proponent of the Fiscal Theory of the Price Level, has voiced his issues with HANK models. However, it appears his concerns stem more from New Keynesian models than the heterogeneous agents themselves (Cochrane 2023). Remember, HANK is just the New Keynesian version of heterogeneous agent models. Indeed, in a recent piece, Cochrane proposed the first HAFT —heterogeneous agent fiscal theory—model, suggesting that heterogeneous agents are valuable for all the reasons listed above, but adapting them to other models may be necessary (Cochrane 2023). Cochrane has pointed to several crucial faults in New Keynesian models that have been left unresolved for decades, such as a fundamentally flawed New Keynesian Phillips Curve that suggests output is high when inflation is expected to fall.
Regardless, the future for heterogeneous agent models, New Keynesian or not, is promising. However, this is not to say that they will replace representative agent models. In the same way that behavioral economics does not replace neoclassical theory, heterogeneous agents supplement economists with another powerful tool. In all their inherently false assumptions, the beauty of simple models like representative agent models is that they can make such false assumptions and still make reasonably good predictions. The key is not to use one over the other but to use them in unison. Ultimately, all economic models are “wrong” in that false assumptions hold them together (Cochrane 2023). But no model can or will ever be 100% right. In times like these, it is important to remember the old aphorism: “All models are wrong, but some are useful.” Because, in the end, improving economic models is the relentless commitment to furthering a field, knowing you will never be right, just trying to be less wrong.
References
Auclert, Adrien. "Monetary Policy and the Redistribution Channel." American Economic Review 109, no. 6 (2019). https://doi.org/10.1257/aer.20160137.
Cochrane, John. "Heterogeneous Agent Fiscal Theory." Entry posted October 13, 2023. https://johnhcochrane.blogspot.com/2023/10/heterogeneous-agent-fiscal-theory.html.
———. "New York Times on HANK, and questions." Entry posted July 11, 2023. https://johnhcochrane.blogspot.com/2023/07/new-york-times-on-hank-and-questions.html.
Coy, Peter. "The 'Representative Agent' Is Always Rational. The Rest of Us Are Not." New York Times, July 10, 2023. https://www.nytimes.com/2023/07/10/opinion/economic-modeling-hank-representative-agent.html.
Ireland, Peter N. "The Monetary Transmission Mechanism." FRB of Boston Working Paper No. 06-1, November 2005. Accessed at SSRN: https://ssrn.com/abstract=887524 or DOI: 10.2139/ssrn.887524.
Lee, Tori, and Max Witynski. "Robert E. Lucas Jr., Nobel laureate and pioneering economist, 1937-2023." Uchicago News. Last modified May 16, 2023. https://news.uchicago.edu/story/robert-e-lucas-jr-nobel-laureate-and-pioneering-economist-1937-2023.
Moll, Benjamin. "Ben Moll on the Rich Interactions between Inequality and the Macroeconomy." EconomicDynamics Research Agenda 21, no. 2 (2020). https://www.economicdynamics.org/research-agenda-moll2020/.
Moll, Benjamin, Greg Kaplan, and Giovanni L. Violante. "Monetary Policy According to HANK." American Economic Review 108, no. 3 (2018): 697-743. https://doi.org/10.1257/aer.20160042.
———. "THE VERY MODEL OF MODERN MONETARY POLICY." Finance and Development. https://www.imf.org/en/Publications/fandd/issues/2023/03/modern-monetary-policy-kaplan-moll-violante.
Violante, Gianluca. "What Have We Learned From HANK Models, Thus Far?" Speech presented at 2021 ECB Forum on Central Banking, Sintra, Portugal, September 28, 2021. European Central Bank. https://www.ecb.europa.eu/pub/conferences/ecbforum/shared/pdf/2021/Violante_presentation.en.pdf.