Tag Archives: employment growth

Not a Good Jobs Report – And Firing the Messenger Will Not Help

Posted on by

Chart 1

Update – August 5, 2025:  In the initial version of this post, I mistakenly said that businesses are required by law to participate in the Current Employment Statistics (CES) survey of the Bureau of Labor Statistics (BLS) – the survey that the BLS jobs numbers are based upon.  This is not correct.  While participation is required in certain states under state laws, there is no federal law on this.  The post below has been corrected to reflect this and has added material on the participation rates.  

On August 1, the Bureau of Labor Statistics (BLS) released its estimate of net job growth in July.  The news was not good.  Net job growth in July was just 73,000, where growth in the number of jobs in health care and social assistance alone came to 73,300.  That is, net job growth for everything else was essentially zero, indeed negative.

But more surprising and concerning than the disappointing growth in July, the updated BLS estimate of total job growth in May was reduced by 120,000 to just 5,000, while the estimate for June was reduced by 133,000 to just 14,000.  The initial job growth estimates for May and June were a healthy 139,000 and 147,000, respectively.  Now, both figures are close to nothing.

Revisions to the monthly job estimates are both automatic and routine.  The BLS always revises the estimates for the most recent two months as each new monthly report is released, as it updates its estimates based on more complete data sent to it by employers who are covered in its Current Employment Statistics (CES) survey.  The CES sample includes approximately 121,000 businesses and government agencies at approximately 631,000 individual worksites – covering a total of about one-third of all nonfarm payroll jobs.  The BLS estimates then rely on timely reporting from these employers.

Participation by employers who have agreed to take part in the survey is high.  While not mandated by federal law (although it is in certain states by state law), approximately 94 to 95% who have agreed to be part of the survey then respond by the time of the final release of the employment estimates (based on response rates over the past year).  But not all respond immediately, and some cannot.

The figures reported are of the number employed by the establishment at some point during the payroll period that covers the 12th day of each month.  These are reported along with figures on the compensation paid by the firm (in dollars) and the total hours worked (and separately the total overtime hours worked).  The firm will not know what these will be until the payroll period is over.  If the payroll period is one week, or even usually two weeks, the firm should be able to file its report (which is normally done online) in time for its data to be included in the initial estimate of employment that the BLS issues each month.  But if the payroll period is for the full month, then by definition the BLS will not have that firm’s data in time for its initial employment estimate of that month.  The BLS thus issues a revised estimate the next month, and then also the following month, as further data arrives.  This is all standard.

The BLS does, however, arrive at an estimate each month for total employment – despite still having only partial reports from its survey – by imputing values for the missing data based on past patterns and relationships.  On average those initial estimates are very good, with the later revisions sometimes higher and sometimes lower.  The average revision since 1979 has been less than 0.01% of the number employed.  In absolute value terms (i.e. in terms of just the revisions themselves, not whether they were positive or negative), the average revision was a still very small 0.05%.

The BLS thus updates its job estimates for any given month at the time of its next monthly report and then again at the time of the monthly report after that, based on the more complete reports it has received by then.  There is also an annual revision of the monthly figures, issued each January, based on updated control totals for overall employment and its composition broken down by such factors as the size of the firm, the sector, and so on.  These control totals are obtained from periodic census information.  The process followed for that annual revision was described in an August 2024 post on this blog.

All this is standard and routine, and follows a methodology that was developed years ago.  Furthermore, it is automated and done by computer, with only a cursory review of the numbers by staff at the end.  The BLS is also fully transparent on the revisions it makes.  One can find on this webpage on the BLS site all the monthly revisions (from both the initial estimate to the second, and from the second to the third), month-by-month, back to January 1979.

Trump paid no attention to this.  Faced with the reality of disappointing jobs numbers for the last several months, Trump decided that the best course of action was to shoot the messenger.  He fired the well-respected and professional head of the BLS, Dr. Erika McEntarfer, asserting (with a clear lack of understanding) that McEntarfer was somehow manipulating the numbers for political reasons.  Trump had also asserted in August 2024 that the BLS under Commissioner McEntarfer had manipulated the annual revision of the benchmark control totals released at that time (which happened also to be a downward revision).  The control totals are released every August, prior to the detailed sectoral and monthly figures then released in the coming January.  See my August 2024 blog post on that episode.

Trump also does not understand that the revisions – while significant – are far from unprecedented in size.  The chart at the top of this post shows the revisions to the monthly figures (defined as the difference between the third estimate and the first, with the sole exception for the June 2025 estimate which is between the second estimate and the first), as a share of the overall number employed in the economy.  It is shown as a share of total employment in order to make meaningful comparisons over time, as there are now 80% more employed in the US than there were in 1979.  The figures for May and June 2025 are shown in red at the far right in the chart, although the red may be hard to see.

The first thing to note from the chart is how small the revisions mostly are, especially over the last quarter century.  Unless there are major economic disruptions underway or soon to be underway (such as from the Covid crisis in 2020 and the recovery from it in 2021, from the 2008/09 Great Recession, and earlier in the years at the end of the Carter presidency and the first term of Reagan, when the economy went through two separate recessions), the range of the revisions is almost always well less than +/- 0.1% of the number employed.  This is extremely small.  Keep in mind that the BLS arrives at its estimates of total employment each month from a zero base, where it asks its (rather large) sample of business establishments how many people they currently employ.  The estimates are not based on a survey asking, for example, what the change in employment at their establishment may have been.

The May and June 2025 revisions were reductions in estimated employment of 0.075% and 0.083% of total employment, respectively.  These were major revisions, but certainly far from unprecedented.  Plotting the revisions onto a histogram with bins of 0.05%, we have:

Chart 2

The May and June revisions fit into the -0.10% to -0.05% bin.  There were a total of 37 monthly cases that fit in that group in the 558 months from January 1979 to June 2025, and there were also 27 monthly cases where the reductions were greater.  That is, revisions similar in magnitude to, or greater than, those in May and June have happened in about 11% of the monthly cases since 1979.

There is no evidence that the BLS or its head Erika McEntarfer somehow manipulated these estimates to arrive at figures that Trump did not like.  Indeed, it would probably be impossible, given how automated the process is.  That does not mean, however, that a new Trump appointee to head the BLS would not be able, over time, to redesign the process in order to give Trump something closer to the numbers he wants.  This will need to be watched closely.

Until recently, the first groups that would be advised of any changes in its methodology that the BLS was considering would have been two advisory panels of outside professionals.  The panels were made up of individuals from universities, research institutes, and private businesses, who were impartial professionals and were not paid (other than for travel expenses).  Those panels – the BLS Technical Advisory Committee and the BLS Data Users Advisory Committee – were, however, dismissed in mid-March by the then new Trump administration.

There were concerns when the panels were dissolved that the Trump administration had plans to politicize the process by which basic economic data is gathered, with the aim of ensuring only flattering figures are released.  The firing of the Commissioner of the BLS appears to be a further step in that process.

Why Voters Are Upset 2: The Proximate Causes of the Underperformance of the US Economy Since the 2008 Crash

Posted on by

Chart 1

A.  Introduction

The previous post on this blog described the slowdown in US growth since the 2008 crash.  GDP fell sharply in the second half of that year – the last year of the Bush administration – due to the crisis in home mortgages leading to a broad collapse in the financial markets.  It led to what has been termed the “Great Recession”.  But unlike in past recessions, GDP never recovered to its previous trend path, even though the unemployment rate fell to lows not seen since the 1960s.  GDP remains well below that previous path today.  The chart above shows how that gap opened up and has persisted since 2008.

The question is why?  The unemployment rate had averaged 4.6% in 2007 – the last full year before the 2008/09 economic and financial collapse.  While the pace of the recovery from the collapse was slowed by federal budget cuts, the economy eventually did return to full employment.  The unemployment rate was at or below 5% in Obama’s last year in office and then continued on the same downward path during the first three years of the Trump administration.  It averaged 3.9% in 2018 and 3.7% in 2019, and hit a low of 3.5% in September 2019.  After the brief but sharp 2020 Covid crisis, the unemployment rate then went even lower under Biden, reaching a low of just 3.4% in April 2023 and averaging just 3.6% in 2022 and again in 2023.  The unemployment rate has not been this low for so long since the 1960s.

In prior times, GDP would have returned to the path it had been on once the economy had recovered to full employment, with resources (in particular labor resources) being fully utilized.  But this time, despite unemployment going even lower than it had been before the downturn, GDP remained far below the path it had been on.  By 2023, real GDP would have been almost 20% above where it in fact was, had it returned to the previous path.  That is not a small difference.

That is, while the economy recovered from the 2008 collapse – in the sense that it returned to the full utilization of the labor and other resources available to it – economic output (real GDP) with that full utilization of resources was stubbornly below (and remained stubbornly below) what it would have been had it returned to its prior growth path.  The economy had followed that path since at least the late 1960s (as seen in the chart above).  Indeed, that same growth path (in per capita terms) can be dated back to 1950 (as the previous post on this blog showed).

This post will examine the proximate factors that led to this.  The post will look first at the growth in available labor.  It has slowed since 2008.  This has not been due to a fall in the labor force participation rates of the various age groups, as some have posited.  We will see below that holding those participation rates constant at what they were in 2007 (for each of the major age groups) would not have had a significant effect on labor force totals.  Rather, labor force growth slowed in part simply because the growth in the overall population slowed, and in part due to demographic shifts:  A growing share of the adult population has been moving into their normal retirement years.  It is not a coincidence that the first of the Baby Boom generation (those born in 1946) turned 62 in 2008 and 65 in 2011.

The second proximate factor is available capital – the machinery, equipment, and everything else that labor uses to produce output.  Capital comes from investment, and we will see below that net investment as a share of GDP has fallen sharply in the decades since the 1960s.  Overall net fixed investment fell by more than half.  This led to a slowdown in capital growth, and especially so after 2008.  There was an especially sharp reduction in public investment.  Since 2008, net public investment as a share of GDP has been only one-quarter of what it was in the 1960s.  It should be no surprise why public infrastructure is so embarrassingly bad in the US.  And net residential investment (as a share of GDP) is only one-third of what it was in the 1960s.  The resulting housing shortage should not be a surprise.

The third proximate factor is productivity.  Labor working with the available capital leads to output.  How much depends on the productivity of the machinery, equipment, and other assets that make up the capital, and that productivity grows over time as technology develops and is incorporated into the machinery and equipment used.  We will see that the rate of growth in productivity fell significantly after 2008.  Given the reduction in net investment and the consequent slowdown in capital accumulation after 2008, it is not surprising that productivity growth also slowed.

For a rough estimate of the relative importance of these three factors – labor, capital, and productivity – I developed an extremely simple Cobb-Douglas production function model to simulate what could be expected.  Despite being simple, it turned out to work surprisingly well both in terms of tracking what actual GDP was (for given employment levels) and in tracking the trend path for GDP given the trend paths of labor, capital, and productivity.

As noted above, the trend level of GDP in 2023 was almost 20% above what GDP actually was in that year – a year when unemployment was at record lows.  Despite being at full employment, the economy was not producing more.  Based on the Cobb-Douglas model, roughly a quarter of the shortfall can be attributed to a slowdown in productivity growth from 2007 onwards.  Of the remaining shortfall, about 60% can be attributed to a smaller stock of capital and 40% to a smaller labor force (both relative to what they would have been had they continued on the same trend paths that they had followed before 2008).

Section B of this post will examine the labor force figures.  Section C will look at what has happened to investment and the resulting growth in available capital.  Section D will then examine the Cobb-Douglas model used to estimate the relative importance of labor and capital both growing more slowly than they had before and the impact of slower productivity growth.  Section E will conclude.

As noted above, labor growth has slowed due to demographic changes as population growth has slowed and as the population has aged.  A rising share of the population (specifically the Baby Boomers) have been moving into their normal retirement years, and this has led to a slower rate of growth in the labor force.  There is nothing wrong with this, it depends primarily on personal choices, and there is no real policy issue here.

In contrast, there are important policy issues to examine on why investment has fallen in recent decades – and especially since 2008 – with the resulting slower rate of capital accumulation as well as slower productivity growth.  But the causes of this are complex, and will not be examined here.  I hope to address them in a subsequent post on this blog.

[Note on the data:  In each chart, I used the most detailed data available for that particular data series, i.e. monthly when available (labor force statistics), quarterly (real GDP), or annual (capital accumulation). The data are current as of the date indicated for when they were downloaded, but some are subject to subsequent revision.]

B.  Growth in the Labor Force

Growth in the US labor force has slowed, but by how much, when did this start, and why?  We will examine this primarily through a series of charts.  Most of these charts will be shown with the vertical axis in logarithms.  As you may remember from your high school math, in such charts a straight line will reflect a constant rate of growth.  The slope of the lines will correspond to that rate of growth, with a steeper line indicating a faster rate of growth.

The trend lines in the charts here (including in the chart at the top of this post) have all been drawn based on what the trends appear to be (i.e. “by eyeball”) in the periods leading up to 2008.  They were not derived from some kind of statistical estimation, nor from a strict peak-to-peak connection, but rather were drawn based on what capacity appeared to be growing at over time.  They were also drawn independently for aggregate real GDP (Chart 1 above), for growth in the labor force (Chart 2 below) and for growth in net fixed assets (Chart 10 below).  Despite being independently drawn, we will see in Section D below that a very simple Cobb-Douglas model finds that they are consistent with each other to a surprising degree, in that the predicted GDP trend corresponds to and can be explained by the trends as drawn for labor and for capital.

Starting with the labor force:

Chart 2

The US labor force grew at a remarkably steady rate from the early 1980s up to 2008.  Prior to the 1980s, it grew at a faster pace (a trend line would be steeper) as women entered the labor force in large numbers and later as the Baby Boomers began to join the labor force in large numbers in the early 1970s.

But then that steady rise in the labor force (of about 1.3% per annum before 2008) decelerated sharply.  The growth rate fell to only 0.5% per year between 2007 and 2023.  Why?

We can start with overall population growth:

Chart 3

Population, too, had grown at a steady pace prior to 2008.  But population growth then slowed.  In this context, it is not surprising to see that growth in the labor force also slowed.

But there is more to it than just this.  Before 2008, the US population had been growing at a similar rate as the labor force, thus leading to a fairly constant share of the labor force in the population (generally in the range of 50 to 51%):

Chart 4

But then, in 2008, the share of the labor force in the US population fell.  Growth in the labor force slowed by more than growth in the US population.  What were the factors behind that?

One assertion that is often made is that labor force participation rates fell.  At an aggregate level this is, almost by definition, true.  As a share of the US adult population (those aged 16 and over), the labor force participation rate fell from 66.0% in 2007 to 62.6% in 2023 (using standard BLS figures).  But one can be misled by focusing on the aggregate participation rate.  The overall participation rate came down not because those in various age groups became less likely to join the labor force, but rather because an increasing share of the population was aging into their normal retirement years.

The BLS provides seasonally adjusted figures for the labor force broken into three age groups: those aged 16 to 24, those aged 25 to 54, and those aged 55 or more.  Labor force participation rates are provided for each of these three groups, and one can calculate what the labor force participation would have been for each had the participation rate always been at that of 2007:

Chart 5

The line in red shows what the labor force then would have been, with the line in blue showing the actual labor force and the line in black the trend (the same trend as in Chart 2 above).  While it would have made a significant difference before the 1980s (as women were not participating in the formal labor force to the same degree then), between 2008 and 2023 it makes very little difference.  The labor force would have still fallen by about the same figures relative to its previous trend.

Rather, the labor force has been aging, with a growing share of the population now in the normal retirement years when labor force participation rates are low.  From the BLS numbers, one can work out the share of the population that are age 55 or older:

Chart 6

The share in the population of those aged 55 or older started to turn sharply upward around 1998.  They thus would have been 65 or older starting around 2008.  And as noted before, this is also when the first of the Baby Boomers (those born in 1946) would have started to reach their normal retirement age.

[Side note:  The discontinuities that one sees at various points in this chart are there because of adjustments made by the BLS in their control totals.  They adjust these control totals once new results are available from the decennial US population censuses.  They need such control totals for the shares of the various demographic groups since the labor force estimates come from its Current Population Survey (CPS), and as with any survey, control totals are needed to generalize from the sample survey results.  But the BLS does not then revise prior CPS estimates once the control totals are updated with each decennial census.  That then leads to these discontinuities.  For our purposes here, those discontinuities are not important.]

Labor force growth thus slowed from 2008 onwards.  This can be explained by basic demographics with an aging population.  This was not due to less willingness to participate in the labor force – an assertion one often sees.  Holding participation rates constant at what they were in 2007 for just three broad age groups led to no significant difference in what the labor force would have been.  Rather, people are just aging into their normal retirement years.

C.  Growth in Capital

Labor works with machinery, equipment, structures, and other fixed assets – which together will be referred to as simply capital – to produce output.  Those assets also reflect the technology that was available and economic (in terms of cost) when they were installed.  Those assets are acquired by investment, and it is important to recognize that net investment has fallen sharply over the last several decades.

This is not often recognized, as most analysts and news reports focus not on net investment but rather on gross investment.  Gross investment figures are provided in the GDP accounts that are released each month, and gross investment as a share of GDP has not varied all that much.  The decade-long averages for gross private fixed investment have varied only between 16 and 18 1/2% of GDP since the 1960s.

But the accumulated stock of capital does not arise simply out of gross investment but rather out of investment net of depreciation – i.e. net investment.  Less attention is paid to net investment figures, and estimating depreciation is not easy.  It is certainly not depreciation as defined by tax law, as tax law as written reflects a deliberate attempt to encourage investment by allowing firms to declare depreciation to be greater than it actually is (e.g. through accelerated depreciation).  Assigning a higher cost to depreciation will reduce reported profit levels and hence reduce what needs to be paid in taxes on that profit income.

For the GDP accounts (NIPA accounts) the BEA needs to record what actual depreciation was, not what depreciation as allowed under the tax code may have been.  The BEA estimates of this are carefully done and are the best available.  However, one still needs to recognize that these are estimates and that there are both conceptual and data issues when estimates of depreciation are made.

Based on the BEA estimates in the NIPA accounts, both public and private net fixed investment levels – as shares of GDP – have fallen sharply since the 1960s:

Chart 7

There are significant year-to-year fluctuations in the shares – especially in the private investment figures – as investment varies significantly over the course of the business cycle.  It falls in recessions and increases when the economy recovers.  The trends may thus be more clearly seen using decade averages of the net investment shares:

Chart 8

Total public and private net fixed investment fell from over 10% of GDP in the 1960s (and almost as much in the 1950s) to just 4.2% of GDP in the period from 2009 to 2023 – a fall of close to 60%.  Total private net fixed investment fell from about 7% of GDP in the 1950s, 60s, and 70s, to just 3.4% since 2009 – a fall by half.  Public net fixed investment fell even more sharply:  from over 3% of GDP in the 1960s to just 0.8% of GDP in recent years – a reduction of three-quarters (in the figures before rounding).  It should be no surprise why public infrastructure is so embarrassingly poor in the US.

The chart also shows private net fixed investment broken down into the share for investment in residential assets (housing) and non-residential assets.  Much of the decline in private net fixed investment was driven by an especially sharp reduction in investment in housing. Still, private investment in assets other than housing has also been cut back substantially, with a reduction of over 40% compared to where it was in the 1980s.

Based on their net fixed investment estimates and other data, the BEA also provides estimates of how the accumulated stock of real fixed capital has changed over time, with those levels shown in terms of quantity indices.  The resulting rates of growth in accumulated capital (which the BEA refers to, more precisely, as the net stock of fixed assets) have declined sharply with the reductions in the net investment shares:

Chart 9

In the 1960s, the annual growth rates varied between 3.5% (for residential fixed assets) and 4.4% (for public fixed assets).  But in the period from 2009 to 2023 those growth rates had fallen to just 1.9% for private non-residential fixed assets, 1.1% for public fixed assets, 0.8% for residential fixed assets, and 1.3% for all fixed assets.  Such a slow rate of capital accumulation will not be supportive of robust growth.

The reductions in the growth rates were especially sharp following the 2008 crisis.  This led capital accumulation to fall well below the trend path that it had previously been on:

Chart 10

As was the case for growth in the labor force, there is again a substantial fall after 2008 in the growth of an important factor in production relative to its previous trend.  This time it is accumulated capital.  It should not be surprising that this slowdown in the growth of both available labor and capital would then be accompanied by a slowdown in the growth of GDP – all relative to their previous trends.  But an open question is how much of the close to 20% shortfall in GDP as of 2023 was due to labor, how much to capital, and how much to the productivity of labor working with the available capital?  This will be examined in the next section.

D.  Modeling GDP:  The Relative Importance of Labor, Capital, and Productivity to the Shortfall

Output (GDP) has fallen relative to the path it was on before – and a 20% shortfall is a lot – as have both the size of the labor force and of accumulated capital.  To estimate how much of the shortfall in GDP can be attributed to the shortfall of labor, how much to the shortfall of capital, and how much to a slowdown in the growth in productivity of that labor and capital, one needs a model.

For this analysis, I used the extremely simple but standard model of production called the Cobb-Douglas.  Its formulation is credited to Paul Douglas (an economist) and Charles Cobb (a mathematician) in 1927, although Douglas recognized and acknowledged that a number of economists before them had worked with a similar relationship.  While extremely simple, it allows us to arrive at an estimate of how much of the shortfall in GDP can be attributed to labor, how much to capital, and how much to a change in productivity growth.  Despite being simple, there was a good fit when the model was tested for its predictions of GDP against what GDP actually was historically.  There was also a very surprisingly good fit against whether the trend growth in GDP was close to what the model predicted based on the trend growth observed for labor and for capital.

The Cobb-Douglas production function is an equation that relates what output (real GDP) would be for given levels of labor and capital as inputs.  The following subsection will provide a brief overview of that equation and of the parameters used.  Those who prefer to avoid equations can skip over this section and go directly to subsection (b) below, where the model was tested via a comparison of the model’s predicted values for GDP to what GDP actually was, both year-by-year and in its trend.

a)  The Cobb-Douglas Equation and Parameters 

The Cobb-Douglas production function can be written as:

Y = A(1+r)tLβK1−β

where Y is real GDP, L is labor, K is capital as measured, r is a rate of growth for the increase in productivity over time (t), A is a scaling factor, and β is an exponent indicating how much output (Y) will increase for a given percentage increase in L as an input.  With constant returns to scale (which is generally assumed), the exponent for K will then be 1- β.  They will also match (under the assumptions of this model) the shares in national income of labor and capital, respectively.  In the NIPA accounts for 2023, the compensation of employees was 62% of national income.  All other income (e.g. basically various forms of profit) was 38% of national income.  I rounded these to just a 60 / 40 split, so β = 0.60 and 1-β = 0.40.

Productivity will grow over time.  That is, the output that can be generated for a given amount of labor and of capital will grow over time.  As technology changes and is reflected in the accumulated stock of capital, labor working with the available machinery and equipment will be able to produce more.  While the contribution of the growth in productivity can be incorporated into the Cobb-Douglas in various different ways, the simplest is to assume that it augments the combination of labor and capital together.  This growth in productivity can then also be referred to as the growth in Total Factor Productivity (TFP).

For the simulations here, I took the year 2007 (the last full year before the 2008 collapse) as the base period, and hence scaled the labor and capital inputs in proportion to what they were in 2007.  Thus they were both set to the value of 1.00 in 2007, and if they were then, say, 10% higher in some future year they would have a value of 1.10 in that year.  The scaling coefficient A would then be equal to real GDP in 2007 ($16,762.4 billion in terms of 2017 constant $).

Finally, the rate of TFP growth was set so that GDP as modeled would roughly track what the actual values for GDP were historically.  It turned out that an annual rate of growth in TFP of 1.20% worked well for the years leading up to 2007, with this then falling to 0.90% per year in the years following 2007 up to and including 2023.  I did not try to fine-tune this to any greater precision (i.e. I looked at annual TFP growth to the nearest 0.1% and not more finely, i.e. to 1.20% or 1.30% but not to 1.21%).  I also constrained the TFP growth to be at just one given rate for all of the years before 2007 (1.20%) and one rate after 2007 (0.90%), even though it is certainly conceivable that it could fluctuate over time.

b)  Comparison of GDP as Modeled by the Cobb-Douglas versus Actual and Trend GDP

The Cobb-Douglas just provides a model, and the first question to address is whether that model appears to track what we know about the economy.  There were two tests to look at:  1)  how well it tracked actual GDP as a function of actual labor employed and capital (net fixed assets), and 2)  how well the model tracked the trend line for GDP growth (as drawn in Chart 1 at the top of this post) as a function of the trend line as drawn for the labor force (Chart 2) and the trend line as drawn for capital (Chart 10).  Keep in mind that these trend lines were drawn independently and “by eyeball” based on what appeared to fit best in the decades leading up to 2008.

This chart shows how well the modeled GDP tracked actual historical GDP:

Chart 11

The line in black shows what actual real GDP was in each year from 1959 to 2023.  The line in red shows what the simple Cobb-Douglas model predicted real GDP would be in each year with the parameters as discussed above and with the labor input based on actual employment in that year rather than the available labor force.  The capital input is always available net fixed assets (as an index, which is all we need for the relative changes), as estimated by the BEA for the NIPA accounts (shown in Chart 10 above).

The line in red for the modeled GDP tracks well the line in black of actual GDP, especially from about the early 1980s onwards.  A reduction in the growth rate for TFP in the years prior to 1980 would have led it to track the earlier years better, but I did not want to try to “fine-tune” the TFP rate.  My main interest is in how well predicted GDP tracks actual GDP over the last several decades.  Over this period, a simple Cobb-Douglas with fixed parameters and with TFP growth of 1.20% for the years before 2007 and 0.90% in the years since, tracked quite well.  And this was over a period when GDP grew from just $7.3 trillion in 1980 (in 2017 constant $) to $22.7 trillion in 2023 – more than tripling.

A second test is whether something close to the GDP trend line (as drawn in Chart 1 at the top of this post) will be generated by the Cobb-Douglas model when the labor force grows on its trend line (as drawn in Chart 2) and capital grows on its trend line (as drawn in Chart 10).  Each of these trend lines were drawn independently and “by eyeball”.

The answer is that it does, and to an astonishing degree.  This may have been the case in part by luck or coincidence, but regardless, was extremely close.  The line for GDP as predicted from the Cobb-Douglas model using labor and capital inputs that each followed their own trend lines, was so close to the GDP trend line that they were on top of each other in the chart and could not be distinguished.

One should keep in mind that, by construction, the predicted GDP in 2007 from the Cobb-Douglas model will be equal to actual GDP in that year.  The scaling factor was set that way.  But the question being examined is whether the predicted GDP (based on the labor and capital trend lines) would drift away from the trend line for GDP (as drawn) over time.  It did not.  Calculating it back over a 60-year period (i.e. equivalent to going back to 1947 from the 2007 base), the predicted GDP was only 0.7% greater than what GDP on the drawn trend line would have been 60 years before.

This is tiny, and indeed so tiny that I at first thought it might be a mistake.  But after simulating what would have been generated by various alternative parameters for the Cobb-Douglas, as well as alternative trend paths for labor and capital, the calculations were confirmed.  The implication is that the trend lines for GDP, labor, and capital – while independently drawn – are consistent with each other and with this simple Cobb-Douglas framework.

The rate of productivity growth – TFP growth – for the years leading up to 2007 was 1.20%.  It was derived, as noted above, by trying various alternatives and seeing which appeared to fit best with the figures for actual GDP in those years.  Going forward from 2007, however, it would have over-predicted what GDP would have been.  What fit well with the data on actual GDP (and based on actual employment and available net fixed assets) was a reduction in the TFP rate from the 1.20% used for the years up to 2007 to a rate of 0.90% for the years after.

The resulting path for actual GDP versus the path as modeled by the Cobb-Douglas can be more clearly seen in the following chart.  It is the same as Chart 11, but now only for the period from 2000 to 2023:

Chart 12

The red line shows the path for the simulated GDP, where from 2007 onwards the assumed TFP growth rate was 0.90%.  The fit is very good, and especially in 2022 and 2023 – the years of most interest to us – when the simulated GDP (from the Cobb-Douglas) is almost identical to actual GDP.  These are both well below the path (the green line) that would have been followed based on the previous trend growth in labor and capital, as well as the continuation of productivity growth at a 1.20% rate rather than falling to 0.90%.

c)  The Causes of the Below Trend Growth of GDP Since 2008

From this simple Cobb-Douglas model, we can try various simulations of what growth in GDP might have been had the labor force continued to grow at the rate it had before 2008, had capital continued to grow at the rate it had before 2008, and had productivity (TFP) continued to grow at the rate it had before 2008.

The results are shown in the following chart:

Chart 13

The resulting paths for GDP are shown as a ratio to what actual GDP was in each year, with the differences expressed in percentage points.  By definition, there will be no difference for actual GDP, so it is a flat line (in black) with a zero difference in each year.  The line in red then shows what the modeled GDP was in each year in terms of the percentage point difference with actual GDP, using actual labor employed in each year and available capital.  The red line shows at most a 2 percentage point difference with actual GDP – and no difference at all in 2022 and 2023.  The model tracks actual GDP well when the labor input is equal to observed employment.

The line in blue then shows what GDP would have been (according to the model) had capital growth continued after 2007 along its pre-2008 trend path (the path drawn in Chart 10 above) while labor grew at the actual rate of employment.  It shows how much the shortfall in GDP was as a consequence of capital accumulation slowing down from 2008 onwards.  As seen in the chart, the impact of this slowdown has grown over time.

The line in orange shows what GDP would have been had labor growth continued after 2007 on its pre-2008 trend path (the path drawn in Chart 2 above), while capital grew not along its trend but rather as measured.  Here one needs to take into account that the growth rate of actual employment and the growth rate of the labor force will only match between periods when the unemployment rate was the same.  Thus comparisons should be limited to periods when the economy was close to full employment, such as between 2007 (when unemployment averaged 4.6%), 2016 to 2019 (annual unemployment rates of 4.9% to 3.7%), and 2022/23 (annual unemployment rates of 3.6%).  That is, the “peaks” seen in the orange line in 2009 and again in 2020 are not significant, as they reflect labor not being fully used.  This was not because the labor force was not available but rather due to the disruptions of the downturns in those years.

The line in burgundy then shows what GDP would have been (in terms of its percentage point difference with actual GDP) had both labor and capital inputs continued to grow (and been used) on their pre-2008 trend paths.  Note that the values here will not be the simple addition of the percentage point contributions of the slower than trend growth of the labor force and the slower than trend growth of capital.  The Cobb-Douglas relationship is a multiplicative one, not a linear one.  But if one does multiply out the two (the blue and orange lines, but as ratios rather than percentage points), and adjust for the model’s tracking error (the red line), one will get the impact of the two together (the burgundy line).

Finally, there is the impact of the slowdown in TFP growth from 1.20% per year before 2007 to 0.90% after.  That will appear as the difference between what GDP would have been had it followed the previous trend path (the green line in the chart) and the impact of labor and capital both slowing down from their respective trends (the burgundy line).  Its impact grows steadily larger over time.

Based on these simulations, as of 2023 approximately 25% of the shortfall in GDP relative to what it would have been had it continued on its pre-2008 trend can be attributed to a fall in the rate of productivity growth (TFP) from 1.20% to 0.90%.  Of the remaining shortfall, approximately 60% was due to the slowdown in investment and hence capital accumulation, and approximately 40% was due to the slowdown in the growth of the labor force.  Or put another way (and keeping in mind that the impacts are not linearly additive, but only approximately so), of the total shortfall in 2023, about 70% was due to the slowdown in productivity growth together with the related slowdown in capital growth, and about 30% was due to the slowdown in labor force growth.

But these figures are for 2023 and will shift over time.  Going forward, and unless something is done to change things, the shortfall in GDP (its deviation from the pre-2008 trend) will be widening, and the shortfall in capital accumulation (due to the fall in investment as a share of GDP) plus the related reduction in productivity growth, can be expected to account for an increasing share of this increasing shortfall in GDP.  These already accounted for about 70% of the shortfall in 2023, and on current patterns that share will grow in the coming years.

E.  Conclusion

GDP fell sharply in the economic and financial collapse that began in the second half of 2008.  But while there was a recovery, with employment eventually returning to full employment levels, GDP never returned to the path it had previously been on.  This was new.  In prior recessions (as seen in Chart 1 at the top of this post), GDP was back close to its earlier path once employment had recovered to full employment levels.  As a consequence, by 2023 GDP would have been close to 20% higher than what it was had GDP returned to its previous path.  And 20% higher GDP is huge.  In terms of current GDP in current prices, that is close to $6 trillion of increased output and incomes each year.  Total federal government spending on everything is about $7 trillion currently.

The proximate causes of this can be broken down into three.  First, the labor force began to grow at a slower rate in the years following 2008.  This was not due to labor force participation rates falling for individual age groups.  Rather, this in part reflected a slowdown in the growth of the overall US population (and to this extent, will then be offset when GDP is looked at in per capita terms).  But in addition, there was the impact of an aging population, with the Baby Boom generation entering into their normal retirement years.

In policy terms, there is not much one can or should want to do about labor force growth.  Population growth is what it is, and an aging population will see an increasing share of the population moving into their retirement years.  These all reflect personal choices.

In contrast, the slowdown in investment and the resulting slowdown in capital accumulation and productivity growth is a policy question that merits a careful review.  Why are firms investing less now than they did before?  Profits (especially after-tax profits) are at record highs and the stock market is booming.  In a market economy where firms are avidly competing with each other, this should have led to an increase – not a decrease – in net investment.

A future post in this series will examine the factors behind this.  But first, a post will examine the specific case of residential investment.  Net residential investment fell especially sharply after 2008 (see Charts 8 and 9 above), while home prices have shot up.  Housing is important, and its rising cost has been the source of much displeasure in recent years by those who do not own a home and must rent.  The rising cost of housing is the primary (indeed, the only) reason why the CPI inflation index remains above the Fed’s target of 2%.  It merits its own review.

The Impact on the Employment Numbers of the August 21 Announcement of the Bureau of Labor Statistics

Posted on by

A.  Introduction

The Bureau of Labor Statistics (BLS) issued an announcement on August 21 that said it had made a preliminary estimate that its figure for total employment as of March 2024 will be revised downwards by 818,000.  Some news media articles treated the announcement as if it were something to be alarmed by, and Trump issued a blast on the social media site he owns.  Trump asserted:  “MASSIVE SCANDAL!  The Harris-Biden Administration has been caught fraudulently manipulating Job Statistics to hide the true extent of the Economic Ruin they have inflicted upon America.  New Data from the Bureau of Labor Statistics shows that the Administration PADDED THE NUMBERS with an extra 818,000 Jobs that DO NOT EXIST, AND NEVER DID.  The real Numbers are much worse …” (sic, and capitalization as in the original).

None of this is true, but we know that accuracy has never been a strong point for Trump.  And such derogatory comments about the professionals at the Bureau of Labor Statistics just doing their jobs are also appalling.  There was nothing scandalous in their work.  A few basic points:

a)  Such a “preliminary benchmark revision” is issued every August, as part of an annual process by which the monthly employment estimates of the BLS are updated and anchored to (benchmarked to) more comprehensive estimates of employment.  This is done on a regular and routine basis every year.

b)  The date of the announcement is certainly not a secret, but rather is set well beforehand.  One will find it, for example, highlighted in a box on page 4 of the July jobs report that was released on August 2.  There was no attempt at a cover-up nor a leak.

c)  The 818,000 jobs figure is not some sort of monthly job number that people normally associate with the monthly jobs reports, but rather reflects an estimate of the change in the total number of people employed in March 2024.  The monthly employment estimates are then anchored to this benchmark, which will be updated again next year to an estimate for March 2025.  Employment still grew – and grew strongly – over the period from March 2023 (the previous benchmark) to March 2024 (which, when finalized, will become the new benchmark), but not by as much as was estimated before.  The previous estimate was of job growth of 2.9 million over this March to March period.  The new estimate (if the preliminary benchmark estimate holds – but bear in mind that it is preliminary and may well change) is of job growth of about 2.1 million.  That is still strong job growth.

d)  Many of the news articles highlighted that the 818,000 revision in estimated overall employment is high.  But one should keep in mind that it is equal only to about 0.5% of total employment.  That is, the revised figure (if the preliminary benchmark figure holds) will be 99.5% of what had been estimated earlier.  The 0.5% revision is also certainly not unprecedented.  Such revisions are part of a regular annual process, and figures the BLS provides going back to 1979 show that there have been revisions of 0.7% twice (in 1994 and 2009), 0.6% twice (1991 and 2006), and 0.5% four times (1979, 1986, 1995, and now in 2024).  That is, there have been such revisions to estimated overall employment by 0.5% or more a total of 8 times in 46 years, or 17% of the time.  A 0.5% change is large compared to what the figures normally are, but it is certainly not unprecedented, and in several years the revisions have been greater.

There is no scandal here.  There is no indication of manipulation.  And if there was some kind of politically motivated manipulation possible, doesn’t Trump realize that it would have made much more sense to manipulate the employment figures to be higher rather than lower?  Did he give even a few seconds of thought to his accusations?  The BLS is just doing the professional job it always has.

With all the publicity that has surrounded the BLS announcement, some may find of interest a description of how this annual updating process of the employment estimates works.  We will review that in the next section below.  The section following will then look at the figure itself – the 818,000 change in estimated overall employment – and what it may imply.  While still preliminary, the final estimate is likely to be close.  And the main message is that the basic story on employment growth during the Biden presidency has not changed.  Employment growth under Biden has been, and continues to be, exceptionally strong.

The chart at the top of this post updates a chart I provided in an article on this blog that was posted on August 21 – the day the BLS announcement came out.  I saw that announcement and the reports on it just after I posted that article.  One focus of that post was on the employment record under Biden and how it compared to the record under Trump.  The chart above replicates one in that August 21 post, but with the addition of what the path of estimated employment may now look like once the new benchmark is taken into account for the recent employment estimates.  That revised path is shown in orange.  It is a very rough estimate as the BLS has not yet worked out and released what the monthly employment figures will be with the new benchmark.  They are working on that now, and will release it – as they always do – in early February as part of the January monthly jobs report.

The path in orange is below the original one in red, but follows the same basic course.  It is still rising at a strong pace, and the basic message remains the same.  Job growth under Biden has been far stronger than what it was under Trump.

B.  The Annual Process of the BLS to Update Its Monthly Employment Estimates

The discussion in this section is based on material the BLS provides on its website on the process it follows in updating its monthly employment estimates to tie them (anchor them) to comprehensive employment estimates arrived at once a year from census-like figures.  The summary description provided here is based primarily on the BLS posts here and here.

The monthly jobs report of the BLS (more formally: “The Employment Situation” report) is eagerly awaited by many.  It provides estimates for what happened to the number of “jobs created” during the past month (more accurately, the change in the estimated number of nonfarm employees between the current month and the month before), as well as the unemployment rate along with numerous other measures of the labor market.

The report is produced on a very tight schedule.  The employment statistics come from a sample of establishments (both public and private, and called the Current Employment Statistics, or CES, survey), where the employing entities report to the BLS the number of employees on their payroll in the week of the month which includes the 12th day of the month.  The BLS jobs report is then issued at 8:30am on the Friday three weeks later, which is usually the first Friday of the following month.

(There are also figures in the monthly Employment Situation report on unemployment, the number in the labor force, and other figures that are obtained through the much smaller Current Population Survey (CPS) of households.  Most of what we will discuss here will be for the CES survey of business establishments, but similar modeling issues arise with the CPS survey, where there is also an annual process to update the model parameters.)

The survey of establishments is a rather comprehensive one, where the reporting entities account for about one-third of all nonfarm payroll jobs.  But it is still a sample survey, and the BLS needs to estimate from this survey the overall number of employees in the country (and hence what the change was from the previous month – the growth in the number employed).

For this, what is mainly needed is a large set of weights that the BLS can use to aggregate the reports it receives from firms of various types.  That is, to estimate the overall totals the BLS will need to know what weight to give to what is found in the survey reports for a particular type of firm (such as of a given size), operating in a particular sector, and perhaps categorized in other ways as well.

For example, small firms with up to 99 employees accounted for (in March 2023) 40.0% of all private employment in the country.  But while 70.4% of the number of private firms sampled by the BLS for the CES were in this category of up to 99 employees, those in the CES survey sample accounted for only 4.6% of total private employment.  Those firms are all small.  In contrast, large firms with employment of 1,000 or more were 6.2% of the number of private firms sampled by the BLS.  But those firms accounted for 68.4% of total private employment in the sample (and 28.8% of the total private employment in the country).

The BLS thus needs to know what weights to assign to each of these categories of firms to determine the overall totals.  The annual benchmarking exercise provides this.  A comprehensive census-type of exercise is needed, and for this the BLS uses primarily the March report of the Quarterly Census of Employment and Wages (QCEW) – which the BLS is also responsible for.  The QCEW is a comprehensive accounting of essentially all workers in the US based on the filings (and unemployment insurance tax payments) all firms are required to provide for the unemployment insurance program.

About 97% of the workers counted in the CES reports will be covered by regular unemployment insurance and hence included in the QCEW reports.  About 3% of workers are not, and the BLS uses various methods to arrive at a count for them.  Such “noncovered employment” (as the BLS labels it) includes, for example, certain workers at nonprofits and religious organizations, certain state and local government workers, railroad workers (where unemployment insurance is covered under the Railroad Retirement Board), paid interns and apprentices, and a range of others.

Keep in mind also that “employment” as reported in the monthly jobs report is for the nonfarm payroll, and thus excludes the self-employed as well as those working on farms (whether as self-employed owners or as employees).  But based on CPS data (the survey of households), those employed on farms (whether as employees or self-employed) only account for 1.4% of total employment.  That is so small that changes in on-farm employment do not have a significant impact on overall employment growth.  More potentially significant are the self-employed, who equal 6.1% of total employment according to the CPS data.  Unemployment insurance does not cover the self-employed, but those who are self-employed are also not employees and hence are not included in the CES definition of the nonfarm payroll.

The BLS then uses the detailed census counts from the March QCEW each year (supplemented by various sources of information for the remaining 3% of employees) to work out the weights to use to aggregate to the global estimates.  The March QCEW figures (as supplemented for the remaining 3%) then serve as an anchor on the employment totals.  It is updated on a routine basis annually on a calendar schedule that is set well ahead of time.  The monthly employment estimates are then worked out over the course of the year relative to the annual anchors of every March.

In addition to working out the weights to use to go from the monthly survey results to the overall totals, the BLS must also estimate the changes over time in the number of firms in each category.  That is, it needs to have an estimate for the number of new firms in each category that have begun operations each month (births), plus the number of firms that have ceased operations (deaths).  The QCEW census data will, by its nature, have nothing on the births and only outdated and now wrong information on the deaths.  The BLS updates its model of firm births and deaths each year as well, as part of its annual process of updating the benchmarks.

There has been speculation that the relatively large estimated reduction in estimated total employment of 818,000 in March 2024 may have been due in part to issues in the estimates of firm births and deaths.  There was an especially large jump in the number of new business establishments that opened in 2021 – a jump of 33% over what it was in 2020 or an increase of 37% over what it was in 2019 – to 1.4 million new firms in that year.  And the number of new firms was again at this record high of 1.4 million in 2022.  But small new firms typically struggle after a year or two, and many close even in the best of times.  It is possible that the BLS model for firm births and deaths did not capture well that this large jump in new business creation in 2021 and again in 2022 was followed by a relatively high number then closing in 2023 and 2024.

The BLS work begins once the March QCEW data become available, and each August it announces its preliminary benchmark revision for total employment in the prior March.  This is what the BLS announced on August 21, that Trump attacked.  The BLS will now work out the month-by-month implications of the new benchmark, adjusting the monthly employment figures that it had earlier estimated to reflect the new benchmark.  These revised monthly figures will be announced as part of the release of the January 2025 jobs report on Friday, February 7, 2025.  It does this in every January jobs report each year.

The benchmark figures on total employment are not seasonally adjusted numbers.  The anchors are the figures for each March, and hence the anchors in the upcoming revision will be for March 2023 (which is unchanged from what was determined before) and March 2024 (the new one).  The non-seasonally adjusted employment numbers will then be revised for the 21 months from April 2023 through to December 2024.  From April 2023 to the new March 2024 benchmark, the monthly employment figures will be adjusted in a simple linear fashion based on what the overall change in employment was between the March 2023 and March 2024 anchors.  If the final estimate turns out to be 818,000 (the same as the preliminary estimate), then that means the April 2023 non-seasonally adjusted employment estimate will be reduced by 68,167 (equal to one-twelfth of 818,000), the May 2023 estimate will be reduced by 136,333 (two-twelfths of 818,000), the June 2023 estimate by 204,500, and so on until the March 2024 employment estimate is reduced by 818,000.

The April 2024 to December 2024 figures for non-seasonally adjusted employment will then be re-estimated based on the models the BLS has updated based on the new March 2024 anchor estimates.  Keep in mind that by the time the January 2025 employment estimates are ready to be released (in early February 2025), the BLS will already have issued estimates for the April to December 2024 figures.  The revised estimates for all of the 2024 estimates are then provided in the Employment Situation report along with the employment figures for January.

The seasonally adjusted employment figures are then also updated.  Seasonally adjusted figures are calculated based on a statistical analysis of the regular annual patterns seen in the non-seasonally adjusted figures, using standard statistical programs.  The model parameters for this are re-estimated once the new non-seasonally adjusted employment figures are determined, and the BLS then goes back and revises the seasonally adjusted monthly employment estimates for a full five years.  Hence, once the January jobs report is released (on February 7 next year), one will find that the seasonally adjusted employment figures for the most recent five years (available online) will have also changed.

The January jobs report also has a section, in the interest of full transparency, showing what the new seasonally adjusted employment estimates are for each month of the past year, what the BLS had previously published, the difference, and the month-to-month employment changes (number of “new jobs”) as revised, as published before, and the difference.  All of this is routine.

The process is well-established and has been followed for at least 46 years (I have not looked farther back).  While the methods constantly evolve and are improved over time, there is no basis for Trump’s attack on the integrity of the BLS.

C.  How Much of an Impact?

The BLS was clear in its announcement that the new benchmark estimate for total employment in March 2024 is preliminary.  It is making this initial estimate available to the public in the interest of transparency, even though it has yet to work out the implications for the month-to-month employment figures.

But while preliminary with month-to-month specifics yet to be estimated, it is possible to get a sense of how significant a change this will likely entail to the pattern of employment growth under Biden.  And the answer is not much.  Furthermore, the change is in the direction one should have expected.  As discussed in my recent post on the economic record of Trump compared to that of Biden and Obama, employment growth during Biden’s term has been extremely fast.  This growth (whether based on the prior estimates or the preliminary revised estimates) has continued at a pace over the last year that is well in excess of separate estimates of growth in the labor force.  Over time, and at a constant unemployment rate, employment can only grow as fast as the labor force does.  In the past year the labor force participation rate rose slightly (from 62.6% of the adult population to 62.7%), which led to somewhat faster growth in the labor force than would be the case with a constant participation rate.  But the longer-term trend has been for the participation rate to drift downwards, as an aging population is leading to a higher share of adults in the usual retirement years.

The current estimate for the period of March 2023 to March 2024 – prior to any benchmark change – has been that total employment grew by 2.90 million.  This is based on the seasonally adjusted figures.  Growth over this period in the non-seasonally adjusted figures was a similar 2.96 million.  The preliminary benchmark change in total employment in March 2024 is 818,000, and formally this is the change in the non-seasonally adjusted figure for employment.  But it makes little difference whether one uses this to adjust the seasonally adjusted figures on employment or the non-seasonally adjusted figures.  With either, one ends up with a new figure for total employment in March 2024 of 2.1 million within round-off.

The month-by-month changes in the total employment estimates have yet to be worked out by the BLS, as noted before, but one can make a very rough estimate of what those might be.  The aim here is simply to give a sense of what the magnitudes are so that one can see – as in the chart at the top of this post – what the path of employment under Biden might then look like in comparison to the paths under Trump and Obama.

A number of assumptions are needed.  First, while the 818,000 adjustment in the benchmark employment total is formally a non-seasonally adjusted figure, I will assume the seasonally adjusted estimate will be similar.  The chart at the top of this post uses seasonally adjusted figures throughout, and the adjusted path for employment growth under Biden will be as well.

Second, for the period from April 2023 to March 2024 I adjusted the month-by-month employment estimates linearly, as the BLS does (although the BLS does this with the non-seasonally adjusted figures for the monthly employment estimates; I am assuming the changes in the seasonally adjusted figures will be similar).  That is, the April 2023 employment total was reduced by 68,167 (one-twelfth of 818,000), the May 2023 total by 136,333 (two-twelfths), and so on to March 2024.

Third, adjusting the figures going forward from March 2024 is more difficult as the BLS will use its updated models to make the revisions to the estimates from April.  Note that while the revised BLS estimates – when they are released as part of the January Employment Situation report – will cover the months through to December, all that we need now are estimates for the months of April, May, June, and July.

While very rough, for this I assumed the revisions for these four months will follow a pattern similar to what was found in the 2019 revision.  This was relatively recent but also pre-Covid (with all of the disruptions of patterns associated with that), and in that year the benchmark employment estimate was reduced by 0.3%.  While less than the 0.5% preliminary revision in the 2024 benchmark estimate, it was a still major revision downward (and during the Trump administration, although I do not recall ever seeing a reference by Trump to that reduction in the job totals).  I then used the month-by-month revisions in the seasonally adjusted employment estimates in 2019 for April through July, rescaled the percentage changes of each by the ratio of 0.5%/0.3% (in fact using the more precise figures of 0.517%/0.341%) and then applied those adjusted percentage changes to the current estimates of total employment in those four months.

The new path for total employment for the period of March 2023 to July 2024 is then shown as the orange line in the chart at the top of this post.  While below the current employment estimates (the line in red), the difference is not large.

The basic story remains the same.  Employment growth has been exceptionally strong under Biden, and has continued.  A downward revision in the benchmark total for March 2024 of 818,000 does not change this.