The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.
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.
January 17, 2018
What Businesses Said about Tax Reform
Many folks are wondering what impact the Tax Cuts and Jobs Act—which was introduced in the House on November 2, 2017, and signed into law a few days before Christmas—will have on the U.S. economy. Well, in a recent speech, Atlanta Fed president Raphael Bostic had this to say: "I'm marking in a positive, but modest, boost to my near-term GDP [gross domestic product] growth profile for the coming year."
Why the measured approach? That might be our fault. As part of President Bostic's research team, we've been curious about the potential impact of this legislation for a while now, especially on how firms were responding to expected policy changes. Back in November 2016 (the week of the election, actually), we started asking firms in our Sixth District Business Inflation Expectations (BIE) survey how optimistic they were (on a 0–100 scale) about the prospects for the U.S. economy and their own firm's financial prospects. We've repeated this special question in three subsequent surveys. For a cleaner, apples-to-apples approach, the charts below show only the results for firms that responded in each survey (though the overall picture is very similar).
As the charts show, firms have become more optimistic about the prospects for the U.S. economy since November 2016, but not since February 2017, and we didn't detect much of a difference in December 2017, after the details of the tax plan became clearer. But optimism is a vague concept and may not necessarily translate into actions that firms could take that would boost overall GDP—namely, increasing capital investment and hiring.
In November, we had two surveys in the field—our BIE survey (undertaken at the beginning of the month) and a national survey conducted jointly by the Atlanta Fed, Nick Bloom of Stanford University, and Steven Davis of the University of Chicago. (That survey was in the field November 13–24.) In both of these surveys, we asked firms how the pending legislation would affect their capital expenditure plans for 2018. In the BIE survey, we also asked how tax reform would affect hiring plans.
The upshot? The typical firm isn't planning on a whole lot of additional capital spending or hiring.
In our national survey, roughly two-thirds of respondents indicated that the tax reform hasn't enticed them into changing their investment plans for 2018, as the following chart shows.
The chart below also makes apparent that small firms (fewer than 100 employees) are more likely to significantly ramp up capital investment in 2018 than midsize and larger firms.
For our regional BIE survey, the capital investment results were similar (you can see them here). And as for hiring, the typical firm doesn't appear to be changing its plans. Interestingly, here too, smaller firms were more likely to say they'd ramp up hiring. Among larger firms (more than 100 employees), nearly 70 percent indicated that they'd leave their hiring plans unchanged.
One interpretation of these survey results is that the potential for a sharp acceleration in GDP growth is limited. And that's also how President Bostic described things in his January 8 speech: "For now, I am treating a more substantial breakout of tax-reform-related growth as an upside risk to my outlook."
September 7, 2017
What Is the "Right" Policy Rate?
What is the right monetary policy rate? The Cleveland Fed, via Michael Derby in the Wall Street Journal, provides one answer—or rather, one set of answers:
The various flavors of monetary policy rules now out there offer formulas that suggest an ideal setting for policy based on economic variables. The best known of these is the Taylor Rule, named for Stanford University's John Taylor, its author. Economists have produced numerous variations on the Taylor Rule that don't always offer a similar story...
There is no agreement in the research literature on a single "best" rule, and different rules can sometimes generate very different values for the federal funds rate, both for the present and for the future, the Cleveland Fed said. Looking across multiple economic forecasts helps to capture some of the uncertainty surrounding the economic outlook and, by extension, monetary policy prospects.
Agreed, and this is the philosophy behind both the Cleveland Fed's calculations based on Seven Simple Monetary Policy Rules and our own Taylor Rule Utility. These two tools complement one another nicely: Cleveland's version emphasizes forecasts for the federal funds rate over different rules and Atlanta's utility focuses on the current setting of the rate over a (different, but overlapping) set of rules for a variety of the key variables that appear in the Taylor Rule (namely, the resource gap, the inflation gap, and the "neutral" policy rate). We update the Taylor Rule Utility twice a month after Consumer Price Index and Personal Income and Outlays reports and use a variety of survey- and model-based nowcasts to fill in yet-to-be released source data for the latest quarter.
We're introducing an enhancement to our Taylor Rule utility page, a "heatmap" that allows the construction of a color-coded view of Taylor Rule prescriptions (relative to a selected benchmark) for five different measures of the resource gap and five different measures of the neutral policy rate. We find the heatmap is a useful way to quickly compare the actual fed funds rate with current prescriptions for the rate from a relatively large number of rules.
In constructing the heatmap, users have options on measuring the inflation gap and setting the value of the "smoothing parameter" in the policy rule, as well establishing the weight placed on the resource gap and the benchmark against which the policy rule is compared. (The inflation gap is the difference between actual inflation and the Federal Open Market Committee's 2 percent longer-term objective. The smoothing parameter is the degree to which the rule is inertial, meaning that it puts weight on maintaining the fed funds rate at its previous value.)
For example, assume we (a) measure inflation using the four-quarter change in the core personal consumption expenditures price index; (b) put a weight of 1 on the resource gap (that is, specify the rule so that a percentage point change in the resource gap implies a 1 percentage point change in the rule's prescribed rate); and (c) specify that the policy rule is not inertial (that is, it places no weight on last period's policy rate). Below is the heatmap corresponding to this policy rule specification, comparing the rules prescription to the current midpoint of the fed funds rate target range:
We should note that all of the terms in the heatmap are described in detail in the "Overview of Data" and "Detailed Description of Data" tabs on the Taylor Rule Utility page. In short, U-3 (the standard unemployment rate) and U-6 are measures of labor underutilization defined here. We introduced ZPOP, the utilization-to-population ratio, in this macroblog post. "Emp-Pop" is the employment-population ratio. The natural (real) interest rate is denoted by r*. The abbreviations for the last three row labels denote estimates of r* from Kathryn Holston, Thomas Laubach, and John C. Williams, Thomas Laubach and John C. Williams, and Thomas Lubik and Christian Matthes.
The color coding (described on the webpage) should be somewhat intuitive. Shades of red mean the midpoint of the current policy rate range is at least 25 basis points above the rule prescription, shades of green mean that the midpoint is more than 25 basis points below the prescription, and shades of white mean the midpoint is within 25 basis points of the rule.
The heatmap above has "variations on the Taylor Rule that don't always offer a similar story" because the colors range from a shade of red to shades of green. But certain themes do emerge. If, for example, you believe that the neutral real rate of interest is quite low (the Laubach-Williams and Lubik-Mathes estimates in the bottom two rows are −0.22 and −0.06) your belief about the magnitude of the resource gap would be critical to determining whether this particular rule suggests that the policy rate is already too high, has a bit more room to increase, or is just about right. On the other hand, if you are an adherent of the original Taylor Rule and its assumption that a long-run neutral rate of 2 percent (the top row of the chart) is the right way to think about policy, there isn't much ambiguity to the conclusion that the current rate is well below what the rule indicates.
"[D]ifferent rules can sometimes generate very different values for the federal funds rate, both for the present and for the future." Indeed.
July 11, 2017
Another Look at the Wage Growth Tracker's Cyclicality
Though Friday's employment report showed that payroll employment rose by a robust 222,000 jobs in June—much higher than most forecasts—enthusiasm for the news was tempered somewhat by average hourly wages coming in below expectations. Is the (ongoing) relatively tepid pace of wage growth a cause for concern? Perhaps, but the ups and downs of average wages over the course of the business cycle—the pattern of expansion-recession-expansion that typifies modern economies—are a bit more complicated than they may seem.
The year-over-the-year growth in the average wage level that we see in the official employment conditions report is influenced by wages paid to people who were employed either today or a year earlier. That is, the wages of those who remained employed (EE) as well as those who entered employment (NE) and those who exited employment (EN). Because the individuals in these groups may command different wages on average—due to experience, for example—the usual wage growth measures confound the effects of changes in the average wage of people with particular types of year-over-year employment histories. In that sense, the usual wage growth statistic may not exactly be comparing apples to apples.
Research by, for example, Solon, Barsky, and Parker 1992 and Daly and Hobjin 2016 explores the effect of the changing composition of workers over time using microdata on individuals with known employment histories. They show that people who enter and exit employment have a lower average wage than those who stay employed over the year and that the net exit/entry flow increases when the labor market is weak—more people leave employment, and fewer people enter it. As a result, the disproportionate increase in the net flow of workers with a lower-than-average wage serves to boost the overall average wage level during recessions.
One approach to making a more apples-to-apples comparison of average wages over time is to strip out the effect that comes from the change in the share of workers who stay employed and who entered or exited employment. Technically speaking, the composition-adjusted wage growth series is determined by adding the change in average log hourly wage within the EE group and the same change within the EN/NE group, while holding constant the respective average population shares in each group. The chart below illustrates the result of this adjustment.
I should note that the change in the average wage uses data only for people who have a known employment status a year earlier, which results in a wage growth series that is somewhat higher than the change in the average wage of all employed people, some of whom have an unknown employment history.
As the chart shows, relative to the adjusted series (the green line), growth in overall average wages (the orange line) stayed up longer during the last recession, then fell by less, and was slower to adjust to improving labor market conditions (falling unemployment) after the recession ended. The correlation between the overall growth in average wages and the inverse of the unemployment rate is 0.72, and this correlation rises to 0.79 using the adjusted wage growth series.
An alternative approach to making a more apples-to-apples comparison of average wages is to ignore the entry/exit margin and only look at people who are employed both today and a year earlier (EE). The Wage Growth Tracker (computed here as the difference in average log hourly wage) does that for the subset of EE people who have an actual wage record in both periods (no earnings information is collected for self-employed workers in the Current Population Survey). The following chart compares this version of the Wage Growth Tracker with the growth in overall average wages.
The Atlanta Fed's Wage Growth Tracker uses the median change in wages rather than the average change, but it displays very similar dynamics.
As the chart shows, the growth in average wages for those who remain in wage and salary jobs (the red line) is a bit smoother than growth in overall average wages (the orange line) and moves more in sync with the inverse of the unemployment rate (the correlation is 0.85). However, its level is quite a bit higher than growth in overall average wages. This disparity is because the average wage for those entering employment is less than for those exiting, so the change in average wages along the entry/exit margin is always negative.
But enough math—let's put this all together. If you want a measure of wage growth that reflects relative labor market strength, then looking at wage growth after controlling for entry/exit composition effects is probably a good idea. The Wage Growth Tracker seems to do that job reasonably well. However, the Wage Growth Tracker almost certainly overstates the growth in per hour wage costs that employers are facing. Most importantly, it ignores the employment exit/entry margin. Hence, one should avoid interpreting the Wage Growth Tracker as a direct measure of growth in labor costs—a point also discussed in this recent Atlanta Fed podcast episode . The next reading from the Wage Growth Tracker will be available when the Census Bureau releases the Current Population Survey microdata, usually within a couple of weeks of the national employment report. Given that the unemployment rate has remained relatively low recently, I would expect the Wage Growth Tracker to stay at a relatively high level. Check back here then and we'll see what we learn.
February 5, 2016
Introducing the Refined Labor Market Spider Chart
In January 2013, Atlanta Fed research director Dave Altig introduced the Atlanta Fed's labor market spider chart in a macroblog post.
In a follow-up post that June, Atlanta Fed colleague Melinda Pitts and I introduced a dedicated page for the spider chart located at the Center for Human Capital Studies (CHCS) webpage. It shows the distribution of 13 labor market indicators relative to their readings just before the 2007–09 recession (December 2007) and the trough of the labor market following that recession (December 2009). The substantial improvement in the labor market during the past three years is quite evident in the spider chart below.
As of December 2012, none of the indicators had yet reached their prerecession levels, and some had a long way to go. Now, many of these indicators are near their prerecession values—and some have blown by them.
To make the spider chart more relevant in an environment with considerably less labor market slack than three years ago, we are introducing a modified version, which you can see here. Below is an example of a chart I created using the menu-bars on the spider chart's web page:
In this chart, I plot the May 2004 and November 2015 percentile ranks of labor market indicators relative to their distributions since March 1994. As with the previous spider chart, indicators such as the unemployment rate, where larger values indicate more labor market slack, have been multiplied by –1. The innermost and outermost rings represent the minimum and maximum values of the variables from March 1994 to January 2016. The three dashed gray rings in between are the 25th, 50th, and 75th percentiles of the distributions. For example, the November 2015 value of 12-month average hourly earnings growth (2.26 percent) is the 23rd percentile of its distribution. This means that 23 percent of the other monthly observations on hourly earnings growth since March 1994 are lower than it is.
I chose May 2004 and November 2015 because they had the last employment situation reports before "liftoffs" of the federal funds rate. November 2015 appears to be stronger than May 2004 for some indicators (job openings, unemployment rate, and initial claims) and weaker for others (hires rate, work part-time for economic reasons, and the 12-month growth rate of the Employment Cost Index).
The average percentile ranks of the variables for these two months are similar, as the chart below depicts:
Also shown in the chart is the Kansas City Fed's Level of Activity Labor Market Conditions Indicator. It is a sum of 24 not equally weighted labor market indicators, standardized over the period from 1992 to the present. In spite of its methodological and source-data differences with the average percentile rank measure plotted above, it tracks quite closely, especially since 2004. However, as shown in the spider chart that I referred to above, there is quite a bit of variation within the indicators that may provide additional information to our analysis of the average trends.
We made a number of other changes to the spider chart to ensure it reflects current labor market issues. These changes are documented in the FAQs and "Indicators" sections of the new spider chart page. Of particular note, users can choose not only the years for which they wish to track information, but also the period of reference that provides the basis of the spider chart. The payroll employment variable is now the three-month average change rather than a level. Temporary help services employment has been dropped, and two measures of 12-month compensation growth and the employment-population ratio (EPOP) for "prime-age workers" (25 to 54 years) have been added.
Some care should be taken when comparing recent labor market data values with those 10 or more years ago as structural changes in the labor market might imply that a "normal" value today is different than a "normal" value in, say, 2004. The variable choices for the refined spider chart were made to mitigate this problem to some extent. For example, we use the prime-age EPOP as a crude adjustment for population aging, putting downward pressure on the labor force participation rate and EPOP over the past 10 years (roughly 2 percentage points). This doesn't entirely resolve the comparability issue since, within the prime-age population, the self-reporting rate of illness or disability as a reason for not wanting a job has increased about 1.5 percentage points since 1998 (see the macroblog posts here and here and the CHCS Labor Force Participation Dynamics webpage). If this increase in disability reporting is partly structural—and a Brookings study by Fed economist Stephanie Aaronson and others concludes it is—some of the decline in the prime-age EPOP since the late 1990s may not be a result of a weaker labor market per se.
Other variables in the spider chart may have had structural changes as well. For example, a study by San Francisco Fed economists Rob Valleta and Catherine van der List concludes that structural factors explain just under half of the rise in the share of workers employed part-time for economic reasons over the 2006 to 2013 period.
To partially account for structural changes in trends, we allow the user to select one of 11 time periods over which the distributions are calculated. The default period is March 1994 to present, which is what was used in the example above, but users can choose a window as short as five years where, presumably, structural changes are less important. A trade-off with using a short window is that a "normal" value may not produce a result close to the median. For example, the median unemployment rate is 5.6 percent since March 1994 and 7.3 percent since February 2011. The latter value is much farther away from the most recent estimates of the natural rate of unemployment from the Congressional Budget Office and the Survey of Professional Forecasters (both 5.0 percent).
In our June 2013 macroblog post introducing the spider chart, we wrote that we would reevaluate our tools and determine a more appropriate way to monitor the labor market when "the labor market has turned a corner into expansion." The new spider chart is our response to the stronger labor market. We hope users find the tool useful.
- Business Cycles
- Business Inflation Expectations
- Capital and Investment
- Capital Markets
- Data Releases
- Economic conditions
- Economic Growth and Development
- Exchange Rates and the Dollar
- Fed Funds Futures
- Federal Debt and Deficits
- Federal Reserve and Monetary Policy
- Financial System
- Fiscal Policy
- Health Care
- Inflation Expectations
- Interest Rates
- Labor Markets
- Latin AmericaSouth America
- Monetary Policy
- Money Markets
- Real Estate
- Saving Capital and Investment
- Small Business
- Social Security
- This That and the Other
- Trade Deficit
- Wage Growth