Questions and answers about
the economy.

How are economic models adapting to rising inequality and the pandemic?

Over the past 15 years economic models have evolved to place the differences between individuals at their heart. This new breed of model is helping us understand the implications of rising inequality, the pandemic and the recovery.

The world is unequal both in its downs and its ups: the costs of recessions hit some harder than others, the benefits of economic growth lift some and leave others behind. Economy-wide events therefore affect inequality between the economy’s inhabitants. In turn, how the overall economy recovers after recessions and grows in the long-run depends on inequality. Angus Deaton captured these interactions succinctly in his 2016 Nobel Prize Lecture:

“While we often must focus on aggregates for macroeconomic policy, it is impossible to think coherently about national well-being while ignoring inequality and poverty, neither of which is visible in aggregate data. Indeed, and except in exceptional cases, macroeconomic aggregates themselves depend on distribution.”

While feedback loops between economy-wide events and inequality have always existed, they have been particularly pronounced during the COVID-19 pandemic. While many households have experienced extreme hardship, others have emerged relatively unscathed or have even benefited from the crisis in economic terms. Can models help us understand the interaction between inequality and the macroeconomy? If so, how can they help us make better policy decisions?

What do macro models try to do?

Most crucial economic questions cannot be answered solely by analysing data. One simple reason is that we do not have enough of it: large shocks, like COVID-19 or the financial crisis, are infrequent so that historical experience, and the data it brings, are limited. In addition, economic measures like growth and inflation affect each other, making it challenging to separate correlation from causation. Unlike natural scientists, we can’t run realistic large scale ‘experiments’ to understand the transmission mechanisms for policy tools like interest rates or government spending. There has been substantial progress in producing soundly ‘identified’ evidence (research that successfully teases correlation from causation), but despite this progress we are far from understanding the macroeconomy through the lens of data alone. We therefore need to use theoretical models to help us tie such evidence into coherent narratives about the workings of the macroeconomy.

Macroeconomic models are not intended to be a perfect replication of the world. Instead, they are tools that set out some of the key mechanisms that affect the economy and help us ensure that our intuitions about how the economy works add up. To see this, consider the analogy of the London Tube map (Attanasio et al, 2017). Laying out tube lines in a simplified way makes planning your route much easier, even though the map is very unrealistic compared to a geographically accurate map. Similarly, in economic models, we capture elements of the macroeconomy that we are interested in - for instance, limits on borrowing or wealth inequality - and make other assumptions to keep the rest of the model simple enough to use. Many assumptions in the model may appear very unrealistic; the aim is that they still allow us to capture the aspects of the macroeconomy that we would like to understand.

When are these simplified models too simple?

One of the simplifying assumptions that many models used since the 1970s make is that there is a ‘representative’ household in the economy. Assuming that individual households act approximately like an ‘average’ household makes these models easier and is a reasonable starting point in some contexts. These models combine this average household with simple versions of how firms, governments and central banks behave and are used to understand the effects of economic shifts, like a change in interest rates.

To some extent, this ‘representative household’ assumption built on earlier work going back to the 1930s when macroeconomics was established as a separate field. Early macroeconomists like John Maynard Keynes often started by thinking about economy-wide aggregates – aggregate consumption, investment and government spending, for instance – and focused less on the individual decision-making that determined these variables. The early models that developed this approach, such as the commonly taught IS-LM model, continued in this vein. After the 1970s, economists began to model more seriously the individual behaviour that underpinned economy-wide aggregates, but they simplified by assuming individual behaviour can be approximated by an `average’ individual.

While this simplification is useful it comes at a cost. First, we can’t answer questions relating to the distribution of economic outcomes - for instance, how does income and wealth inequality change during a recession? Second, there are many instances in which the assumption that differences between households ‘wash out’ overall may be incorrect. In these instances, macroeconomic outcomes—and as a result the policy recommendations we make—would be different.

In reality, there are important differences in the economic circumstances that individuals and households face. Incomes vary substantially and might change suddenly after a job loss or promotion. Some households live paycheck-to-paycheck, while others have large pool of savings to rely on or find it easy to borrow on their credit cards or against a house if needed.

How do models take account of inequality?

Economists have long sought to reflect this: there is a history of models designed to understand when and how these inequalities may affect the macroeconomy. Some of the first examples of models of the distribution of income were developed by Kaldor, Pasinetti and others after the Second World War. These often focused on the differences in income between classes, in particular workers versus capitalists.

Beginning in the late 1980s and 1990s, more modern macroeconomic models of the distribution of income and wealth began to be developed in work by Aiyagari (1994), Bewley (1986), Huggett (1993), Imrohoroglu (1989), Den Haan (1997), Krusell and Smith (1998) and others. Typically called ‘heterogenous agent’ (HA) models they include more realistic differences between households, for example differences in income and uncertainty over future income.

Despite these differences most households in these models act quite similarly after a macroeconomic shock. A poorer household consumes less than a richer household, but in response to a recession makes similar changes to how they spend and save. As a result, in many of these models the distributions of income and wealth did not substantially affect macroeconomic outcomes (though, even at the time, there were some important exceptions (Galor and Zeira, 1993; section 4 of Krusell and Smith, 1998).

How have models changed since the 2008 financial crisis?

The financial crisis laid bare the fact that households’ differing financial health (for instance, their levels of mortgage debt or access to liquid savings) are important. These differences have a key role to play in many of the macroeconomic models developed since. One example are models referred to as ‘Heterogenous Agent New Keynesian’ (HANK) models (see for example Kaplan et al (2018) and the more informal discussion here). This approach builds on HA models but additionally includes sticky wages or sticky prices (the slow adjustment of pay and prices was a property of economies that John Maynard Keynes wrote and worried about - hence their ‘New Keynesian’ label).

As well as inequality of income and wealth, these new models include details of household borrowing and balance sheets such as illiquid assets (things like pension wealth and houses, which aren’t easy to draw on in times of financial stress). In these models, different households respond very differently to economic shocks. Some households live paycheck-to-paycheck and must cut their spending substantially if their income falls suddenly, either because they have small saving buffers or because they aren’t able to rapidly sell their assets. Other households can use their savings and ‘smooth’ their consumption.

These differences mean that inequality plays a major role in the response to shocks or policy changes. For example, an interest rate cut affects households not only directly (lower interest rates motivate households to save less or borrow more and thus spend more) but importantly also indirectly by affecting their labour incomes which then results in additional spending. On the one hand, a country’s wealth distribution and its inhabitants’ portfolio composition (i.e. the nature of the assets households can call on) determine the strength of these direct and indirect channels. This is because these factors determine how much of an unexpected change in income or wealth households will spend, vital measures known as marginal propensities to consume, or MPCs. On the other hand, how the distribution and composition of individual incomes comove with GDP determines the impact of monetary policy and its redistributive consequences.

How well do these models fit the data?

A challenge for modern economic models is to match the data on how both individuals and the economy as a whole respond to shocks or changes in policy. For example, in the previous representative household and HA models, too many households were able to smooth consumption after shocks, which was at odds with the data. Though it is difficult to differentiate between causation and correlation in economics, substantial progress has been made. One example is understanding the effects of government spending using wars. Many wars began for reasons unrelated to the performance of the domestic economy, but military spending is nevertheless often found to have substantial effects on subsequent growth and inflation. Additionally, there is now extensive micro-level cross-sectional evidence on how individuals respond to shocks and policies like fiscal stimulus payments. Both types of evidence can be used to decide between competing economic models. A model which is consistent with the empirical evidence at both the micro and macro level is a much more reliable tool for understanding how an economic shock feeds through the economy (for more discussion, see for example Nakamura and Steinsson (2018)).

An example of this is evidence on the marginal propensity to consume (MPC). There is robust micro evidence which shows that households whose easily accessible wealth is low tend to have high MPCs; if they received extra income or a government transfer, they would spend most or all of it fairly quickly (see for instance Ganong et al (2020) and Fagereng et al (forthcoming)). A large portion of the population (around 30% in the UK and US) behave like this, either because they have low incomes or because their wealth is in assets like housing which are difficult to sell quickly (Kaplan and Violante, 2014). The behaviour of these high MPC households is an important determinant of overall responses to economic shocks (Cloyne et al, 2020). HANK models use and match this micro and macro evidence, which makes them credible tools for understanding vital policy questions, including the transmission of monetary and fiscal policy.

How can models help us understand the Covid-19 crisis?

The coronavirus pandemic has caused an unprecedented economic crisis, and one with highly unequal impacts on different households. Many of those who have been hardest hit by the pandemic are those that have low liquid savings and are therefore financially vulnerable. Lower income households were more likely to work in sectors of the economy, such as restaurants and hotels, that were most affected by the pandemic and subsequent lockdowns. High frequency payroll and transaction data have shown that these poorer households saw a much larger fall in wages (see Hacioglu et al (2021) for the UK, Cajner et al (2020), Chetty et al (2020) and Cox et al (2020) for the US). As they also had limited savings, they were particularly exposed, though government benefits partly mitigated the hardship. Richer households were more likely to be able to work from home, saw less of a fall in income, and ended up saving much more as they reduced consumption.

Heterogenous agent models can help us understand the implications of these inequalities—the challenge is to bring economics and epidemiology together. This can be done by combining an epidemiological model of virus transmission into a heterogenous agent model with multiple economic sectors (Kaplan et al, 2020). This helps us understand why the impact of the pandemic is so unequal and to evaluate possible policy responses. Figure 1 below illustrates the effects of different policy options and the associated trade-offs between lives and livelihoods and who bears these costs for the US. The red line shows the trade-offs between lives lost and economic costs of lockdowns, as the lockdown length changes. The blue line shows the lower costs when fiscal support is used. The bands around these lines show the inequality in economic costs across the distribution of households, demonstrating just how much heterogeneity and inequality there is in who bears the economic costs of the pandemic and how this varies with different policy options. The model shows that the middle class are hit hardest; the poorest receive more government transfers, which help insulate them from the shock, while the rich are less affected by the pandemic.

Figure 1: Pandemic Possibility Frontier

Source: Kaplan et al (2020)
Note: Laissezfaire: no lockdown and no fiscal support. U.S. lockdown: lockdown without fiscal stimulus. U.S. policy: lockdown plus fiscal stimulus. Each point corresponds to lockdowns of different durations.

What’s next for macroeconomics?

Although there has been progress in understanding the implications of inequality, only fairly specific examples have been explored in the models mentioned in this article. These models primarily focus on inequalities of wealth and income, but there are many other dimensions of inequality that may matter for the macroeconomy. Similarly, these models typically focus on economic shocks over a timescale of months or years, rather than long-term shifts. Popular discontent with inequality often focuses, instead, on longer term changes in wealth inequality, racial inequality, geographic inequalities, inequalities of opportunity and intergenerational inequalities. Many of these issues have important macroeconomic implications, and are areas of active research: for instance, Mian et al (2021) explore how long-term rises in inequality may contribute to low interest rates, Hsieh et al (2019) examine how increases in gender and racial equality across occupations has increased long-run growth, and Auclert et al (2020) explore the implications of demographic shifts on inequalities and imbalances.

The pandemic highlighted the fact that challenges for the macroeconomy may - increasingly - come from new sources. Climate change is a clear risk which is becoming a key focus - both in terms of how the economy will be affected and how to help the economy adapt. Asset price fluctuations are a perennial source of risk, and remain crucial drivers of economic cycles and wealth inequality. Despite progress since the financial crisis, understanding the causes and consequences of booms and busts in asset prices is still a focus for economic researchers.

Current research is also exploring how insights from behavioural economics affect our understanding of the economy and policy. The heterogeneous agent approach is well-suited for this endeavour because it aims to build models “from the ground up”, taking seriously what we know about household behaviour at the micro level. A simplifying assumption commonly used in models is that households use all the information available about the economy and then make the best possible decisions about how to spend and save. Evidence from behavioural economics and household finance suggests that this is often not the case. Even when making important decisions about how much to save for retirement or when to re-mortgage, households often make decisions which are hard to rationalise with standard models. These different household decision making processes can potentially have large implications for the macroeconomy. Both the recent advances set out in this article and the changes new research will surely bring show that models, like the economies we seek to understand, are in a constant state of evolution.

Where can I find out more?

Who are experts on this question?

Author: Ben Moll, Natalie Rickard
Photo by Dan Burton on Unsplash
Recent Questions
View all articles
Do you have a question surrounding any of these topics? Or are you an economist and have an answer?
Ask a Question
Submit Evidence