The way we were: Price-level shocks and consumers’ memories

Key takeaways

• The affordability crisis isn't just about inflation, it’s also about the price level — what things cost — and the fact that, especially when prices suddenly spike upwards, consumers become lastingly annoyed.
• The price-level shock helps explain lingering consumer pessimism that is out-of-step with the macroeconomic indicators — a puzzle previously identified as the “vibecession.”
• Consumers’ memory for prices following a sudden shock is particularly salient when juxtaposed with what has happened to wages and incomes: wages and incomes used to grow much faster than the prices of things like shelter, groceries, and electricity. In recent years, these prices have grown as fast or faster than earnings.
• With the understanding of affordability advanced in this brief, policymakers should not try to lower overall price levels, but instead target the affordability of some of the most relevant goods and services to working families, like housing, childcare, and healthcare.

Affordability concerns continue to pose a challenge to the economic lives of American families. Recent polling shows that most people are worried about the cost-of-living, being able to make ends meet, and inflation (which we distinguish from affordability in important ways articulated below). Such concerns are often most pronounced around the lack of affordable housing, childcare, healthcare, utilities, and groceries, though this is by no means a complete list.

Though numerous authors have tried to bring more rigor to the definition of affordability, the concept remains somewhat vague and elastic. Some economists have argued that if the goal of an affordability policy agenda is to lower the broad, economy-wide price level, then the most reliable way to get there is a deep recession. Bernstein (2025) pointed out that this argument misreads ongoing policy work to craft an affordability agenda, which instead is focused on ameliorating costs in a few key areas where markets are flawed in ways that policy can help correct.

Other analysts have correctly pointed out that it makes little sense to focus solely on the price side of affordability. Affordability is a function of both prices and incomes, and a comprehensive agenda must consider both.

In this brief, we attempt to quantify various aspects of affordability as we understand it. We begin by trying to quantitatively understand what the concept means to people. This leads us to establish an important distinction between inflation and price levels. Coming out of the pandemic-induced price-spike in 2021-22, consumers appeared to focus much more on the historically sharp and fast increase in price levels than on the ups and downs of the inflation rate. For example, we show that we cannot track negative consumer sentiment with the traditional “misery index” (the sum of the unemployment and year-over-year inflation rates), but we can do so if we incorporate a longer memory of inflation (using four-year inflation rather than the more often-used annual measure). We also show that adding a price-level-spike variable to our “affordability gap” (i.e., the historically large gap between actual and predicted economic sentiment) regressions helps explain sentiment in a way year-over-year inflation alone cannot. We find that this framework does not apply to the 1970s’ price spikes, suggesting that the 2022 price spike was uniquely challenging relative to historical antecedents.

To bring in the income side of affordability, we conclude by analyzing the trajectory of various nominal income and wage series relative to not just the overall Consumer Price Index (CPI), but price sub-indices directly related to the big-ticket items on the affordability list. We find that for many of the goods and services we analyze, price growth relative to pre-pandemic trends have outpaced wage and income growth. We apply the same price spike framework to these price sub-indices, wages, and incomes to isolate the “surprise” component of post-COVID shocks on either side of the affordability equation. We find that while there was a positive post-COVID shock to wages, this shock was smaller than the post-COVID shocks to prices for numerous salient and high-frequency goods and services like groceries, electricity, and rent.

Measuring the affordability (vibes) gap

Various analysts, including Wilcox (2021) and Cummings et al (2024), have shown that a standard set of off-the-shelf macroeconomic variables (like unemployment and inflation) have historically done a good job of predicting prominent economic sentiment and confidence indices. Figure 1 shows the University of Michigan Index of Consumer Sentiment (from their Surveys of Consumers), along with the predicted values – based on an OLS regression of sentiment on the year-over-year inflation rate, unemployment rate, S&P500 quarterly returns, and monthly changes in real aggregate consumer spending. We estimate the model on data through the end of 2019, right before the COVID pandemic, and then make out-of-sample predictions of the consumer sentiment given the four variables’ post-COVID values. Whereas our model performs well for the 30 years spanning 1990 to 2019, starting around 2021-22 a large gap appears with predicted values well above actual values. This means people feel worse about the economy than we would expect given observed macroeconomic conditions. This affordability gap, which we also call a vibes gap, still appears only after 2020 if we cut off the training window in 2015; in this specification, the model performs well out-of-sample for 2015-19 before a large gap opens up post-2020.[1]

Figure 1. The affordability gap, measured using the University of Michigan Index of Consumer Sentiment

This gap is robust to measurement of sentiment: Appendix Figure A-1 documents the same regression’s results when replacing the consumer sentiment index with the Conference Board’s Consumer Confidence Index (CCI), another prominent measure of economic mood. While the gap is smaller (according to the Conference Board’s measure, consumers are less pessimistic than according to the University of Michigan’s index), it is there, nonetheless.[2]

These graphs reveal that the price shock of the pandemic and ensuing recovery led to a large and persistent residual between the actual and predicted sentiment. This gap’s appearance has coincided with the affordability crisis and elevated interest in affordability as a political and economic issue, exemplified by political campaigns like Zohran Mamdani’s mayoral campaign in New York City, and Abigail Spanberger and Mikie Sherrill’s gubernatorial campaigns in Virginia and New Jersey, respectively. Figure 2 plots web search trends for “affordability” against the affordability gap as measured in our model (using the University of Michigan Index of Consumer Sentiment); we can see that the two, each separately standardized to 100 in January 2020, have moved together post-2022.

Figure 2. Web search trends for “affordability” and measured affordability gap (Jan 2020 = 100)

It is widely argued that higher inflation explains this residual gap; however, yearly inflation is a regressor in the model, so any effects of inflation on sentiment would already be accounted for in the model. Instead, our commonly-used inflation measures—month-over-month or year-over-year percent changes—lack a memory of the price level.  We hypothesize that a memory-infused price variable is a key missing ingredient from our models of consumer sentiment, especially following a price shock.

To further elevate this point, consider the shortcoming of the traditional “misery index” (pioneered by economist Arthur Okun) in tracking consumer sentiment. This basic misery index is the sum of the unemployment and year-over-year inflation rates. As Figure 3 shows, this traditional index has done a poor job of tracking consumer sentiment after 2022, after more than a decade of moving together with sentiment. While the misery index is roughly at 2016-19 levels, sentiment is more pessimistic than at any point since the immediate aftermath of the Great Recession.

We propose a simple augmentation. In the case of a sharp price spike, we suspect that people’s memory of pre-spike prices persists, such that nearer term price changes, like the change over the past year, are less salient to sentiment formation.

We thus build a misery index with a longer memory, using the 4-year, rather than annual, percent change in prices. This version of the index, as Figure 3 shows, is more effective at tracking sentiment; in particular, the misery index with a memory has remained elevated as consumer pessimism has persisted.[3] While our alternative index tracks sentiment better than the original, it doesn’t match the directionality of sentiment: the misery index with memory rose in 2023-24 while sentiment concurrently improved, and since 2024 it has precipitously fallen while sentiment has worsened again. Clearly, our theory — that consumers have a long memory of prices — requires more careful operationalization than simply extending backwards the horizon of calculating inflation.

Figure 3. Traditional and memory-based misery indices and (inverse) consumer sentiment

For these reasons, while the MIM is conceptually closer to what we think is driving the affordability gap, a better approach would be to try to model the price-level shock relative to price-level expectations. We construct a price-spike term to infer deviations from price-level expectations. We start with the 1990–2019 trend in price level and, using the average growth rate of this trend from 2014-19, we linearly extrapolate it out to the present. Next, we find the log residual between this counterfactual trend line and the observed price level. This residual captures, in effect, the difference between where price levels are and where they would be had prices continued to grow at the rate that households had grown accustomed to over the post-Great Recession, pre-pandemic period. To illustrate how this variable has evolved over time, Figure 4 plots the extrapolated trend line and the observed price level over time.

Figure 4. The price-spike term: CPI vs. pre-pandemic trend

We incorporate this variable (the log difference between the actual and trend values) into the affordability gap regressions and plot those results in Figure 5. When including this trend-break/price-spike term, the gap entirely closes. In fact, sentiment is better than predicted in 2024-25 once we include the price-spike term. Importantly, the price-spike term does not change how the model is fit between 1990 and 2019; these results suggest that the price-level trend break of 2021-22 was an unprecedented shock in the modern era, and, at least as far as the University of Michigan Consumer Sentiment Index is concerned, households have still not fully updated their price-level expectations (which, to be clear, are not the same thing as the much more commonly cited “inflationary expectations”).[4]

Figure 5. The affordability gap, including the price-spike term

Applying price-spike methodology to the 1970s inflation

Applying our price-level expectations method to earlier inflationary spikes reveals interesting information about how the post-COVID shock differed from earlier shocks. After relatively low inflation characterized the early 1960s, inflation remained persistently high from 1965-82 — a period known as the “Great Inflation.” This era featured two pronounced inflation spikes in 1973-74 and 1979-80, when inflation peaked at 12.2 percent and 14.6 percent, respectively. We focus on 1979-80 here.[5]

In Figure 6, we replicate the methodology used to estimate the affordability gap shown in Figure 5, training alternative models (one including the price-spike term shown in Appendix Figure A-2 and the other excluding that price-spike term) on data from 1959-79. To clarify, we build the price-spike variable by running a smooth trend through the price-level series, taking the average annual growth rate from the years 1973-78, and extrapolating the trend at that rate from 1979-84, the results of which are shown in Figure A-2. Again, the expanded model incorporates the log difference between trend and actual to the baseline model as an additional regressor.

Figure 6. Predicted vs. actual consumer sentiment: 1979 inflationary episode

The results shown in Figure 6 are quite different from those in the post-pandemic period. First, there never was a “vibes gap” in the late 70s or early 80s. Consumer sentiment formation did not fundamentally change as prices spiked: unlike in the present, the baseline model closely tracks actual sentiment out of sample. Second, a consumer sentiment model including the price-spike variable performed slightly but noticeably worse than the model without a price-spike variable.

While we want to be careful not to extrapolate too much from these few episodes, we suspect that the post-COVID price spike was uniquely shocking to consumers relative to historical antecedents. Note that the variance in yearly CPI inflation from 1960-89 was 7.7 percent, whereas the inflation rate’s variance since 1989 has only been 1.7 percent. In other words, in recent decades — in part due to the success of the Federal Reserve in anchoring inflationary expectations — consumers simply hadn’t experienced much in the way of price shocks or price volatility. On the other hand, consumers in the 1970s had experienced both the steady rise of inflation from the mid 1960s to 1973, as well as the price spike of 1973-74. Inflation spiked in November 1974, reaching 11.5 percent; consider that someone 18 years of age at the time of that spike would only have been 23 in 1979. Someone 18 years old in 1979 would have been 61 in 2022. Nobody under the age of 43 in 2022 would have been alive during the last time inflation breached 7.5 percent, much less actively participating in the economy.

In other words, the results from this part of our research underscore that one cannot begin to understand how people experience price changes by simply looking at the ups and downs of yearly inflation, especially around a shock. It is also essential to account for people’s memories and expectations, not solely with respect to the monthly or yearly change in prices, but also regarding price levels. It is also important to account for their longer-term experiences of price movements. Our findings suggest that a huge storm after a long calm can be more upsetting to people who are not used to bad weather.

Affordability, wages, and incomes

Affordability is, of course, not a function of only prices, but rather of prices relative to incomes. In this section, we examine the recent history of the buying power of wages and incomes, focusing on wage growth relative to price growth of the goods and services most salient to the affordability agenda and people’s everyday interactions with the economy. We focus on hourly wages and post-transfer pre-tax household incomes, with the latter more representative of a household’s purchasing power for a given year as it incorporates changes in household hours worked in the paid labor market.

First, Figure 7 below compares nominal hourly earnings of middle/low-wage workers from the monthly Bureau of Labor Statistics survey of establishments with price changes for groceries, electricity, health insurance premiums, housing prices, and rents — categories of goods whose prices everyone in the economy is regularly exposed to (the overall CPI is also in the figure).[6] All prices are from the Consumer Price Index, except for insurance premiums — which are measured by Kaiser Family Foundation’s (KFF) Employer Health Benefits Survey [7] — and housing prices — which are measured by the Case-Shiller Home Price Index. Each bar captures cumulative percentage growth over a four-year period; the figure examines two periods, pre- and post-pandemic, with the latter corresponding to the affordability crisis.

Figure 7. Cumulative changes in nominal hourly wages and prices of select goods

In the pre-pandemic period, hourly wage growth significantly outpaced the growth of grocery and electricity prices, while lagging in insurance and housing. Grocery inflation, in particular, was effectively zero over these years. The second set of bars tells a very different story. While mid-level nominal wages did accelerate relative to their pre-pandemic growth, grocery and electricity prices accelerated even faster. Insurance premiums maintained roughly the same growth rate, while the housing market ran extremely hot, both in terms of rent and home prices. While it’s true that wage gains slightly outpaced grocery costs over these years, they were barely doing so in the recent period compared to the very large gap between wage and price growth in the earlier period. Electricity prices went from significantly lagging behind wages to outpacing them, exhibiting nearly double the growth we see in wages. It’s clear that averaging out the basket of goods, which is what the all-items CPI inflation rate does, masks heterogeneities in how purchasing power has evolved for consumers with respect to specific, very salient goods.

These two very different wage/price regimes explain a lot about the perceived affordability problem, one that some commentators miss because they discount this context. A recent Washington Post editorial, for example, argued that because grocery prices and wages were now growing at about the same rate, there’s no reason to consider grocery prices as part of an affordability agenda. But this overlooks the memory principles we focus on throughout this piece. People know that their paychecks used to go a lot further at the supermarket, and if you fail to grasp the salience of the flip in the wage/price dynamics shown in Figure 7, you will fail to grasp why a lot of people are still unhappy about the current level of grocery (and electricity) prices.

While the hourly wage is a foundational variable in the affordability context, it also makes sense to look at family income, which is a function of the wage in tandem with household labor supply. For example, if your hourly wage grows but you cannot find the hours of work you seek, those conflicting dynamics show up as lower family income. We use the Census Bureau’s household income data, which allows us to incorporate into the analysis the income distribution. Figure 8 shows average nominal income growth for the bottom fifth, the middle fifth, and the top 5 percent of households, compared to the same prices shown in the previous figure. The figure illustrates a subtly different trend from Figure 7: whereas nominal wage growth has accelerated post-pandemic, nominal income growth was comparable in both periods, though of course inflation grew much faster in the latter period. Post-pandemic incomes have largely only held steady against overall inflation, groceries, and rent —while lagging electricity and housing prices.

Figure 8. Cumulative changes in nominal household incomes and prices of select goods

The preceding two figures underscore the point that while affordability, at least in popular discourse, is most frequently discussed as a function of higher prices, the decline in real wages and incomes in the post- relative to the pre-pandemic period was also very much in play. In the wage case, nominal hourly pay accelerated for low/mid-wage workers, but at a considerably slower pace than most of the prices of goods and services that typically populate the affordability agenda. In the case of incomes, nominal growth was relatively constant, and therefore real growth significantly slowed.

Note that the cost growth of health insurance premiums was flat between the two periods. In our broad framework, this price should have been experienced as more affordable, post-pandemic. The fact that this is not the case probably reflects the fact that households have long viewed health premiums as a cost burden, well before the affordability debate arose.

A good way to analyze these income-price dynamics is to return to the method we used to measure the price-spike variable in the economic sentiment regression above. In Figure 9, we calculate a smooth trend and extrapolate forward the average 2014-19 growth trend through to the present. We then plot the log difference between the actual value and the underlying trend, a quantity we think of as the “surprise” component (deviations from the trend to which consumers had grown accustomed) of changes in earnings and prices. When this measure is near zero, it means that the price or income level is near the pre-pandemic trend. This approach allows us to infer price-level and earnings or income expectations.

Figure 9 reveals that nominal income growth at the bottom, middle, and top of the distribution grew roughly at trend from 2020-24. Yet, post-COVID, some of the major price categories in the affordability bucket experienced large shocks relative to consumers’ expectations based on the pre-pandemic trend, with these shocks particularly concentrated among electricity, groceries, and housing prices. These dynamics speak to what we see as a pressing question that further research on the affordability crisis needs to address — the temporal dynamics of price expectations, and how price and income expectations interact to help determine consumer sentiment.

Figure 9. The surprise component of post-COVID shocks to prices and incomes

Conclusion

Affordability is a broad, often ill-defined topic, though one that’s getting a great deal of policy attention. In that regard, a useful role policy-oriented economists can play is to add some rigor to the concept by trying to quantify the nature of the problem. We do so here by constructing an “affordability gap” measure and showing how that measure is affected by variables that we believe represent the underlying source of the crisis. For example, we find that what we call a price-spike variable is critical to explaining the post-pandemic affordability gap.

While much of the affordability agenda focuses on prices, wages and incomes are of course equally germane to the buying power of households. We examine wage and income growth along different points of the distribution. This analysis shows that, while in some cases, incomes kept pace with rising prices, wage/price dynamics were very different before the pandemic price shock. For example, wage growth far outpaced grocery prices before the pandemic price spike but barely outpaced them afterwards. This more holistic look at incomes and prices further underscores the nature of the affordability crisis as the result of acute price shocks and consumer memory.

There are at least two strains of follow-up research suggested by our findings. First, we say nothing herein about what policies might help to ameliorate the affordability challenges faced by so many households today.[8] Second, throughout this analysis, we have asserted that the affordability crisis was formed as a reaction to the price shock that occurred in 2021-22. Our hypothesis is that the shock in the price level, versus the ups and downs of inflation, was instrumental in the evolution of the affordability focus. This assertion requires both theoretical and empirical backing, which we hope to develop in future work.

Appendix

Figure A-1. The affordability gap, measured using the Conference Board Consumer Confidence Index

Figure A-2. The price-spike term in the 1979 inflationary episode

References

[1] We have also stress-tested this model with rolling out of sample predictions; while there have been vibes gap in previous periods, most notably the Great Financial Crisis, in these historical episodes the gap closed relatively quickly as the model’s parameters updated. After COVID, the vibes gap has proven remarkably persistent even for iterations of the model trained on very recent months of data.

[2] There is some evidence that the University of Michigan Surveys of Consumers focuses more on prices relative the Conference Board survey which focuses more on jobs. This evidence is consistent with the smaller magnitude of the gap in the latter case.

[3] Note that in Figure 3 we take the inverse of the consumer sentiment; the higher the inverse sentiment is, the more pessimistic consumers are (the opposite direction interpretation from the unadjusted consumer sentiment index). This matches the directionality of the misery index, for which higher values reflect a worse economy.

[4] Interestingly, when using the Conference Board Consumer Confidence Index as the outcome variable, we find the prediction is now below the actual, meaning that by this index, Americans feel better than we’d expect about the economy, given the price shock.

[5] Our data begins in 1959, although the University of Michigan only began to report its Consumer Sentiment Index monthly in 1978 (it was reported quarterly beforehand). Focusing on 1979-80 as opposed to 1973-74 allows for a large increase (n = 86 vs. n = 54, a 59% increase) in our training sample.

[6] For hourly wages, we use the Bureau of Labor Statistic (BLS) series capturing the hourly wages of production, non-supervisory workers.

[7] Specifically, we focus on the average employee contribution to health care coverage for a family of four.

[8] We have done so elsewhere, including a SIEPR policy brief by Bernstein and Mahoney (2025), as well as Center for American Progress policy briefs on housing and groceries.