A falling unemployment rate is generally good news, but when it’s accompanied by a falling labor participation rate, as has generally been the case in the US since the year 2000, the unemployment rate may not necessarily indicate strength, but weakness, if it is because workers are dropping off the unemployment lines because they are taking themselves out of the labor force. Conversely, a “high” unemployment may not necessarily be a negative sign for the economy if it is accompanied by an increasing labor participation rate. Such occurred during the 1980s when the unemployment rate was higher than today throughout the decade, but labor participation rates were strongly rising, not falling.
Population and Labor Force Growth
Long-term job growth is mostly a function of population growth, or more accurately, labor force growth. The rate of job growth is only meaningful once it is compared to the overall trend of population growth. Imagine a country or state of say, 10 million people, where overnight the population immediately doubled. Simply based on the additional resources required for the newcomers to function – houses, grocery stores, utility plants, office construction, retail stores, restaurants, etc. – to achieve the same standard of living for the larger population base, there would in relatively short order be roughly a doubling in employment. While technically, the job market “grew” by millions of jobs, it did not necessarily result in an increased standard of living. Only after it is compared to the population growth can employment growth properly be analyzed.
With the US population growing about 0.7% per year, or about 185,000 per month, a majority of this population growth will become labor pool growth. A ballpark estimate, and close to the one quoted by the Atlanta Fed a couple years ago is that the US needs to gain about 120,000 jobs each month just to maintain the country’s standard of living and keep the unemployment rate stable. So a gain in jobs of 200,000 in a month would mean an “extra” 80,000 or so jobs above what is necessary to sustain the growing population. But a monthly gain in jobs of say, 50,000, would actually be below the growth rate of the labor force, and likely result in a rising unemployment rate, and eventually, a lower collective standard of living.
In fact, if the population of an area grew by 10 million people and the number of jobs grew by a much smaller amount, say 1 million, it is very likely that the economy has suffered a severe material setback and on a per person basis, the economy has likely shrunk significantly. When the percentage increase in the number of jobs in an area is less than the increase in population growth, it would be misleading and possibly deceptive, to refer to the increase in jobs as “job growth”, or “economic growth”. In an economy where employment is increasing no more than the increase in the potential labor force a neutral observer would not properly call this a growth economy, at least not from the perspective of employment alone.
Below is a graph, using data from the Bureau of Labor Statistics, of US employment over the last decade. At the beginning of 2006, non-farm employment was about 135 million. Ten years later, by January 2016, employment was said to have reached 143 million, for an increase of about 8 million jobs. That is 8 million more jobs in the economy, but not that meaningful without comparing it to the population change over the same time period.
According to the US Census Bureau, the population grew from about 298 million at the beginning of 2006 to about 323 million as of the beginning of 2016, an increase of about 25 million. According to the Bureau of Labor Statistics, about a third of adults are not in the labor force at all, mostly due to family circumstances. Combining adults out of the labor force with the number of children (obviously also out of the labor force), suggests a given increase in the growth of a population will result in about half of that population immediately becoming part of the labor force.
With the US population increase over the decade of about 25 million, suggests the labor force, and employment, should have increased by approximately 12 million to 13 million. (Such an increase compares well with the earlier estimate cited by the Atlanta Fed of about 120,000 jobs a month needed to compensate solely for increases in population growth/labor force.) Comparing the actual increase of 8 million jobs over the decade does not indicate a “strong” labor market - it indicates a weak one.
Below I’ve graphed actual job numbers (the green bars) with projected employment if growth had matched the projected increase in the labor force of 120,000 per month. Spanning the decade from 2006, we see that employment growth still has not matched the hypothetical growth in the labor force. Note the red bar does not represent “strong” growth in employment, but would simply match the increases in population growth and the probable labor market growth resulting from that population growth. Even at stronger growth rates in recent years, it will be until the year 2023 when employment reaches the level predicted from population alone. Quite simply, over the full decade, employment growth compared to population growth or the probable labor force, has been very weak.
Compared to the often quoted statistic of 13 or 14 million jobs “created” since the recession ended, if you take away the jobs that would have been added based solely on population growth, the incremental job growth is more like 4.8 million. And we should also remember that since nearly 9 million jobs were lost from 2008-2010, millions of jobs that were “created” simply filled those that were eliminated during the prior recession years. Compared to economic recoveries in prior decades, employment growth has been weak. Over the 12 US recessions since the Great Depression of the 1930s, this latest recovery required the longest amount of time, nearly 7 years, to reach the pre-recession peak in employment (the majority of US recoveries after recessions see employment surpass pre-recession levels within two or three years ).
Margin of Error
While employment statistics over long periods can provide at least some insight as to the direction or strength of the economy, in the short-term, the noise surrounding employment statistics provides almost no useful information, despite the elevated anticipation they stir in commentators and investors.
Every month before the official unemployment report is released there are a great many economists from financial institutions jockeying for position to give their estimate of the number of jobs added the prior month. One estimate will be 120,000 jobs, another 165,000, another 202,000 and so on. Inevitably when the “official” number is released, the analyst that estimated closest to the actual number is celebrated as having incredible analytical skills. In addition, the official number of jobs is compared to the average of the economists’ estimates, which is then deemed cause for celebration or sorrow. So if the average estimate among economists and analysts was that 170,000 jobs were added, but the official report was that 195,000 jobs were added, this is typically seen as a strong indicator that the economy was doing better than expected.
But such comparisons ignore the fact that these are statistical estimates. Statistical estimates have statistical errors. Those errors are based on the sampling size. In the case of US employment statistics, the government does not have the time or resources to ask every person in the country each month if he or she is working and the circumstances of such employment. Therefore statistical sampling is used. The US government uses two surveys to report and estimate employment.
The first is the Payroll Survey. Although it takes a relatively large sample size (more than 50,000 employers), it is still a small fraction of the population. As a result, the standard error in any given month is 114,000 jobs. That means that in a given monthly report, the range of possibilities could be anywhere from the official estimate plus or minus 114,000 jobs. So if the official estimate was 200,000 jobs, the range of possibilities could be anywhere from 86,000 to 314,000. With such an estimate, any of the numbers within the confidence interval could be actual number of jobs added. In fact, with such a wide range, it’s almost a certainty that the number of jobs actually added does not equal the headline estimate, even though that is the one that analysts, commentators and politicians will refer to in analyzing the economy.
Below I’ve shown the average number of payroll gains from May 2015 to May 2016 (using the Payroll Survey). The red bars show the monthly changes in employment, the “headline” numbers. But in addition, I’ve also added upper and lower boundaries to reflect the confidence interval and range of estimates.
The second report used to report employment, which some economists prefer, is the Household Survey. But because of the small sample size, the error is an incredible 500,000 jobs. That means that an estimate of 200,000 jobs in a given month could actually be anywhere from -300,000 to 700,000 jobs. Such an estimate is hardly meaningful at all. Even the Payroll Survey has nearly too wide a range of estimates to be useful on a monthly basis. After all, an economy adding 50,000 jobs probably shows a deteriorating one while one gaining 250,000 signifies a rather robust one. Such a range falls within the wide range of estimation.
There are other problems with the employment reports that more astute commentators have observed. The level of “seasonal adjustments” in the statistics is often immense, making it too easy for the numbers to be massaged to whatever degree wished. The same could be said for the “Birth/Death” model of business creation, which is not only impossible to accurately assess, but are exaggerated at turning points in the economy.
In addition, every month the prior two monthly employment reports are revised, further eroding their value as a short-term clue of economic activity. Finally, as I’ve described before, the employment report is a lagging indicator, not a leading one. The last recession was said to have ended in June of 2009, but employment in the US did not bottom until February the following year. Even if the official employment figures are accurate, they are mostly telling us about business conditions six to twelve months before.
The headline number from the jobs report also does not tell us the makeup of jobs, although some of that information is found deeper in the employment report (though most do not look deeper than the headline). The number of jobs does not define the employment situation of an economy. If a worker loses a $50,000 a year job, and is only able to replace it with a job paying $25,000, the level of employment or unemployment will not change, though it means a reduction in economic income and probably output as well. Such a scenario has been playing itself out over much of the past decade as the number of higher-paying jobs has been in decline, but lower-paying jobs have been increasingly plentiful.
Over the long-term, the employment report is moderately useful for telling us what happened or to what extent. Over the short-term it provides almost no useful information. The nature of such a large and dynamic economy makes it nearly impossible that such a survey will provide any objectively useful information. The reports can certainly have an effect on the short-term psychology of traders and investors, but the noise surrounding them provides almost no useful information as to the real output and direction of the economy.