We use cookies on our website to give you the best online experience. Please know that if you continue to browse on our site, you agree to this use. You can always block or disable cookies using your browser settings. To find out more, please review our privacy policy.

COVID-19 RESOURCES AND INFORMATION: See the Atlanta Fed's list of publications, information, and resources; listen to our Pandemic Response webinar series.

About


Policy Hub: Macroblog provides concise commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues for a broad audience.

Authors for Policy Hub: Macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

Comment Standards:
Comments are moderated and will not appear until the moderator has approved them.

Please submit appropriate comments. Inappropriate comments include content that is abusive, harassing, or threatening; obscene, vulgar, or profane; an attack of a personal nature; or overtly political.

In addition, no off-topic remarks or spam is permitted.

November 12, 2020

A Dashboard Approach to Monitoring Underlying Inflation

Editor's note: In December, macroblog will become part of the Atlanta Fed's Policy Hub publication.

Measuring progress toward the Federal Open Market Committee's (FOMC) dual mandates of maximum sustainable employment and price stability is often reduced to shorthand: Simply, monitoring the level of the unemployment rate relative to its longer-run trend, and tracking the level of a specific measure of underlying inflation—the so-called "core" measure of personal consumption expenditure (PCE) inflation (which excludes food and energy prices)—relative to the FOMC's 2 percent price stability goal. This ex food and energy "core" measure is often taken as a de facto proxy for the trend in overall (or "headline") inflation.  While the unemployment rate and core PCE inflation are, indeed, useful in measuring the FOMC's progress toward its objectives, they are not perfect.

In an influential speech in 2013, Janet Yellen (then chair of the FOMC) argued for a broader, more inclusive approach to monitoring the health of the labor market. The unemployment rate, she pointed out, had significant shortcomings—namely, when unemployed workers became discouraged and stopped looking for work, the unemployment rate would decline. Hence, monitoring a basket or dashboard of indicators (such as payroll employment, data on gross job flows, and quits rates) could help paint a more accurate picture of the health of the labor market.

And, much like former Chair Yellen highlighted the need for a dashboard approach to monitoring the employment mandate, simply using core PCE inflation to track the underlying inflation trend is insufficient. An analogous dashboard approach is needed to monitor progress toward the FOMC's price stability mandate.

For some obvious reasons, movements in the aforementioned core PCE price measure do not always reflect changes in inflation. Explaining why gets a bit academic, but embedded in every price change are, at least, two components. The first is a real component, reflecting changes in the supply and demand for a particular good or service relative to others in the consumers' market basket. The second component is a nominal component, reflecting the supply of money (or the stance of monetary policy) relative to what's needed to facilitate purchases of goods and services in the economy during a given time period. It is that second component—the inflation component—that we are attempting to uncover. Efforts to do this rely on discerning measures of underlying inflation—measures that attempt to remove transitory effects and noise from the price data.

Implicit in using an underlying inflation measure that excludes only food and energy prices is the assumption that every other price change is a reflection of a change in underlying inflation. However, that assumption is off base. Large relative price changes outside of food and energy items, having nothing to do with the FOMC's price stability mandate, often occur. In the abstract, these can be any relative price change, such as a sharp increase in excise taxes, subsidies on prices in particular markets, or temporary supply chain disruptions resulting from natural disasters, pandemics, or a disruption in global trade.

Specifically, over the past six years or so, we've seen at least three of these large and salient relative price changes that have materially affected core PCE inflation. The first example is changes in administered health care prices that lead to a sharp slowing in the price index for hospital services (and thus health care prices in general) shortly after the passage of the Affordable Care Act (often referred to as Obamacare). The second is a methodological change that the U.S. Bureau of Labor Statistics enacted in January 2017, which made cell phone service prices more sensitive to quality changes. In March 2017, a few large national carriers switched to offering largely unlimited data packages, yielding a 50 percent (annualized) decline in this component and having a striking impact on year-over-year core PCE inflation through March 2018. The third is a series of huge price swings in imputed financial services prices in late 2018 and early 2019.

These are just a few of the most salient relative price changes that have altered a "core PCE-centric" view of inflation over the prior expansion, but there have been many more. Thankfully—and in large part due to great work throughout the Fed system to understand and measure inflation—we have a variety of alternative inflation measures designed to limit the influence of these large, idiosyncratic price changes. A few of the better-known ones are the Cleveland Fed's median and 16 percent trimmed-mean CPI and the Dallas Fed's trimmed-mean PCE measure. These measures remove the influence of sharp component price swings on a monthly basis and, as a result, tend to have a lower variance than the usual "core" measures, leading to superior forecasting performance over most time horizons and in a variety of inflation forecasting models.

Another set of inflation indicators reweights, or classifies, detailed components in the consumers' market basket into different groups based on characteristics that a monetary authority (such as the Federal Reserve) should be interested in. First, let's consider indicators such as the San Francisco Fed's Cyclical Core PCE Inflation index and Stock and Watson's Cyclically Sensitive Inflation Index. These measures either exclude or deemphasize the weight of prices that do not move in tandem with the business cycle. The argument goes that certain components of the consumer's market basket (such as health care and education prices) follow strong, idiosyncratic trends, and therefore price changes among these components are more likely to reflect relative price changes rather than the influence of monetary policy working through the pricing mechanism.

A third type of inflation indicator is the Atlanta Fed's Sticky Price CPI. This measure tracks a set of CPI components that are slow to react to changing economic conditions (and hence are "stickier" than other components) and appear to incorporate expectations about future inflation to a greater degree than prices that change on a frequent basis. This measure tends to forecast inflation over longer time horizons more accurately than headline or core inflation measures.

We have pulled together these various measures into the Underlying Inflation Dashboard, which allows users to see a more complete picture of underlying inflation. A quick overview of this dashboard is in order (see chart 1). The first section of the table shows the 12-month growth rate of each underlying inflation measure. It compares the most current data to the value of a measure from one year prior. Each measure is color-coded, in 25 basis point (bp) increments, relative to its price stability target. Admittedly, the choice for each measure's price stability target is somewhat arbitrary, but given the primacy of core PCE inflation in the communications of the FOMC and the Federal Reserve, we chose to express each measure's target as 2 percent plus its average difference with core PCE inflation over the past decade. For example, the growth rate in the core CPI over the past 10 years is 30 bp higher than that of the core PCE, yielding a price stability target of 2.3 percent.

Chart 1: A Broader Approach to Tracking Inflation

What's interesting to see is the narrative that emerges by taking the dashboard approach to monitoring underlying inflation.

Prior to the onset of COVID-19 earlier this year, the consensus view surrounding inflation was one of persistent shortfall, even after (at least) achieving maximum employment by most measures. Indeed, at least through the beginning of 2020, core PCE inflation continued to trend below target (see chart 2).

Chart 2: Core PCE Inflation Remained below Target throughout the Last Expansion

However, every other measure of underlying inflation in the dashboard had converged to a growth rate consistent with the Fed's price stability mandate by early in 2018 and stayed there up until the onset of the pandemic (see charts 1, 3, and 4). Taking a dashboard approach leads us to the conclusion that, while it took some time following the Great Recession, inflation had converged to our price stability target and remained on target until March 2020.

Chart 3: Trimmed-mean PCE Was on Target

Chart 4: Cyclically Sensitive Inflation Was on Target Too

The previous discussion made the case for using a dashboard approach when evaluating underlying inflation prior to the onset of the COVID-19 pandemic. Since then, a series of dramatic relative price swings, along with direct complications in the physical measurement of prices, have further complicated the measurement of underlying inflation.

One particularly salient example comes from used auto prices. Auto prices have surged, rising at a record annualized rate of 75 percent from July to September. This spike is likely the result of a combination of increased demand stemming from commuters attempting to avoid mass-transit options, less confidence over future incomes, and a temporarily reduced supply of new vehicles. This relative price change alone has pushed the 12-month growth rate in core PCE goods prices up by nearly a full percentage point and added nearly three tenths of a percent to the 12-month trend in core PCE inflation. In contrast, trimmed-mean estimators (such as the Cleveland Fed's 16 percent trimmed-mean CPI and the Dallas Fed's trimmed-mean PCE) have largely ignored (hence the "trimmed") the influence of this rather dramatic price swing over the prior three months—leading to much more stable month-to-month estimates of underlying inflation.

It is precisely these types of dramatic relative price swings that argue for the broader approach we've sought to provide with the Underlying Inflation Dashboard. We hope you'll give it a whirl and let us know what you think.

 

October 7, 2020

Two Quite Different Paths for U.S. Unemployment

Editor's note: In December, macroblog will become part of the Atlanta Fed's Policy Hub publication.

Here are two charts that I think are very telling for the recovery of the U.S. labor market. Chart 1 shows the unemployment rate for people who reported being temporarily laid off from their job and anticipate being recalled. Chart 2 shows the unemployment rate for those who reported being laid off permanently, with no prospect of being recalled. They are on very different trajectories.

Chart 1: Temporary Layoff Unemployment Rate
Chart 2: Permanent Layoff Unemployment Rate

I've computed these rates as a share of the civilian labor force. Other reasons for unemployment include reentrants or new entrants to the labor force as well as those completing temporary jobs and are not shown.

The good news is that after increasing to a never-before-seen level in April, the temporary unemployment rate has improved markedly as many businesses have reopened and recalled their temporarily laid-off staff. The bad news is that as the pandemic has unfolded, an increasing number of unemployed workers are reporting being laid off permanently—and they account for a rising share of the labor force. Those on permanent layoff have a lower rate of reemployment in general than those on temporary layoff, and the flow into employment is currently similar to the low level seen in the wake of the Great Recession. Also troubling is the fact that the reemployment rate of those on temporary layoff is also lower than normal—meaning that for some, temporary is starting to look more permanent.

While the level of permanent layoffs is not close to that seen during the Great Recession, the increase indicates that a near-term return to prepandemic labor market conditions is unlikely. In fact, as last week's macroblog post pointed out, survey evidence suggests that many firms don't anticipate getting back to prepandemic employment levels for several years.

May 28, 2020

Firms Expect Working from Home to Triple

The coronavirus and efforts to mitigate its impact are having a transformative impact on many aspects of economic life, intensifying trends like shopping online rather than visiting brick-and-mortar stores and increasing the incidence of working from home. Indeed, many tech giants have already made working from home a permanent option for employees.

Working from home, or telecommuting, is not a new phenomenon. According to a survey by the U.S. Bureau of Labor Statistics (BLS), around 8 percent of all employees worked from home at least one day a week before the arrival of COVID-19. However, only 2.5 percent worked from home full-time in the 2017–18 survey period.

Working from home has surged in the wake of social distancing and other efforts to contain the virus, and this surge brings up a good question: How many jobs can be done at home? Some careful research by Jonathan Dingel and Brent Neiman indicates that nearly 40 percent of U.S. jobs can be done at home.

While this provides an upper bound, can does not mean will, so a natural follow-up question is: How many jobs willbe done at home? To get a sense of how many jobs and how many working days will beperformedat home after the pandemic recedes, we turn to our Survey of Business Uncertainty (SBU). To preview our conclusion, the share of working days spent at home is expected to triple after the COVID-19 crisis ends compared to before the pandemic hit, but with considerable variation across industries.

In the May SBU, we asked two questions to gauge how firms anticipate working from home to change. To get a pre-pandemic starting point, we asked panelists, "What percentage of your full-time employees worked from home in 2019?" And to gauge how that's likely to change after the crisis ends, we asked, "What percentage of your full-time employees will work from home after the coronavirus pandemic?" We asked firms to sort the fraction of their full-time workforce into four categories, ranging from those employees working from home five full days per week to those who rarely or never work from home.

Chart 1 summarizes firms' responses to these two questions. It also summarizes the responses by workers to questions about working from home in the BLS's 2017–18 American Time Use Survey. For the period preceding COVID-19, SBU results and the Time Use Survey results are remarkably similar. Both surveys say 90 percent of employees rarely or never worked from home, and a very small fraction worked from home five full days per week. As reported in the chart's rightmost column, about 5 to 6 percent of all working days happened at home before the pandemic hit.

Chart 1: Working From Home, Pre- and Post-COVID

According to the SBU results, the anticipated share of working days at home is set to triple after the pandemic ends—rising from 5.5 percent to 16.6 percent of all working days. Perhaps even more striking, firms anticipate that 10 percent of their full-time workforce will be working from home five days a week.

Overall, firms say that about 10 percent of their full-time employees worked from home at least one day a week in 2019. That fraction is expected to jump to nearly 30 percent after the crisis ends (well below the upper bound estimated by Dingel and Neiman). Chart 2 gives a look at firm's working-from-home expectations for major industry groups.

Chart 2: Working From Home at Least One Full Day Per Week, Pre- and Post-COVID, by Industry

The share of people working from home at least one day a week is expected to jump markedly in the construction, real estate, and mining and utilities sectors, presumably by granting front-office staff working-from-home status. It is also expected to jump markedly in health care, education, leisure and hospitality, and other services, possibly by relying more heavily on remote-delivery options (for example, online education and virtual doctor's visits). Firms in the business services sector anticipate that working from home will rise to nearly 45 percent.

For the industries we can match directly to American Time Use Survey statistics, the two data sources imply a similar incidence of working from home before COVID-19. For manufacturing, SBU data indicate that 9 percent of employees worked at home at least one day a week prior to COVID-19, and the American Time Use Survey indicates that 7.3 percent did so. For retail and wholesale trade, the corresponding figures are 4.1 percent and 4.0 percent, respectively.

To summarize, our survey indicates that, compared to before the pandemic, the share of working days spent at home by full-time workers will triple after the pandemic. Our results also say that this shift will happen across major industry sectors. These changes in the location of work are also likely to exert powerful effects on the future of cities and the demand for high-rise office space (more on that next month).

Regarding the long-run impact of the shift to working from home, there are grounds for optimism, including a potential boost to productivity—although if you're juggling kids at home and working from your couch or bedroom, we can understand if it's hard to imagine right now.

 

April 17, 2020

Businesses Are in Uncharted Waters

Inflation expectations in our April Business Inflation Expectations (BIE) survey fell to an all-time low (going back to October 2011) of 1.4 percent, plunging far below its next lowest level of 1.7 percent (most recently observed in February 2020). Perhaps unsurprisingly, firms have bigger worries on their minds. And our boss, President Raphael Bostic, agreed, noting on WednesdayOff-site link that "inflation at this point is not something I'm particularly worried about."

The drop in inflation expectations was not the only historical low that our survey results uncovered. Firms' assessments of current sales levels relative to what they consider "normal" levels fell precipitously. Recovering from the 2007–09 financial crisis and recession, this quantitative sales gap measure had slowly been moving toward zero (or "normal" sales levels) alongside solid gains in gross domestic product growth and previously strong job gains. However, that all changed in April. Our survey, which was in the field from April 6 to 10, showed an extraordinarily large decline in sales levels relative to normal—from 2.5 percent below normal in the first quarter to 32 percent below normal in April (see the charts). The decline in sales had an impact on firms of all sizes, but smaller firms reported a much larger hit to sales than did firms with more than 100 employees.

Our survey's special questions this month focused on the level of disruption the coronavirus outbreak was causing for southeastern firms. We asked participating firms to assess disruption to their business operations and sales activity, on a scale of "no disruption" to "severe disruption," and it's obvious that a majority of firms in our panel have experienced severe disruption to their sales activity (see chart 2). The table indicates how disrupted firms' operations and sales were. Among those firms experiencing severe disruption, current sales levels have been roughly halved relative to normal conditions. The results suggest that the disruption associated with the outbreak has not hit all firms equally. There is also some evidence of dispersion (reallocation) across firms, as a small share of firms that indicated they are experiencing low levels of disruption are seeing stronger-than-usual sales levels.

As firms continue to grapple with the unprecedented impact and uncertainty that the coronavirus outbreak has inflicted, we wanted to get a rough sense of how long they expect these unusual conditions to persist and how long they can weather the economic shutdown without seeking new sources of funding. The left-hand graph in chart 3 shows the cumulative distribution function (CDF) for how many months before business operations return to normal. The CDF on the right-hand side plots how long firms can continue to operate in the current environment without seeking additional funding to backstop operations.

The typical (median) response was an expectation that it will be about four more months for business operations to return to normal (though the tail is long, and about 10 percent thought a year or longer is in order). Perhaps the silver lining here is that the typical response was an expectation to be able to operate for another six months before needing to tap additional sources of funding. Assuming that much of the economic activity that has been shuttered begins to resume by the beginning of the fourth quarter and conditions do not deteriorate further, the "typical" firm in our panel should be able to continue to operate.

However, digging into the individual responses reveals some nuance in this relationship. The cross-sectional relationship between a business decision-maker's assessments of the length of time he or she can continue to operate without securing additional funding and the length of time before resuming normal activity carries a correlation coefficient of just 0.2. (This finding essentially means that survey respondents often had different notions of when they would be able to resume normal business operations and need to tap additional funding.) The typical firm expects to be able to resume normal operations about a week or so before they need to tap additional funding. And, perhaps more importantly, nearly 40 percent of firms in our sample expect they'll need to secure additional funding before their operations return to normal.

Finally, although inflation isn't the first thing on everyone's mind at the moment, we did ask firms about their price expectations (see chart 4). While roughly 60 percent expect to hold steady on prices over the next six months, roughly a quarter of the panel expect to lower prices, and just 15 percent expect to increase them. On average, firms expect to lower prices by 2.2 percent, and there appears to be a relationship between COVID-related disruption to sales activity and expected price declines.

Across many dimensions, the disruption caused by the current pandemic is without precedent. Many firms headquartered in the Southeast have indicated severely disrupted business operations and sales activity, disruptions that appear to have caused incredibly sharp declines in sales levels. The typical firm in the panel expects this disruption to persist at least through the summer months (which may foreshadow the likely shape of the recoveryOff-site link). And—though not a primary concern at the moment—inflation expectations are the lowest we've recorded in more than 100 consecutive months of conducting this survey. In many ways, we appear to be in uncharted waters.