About aneconomicsense

Economist

Long-Term Structural Change in the US Economy: Manufacturing is Simply Following the Path of Agriculture

A.  Introduction

A major theme of Trump, both during his campaign and now as president, has been that jobs in manufacturing have been decimated as a direct consequence of the free trade agreements that started with NAFTA.  He repeated the assertion in his speech to Congress of February 28, where he complained that “we’ve lost more than one-fourth of our manufacturing jobs since NAFTA was approved”, but that because of him “Dying industries will come roaring back to life”.  He is confused.  But to be fair, there are those on the political left as well who are similarly confused.

All this reflects a sad lack of understanding of history.  Manufacturing jobs have indeed been declining in recent decades, and as the chart above shows, they have been declining as a share of total jobs in the economy since the 1940s.  Of all those employed, the share employed in manufacturing (including mining) fell by 7.6% points between 1994 (when NAFTA entered into effect) and 2015 (the most recent year in the sector data of the Bureau of Economic Analysis, used for consistency throughout this post), a period of 21 years. But the share employed in manufacturing fell by an even steeper 9.2% points in the 21 years before 1994.  The decline in manufacturing jobs (both as a share and in absolute number) is nothing new, and it is wrong to blame it on NAFTA.

It is also the case that manufacturing production has been growing steadily over this period.  Total manufacturing production (measured in real value-added terms) rose by 64% over the 21 years since NAFTA went into effect in 1994.  And this is also substantially higher than the 42% real growth in the 21 years prior to 1994.  Blaming NAFTA (and the other free trade agreements of recent decades) for a decline in manufacturing is absurd.  Manufacturing production has grown.

For those only interested in the assertion by Trump that NAFTA and the other free trade agreements have killed manufacturing in the US and with it the manufacturing jobs, one could stop here.  Manufacturing has actually grown strongly since NAFTA went into effect, and there are fewer manufacturing jobs now than before not because manufacturing has declined, but because workers in manufacturing are now more productive than ever before (with this a continuation of the pattern underway over at least the entire post-World War II period, and not something new).  But the full story is a bit more complex, as one also needs to examine why manufacturing production is at the level that it is.  For this, one needs to bring in the rest of the economy, in particular services. The rest of this blog post will address this broader issue,

Manufacturing jobs have nonetheless indeed declined.  To understand why, one needs to look at what has happened to productivity, not only in manufacturing but also in the other sectors of the economy (in particular in services).  And I would suggest that one could learn much by an examination of the similar factors behind the even steeper decline over the years in the share of jobs in agriculture.  It is not because of adverse effects of free trade.  The US is in fact the largest exporter of food products in the world.  Yet the share of workers employed in the agricultural sectors (including forestry and fishing) is now just 0.9% of the total.  It used to be higher:  4.3% in 1947 and 8.4% in 1929 (using the BEA data).  If one wants to go really far back, academics have estimated that agricultural employment accounted for 74% of all US employment in 1800, with this still at 56% in 1860.

Employment in agriculture has declined so much, from 74% of total employment in 1800 to 8.4% in 1929 to less than 1% today, because those employed in agriculture are far more productive today than they were before.  And while it leads to less employment in the sector, whether as a share of total employment or in absolute numbers, higher productivity is a good thing.  The US could hardly enjoy a modern standard of living if 74% of those employed still had to be working in agriculture in order to provide us food to eat. And while stretching the analysis back to 1800 is extreme, one can learn much by examining and understanding the factors behind the long-term trends in agricultural employment.  Manufacturing is following the same basic path.  And there is nothing wrong with that.  Indeed, that is exactly what one would hope for in order for the economy to grow and develop.

Furthermore, the effects of foreign trade on employment in the sectors, positive or negative, are minor compared to the long-term impacts of higher productivity.  In the post below we will look at what would have happened to employment if net trade would somehow be forced to zero by Trumpian policies.  The impact relative to the long term trends would be trivial.

This post will focus on the period since 1947, the earliest date for which the BEA has issued data on both sector outputs and employment.  The shares of agriculture as well as of manufacturing in both total employment and in output (with output measured in current prices) have both declined sharply over this period, but not because those sectors are producing less than before.  Indeed, their production in real terms are both far higher. Employment in those sectors has nevertheless declined in absolute numbers.  The reason is their high rates of productivity growth.  Importantly, productivity in those two sectors has grown at a faster pace than in the services sector (the rest of the economy).  As we will discuss, it is this differential rate of productivity growth (faster in agriculture and in manufacturing than in services) which explains the decline in the share employed in agriculture and manufacturing.

These structural changes, resulting ultimately from the differing rates of productivity growth in the sectors, can nonetheless be disruptive.  With fewer workers needed in a sector because of a high rate of productivity growth, while more workers are needed in those sectors where productivity is growing more slowly (although still positively and possibly strongly, just relatively less strongly), there is a need for workers to transfer from one sector to another.  This can be difficult, in particular for individuals who are older or who have fewer general skills.  But this was achieved before in the US as well as in other now-rich countries, as workers shifted out of agriculture and into manufacturing a century to two centuries ago.  Critically important was the development of the modern public school educational system, leading to almost universal education up through high school. The question the country faces now is whether the educational system can be similarly extended today to educate the workers needed for jobs in the modern services economy.

First, however, is the need to understand how the economy has reached the position it is now in, and the role of productivity growth in this.

B.  Sector Shares and Prices

As Chart 1 at the top of this post shows, employment in agriculture and in manufacturing have been falling steadily as a share of total employment since the 1940s, while jobs in services have risen.

[A note on the data:  The data here comes from the Bureau of Economic Analysis (BEA), which, as part of its National Income and Product Accounts (NIPA), estimates sector outputs as well as employment.  Employment is measured in full-time equivalent terms (so that two half-time workers, say, count as the equivalent of one full-time worker), which is important for measuring productivity growth.

And while the BEA provides figures on its web site for employment going all the way back to 1929, the figures for sector output on its web site only go back to 1947.  Thus while the chart at the top of this post goes back to 1929, all the analysis shown below will cover the period from 1947 only.  Note also that there is a break in the employment series in 1998, when the BEA redefined slightly how some of the detailed sectors would be categorized. They unfortunately did not then go back to re-do the categorizations in a consistent way in the years prior to that, but the changes are small enough not to matter greatly to this analysis.  And there were indeed similar breaks in the employment series in 1948 and again in 1987, but the changes there were so small (at the level of aggregation of the sectors used here) as not to be noticeable at all.

Also, for the purposes here the sector components of GDP have been aggregated to just three, with forestry and fishing included with agriculture, mining included with manufacturing, and construction included with services.  As a short hand, these sectors will at times be referred to simply as agriculture, manufacturing, and services.

Finally, the figures on sector outputs in real terms provided by the BEA data are calculated based on what are called “chain-weighted” indices of prices.  Chain-weighted indices are calculated based on moving shares of sector outputs (whatever the share is in any given period) rather than on fixed shares (i.e. the shares at the beginning or the end of the time period examined).  Chain-weighted indices are the best to use over extended periods, but are unfortunately not additive, where a sum (such as real GDP) will not necessarily equal exactly the sum of the estimates of the underlying sector figures (in real terms).  The issue is however not an important one for the questions being examined in this post.  While we will show the estimates in the charts for real GDP (based on a sum of the figures for the three sectors), there is no need to focus on it in the analysis.  Now back to the main text.]

The pattern in a chart of sector outputs as shares of GDP (measured in current prices by the value-added of each sector), is similar to that seen in Chart 1 above for the employment shares:

Agriculture is falling, and falling to an extremely small share of GDP (to less than 1% of GDP in 2015).  Manufacturing and mining is similarly falling from the mid-1950s, while services and construction is rising more or less steadily.  On the surface, all this appears to be similar to what was seen in Chart 1 for employment shares.  It also might look like the employment shares are simply following the shifts in output shares.

But there is a critical difference.  The shares of workers employed is a measure of numbers of workers (in full-time equivalent terms) as a share of the total.  That is, it is a measure in real terms.  But the shares of sector outputs in Chart 2 above is a measure of the shares in terms of current prices.  They do not tell us what is happening to sector outputs in real terms.

For sector outputs in real terms (based on the prices in the initial year, or 1947 here), one finds a very different chart:

Here, the output shares are not changing all that much.  There is only a small decline in agriculture (from 8% of the total in 1947 to 7% in 2015), some in manufacturing (from 28% to 22%), and then the mirror image of this in services (from 64% to 72%).  The changes in the shares were much greater in Chart 2 above for sector output shares in current prices.

Many might find the relatively modest shifts in the shares of sector outputs when measured in constant price terms to be surprising.  We were all taught in our introductory Economics 101 class of Engel Curve effects.  Ernst Engel was a German statistician who, in 1857, found that at the level of households, the share of expenditures on basic nourishment (food) fell the richer the household.  Poorer households spent a relatively higher share of their income on food, while better off households spent less.  One might then postulate that as a nation becomes richer, it will see a lower share of expenditures on food items, and hence that the share of agriculture will decline.

But there are several problems with this theory.  First, for various reasons it may not apply to changes over time as general income levels rise (including that consumption patterns might be driven mostly by what one observes other households to be consuming at the time; i.e. “keeping up with the Joneses” dominates).  Second, agricultural production spans a wide range of goods, from basic foodstuffs to luxury items such as steak.  The Engel Curve effects might mostly be appearing in the mix of food items purchased.

Third, and perhaps most importantly, the Engel Curve effects, if they exist, would affect production only in a closed economy where it was not possible to export or import agricultural items.  But one can in fact trade such agricultural goods internationally. Hence, even if domestic demand fell over time (due perhaps to Engel Curve effects, or for whatever reason), domestic producers could shift to exporting a higher share of their production.  There is therefore no basis for a presumption that the share of agricultural production in total output, in real terms, should be expected to fall over time due to demand effects.

The same holds true for manufacturing and mining.  Their production can be traded internationally as well.

If the shares of agriculture and manufacturing fell sharply over time in terms of current prices, but not in terms of constant prices (with services then the mirror image), the implication is that the relative prices of agriculture as well as manufacturing fell relative to the price of services.  This is indeed precisely what one sees:

These are the changes in the price indices published by the BEA, with all set to 1947 = 1.0.  Compared to the others, the change in agricultural prices over this 68 year period is relatively small.  The price of manufacturing and mining production rose by far more.  And while a significant part of this was due to the rise in the 1970s of the prices of mined products (in particular oil, with the two oil crises of the period, but also in the prices of coal and other mined commodities), it still holds true for manufacturing alone.  Even if one excludes the mining component, the price index rose by far more than that of agriculture.

But far greater was the change in the price of services.  It rose to an index value of 12.5 in 2015, versus an index value of just 1.6 for agriculture in that year.  And the price of services rose by double what the price of manufacturing and mining rose by (and even more for manufacturing alone).

With the price of services rising relative to the others, the share of services in GDP (in current prices) will then rise, and substantially so given the extent of the increase in its relative price, despite the modest change in its share in constant price terms.  Similarly, the fall in the shares of agriculture and of manufacturing (in current price terms) will follow directly from the fall in their prices (relative to the price of services), despite just a modest reduction in their shares in real terms.

The question then is why have we seen such a change in relative prices.  And this is where productivity enters.

C.  Growth in Output, Employment, and Productivity

First, it is useful to look at what happened to the growth in real sector outputs relative to 1947:

All sector outputs rose, and by substantial amounts.  While Trump has asserted that manufacturing is dying (due to free trade treaties), this is not the case at all.  Manufacturing (including mining) is now producing 5.3 times (in real terms) what it was producing in 1947.  Furthermore, manufacturing production was 64% higher in real terms in 2015 than it was in 1994, the year NAFTA went into effect.  This is far from a collapse.  The 64% increase over the 21 years between 1994 and 2015 was also higher than the 42% increase in manufacturing production of the preceding 21 year period of 1973 to 1994. There was of course much more going on than any free trade treaties, but to blame free trade treaties on a collapse in manufacturing is absurd.  There was no collapse.

Production in agriculture also rose, and while there was greater volatility (as one would expect due to the importance of weather), the increase in real output over the full period was in fact very similar to the increase seen for manufacturing.

But the biggest increase was for services.  Production of services was 7.6 times higher in 2015 than in 1947.

The second step is to look at employment, with workers measured here in full-time equivalent terms:

Despite the large increases in sector production over this period, employment in agriculture fell as did employment in manufacturing.  One unfortunately cannot say with precision by how much, given the break in the employment series in 1998.  However, there were drops in the absolute numbers employed in manufacturing both before and after the 1998 break in the series, while in agriculture there was a fall before 1998 (relative to 1947) and a fairly flat series after.  The change in the agriculture employment numbers in 1998 was relatively large for the sector, but since agricultural employment was such a small share of the total (only 1%), this does not make a big difference overall.

In contrast to the falls seen for agriculture and manufacturing, employment in the services sector grew substantially.  This is where the new jobs are arising, and this has been true for decades.  Indeed, services accounted for more than 100% of the new jobs over the period.

But one cannot attribute the decline in employment in agriculture and in manufacturing to the effects of international trade.  The points marked with a “+” in Chart 6 show what employment in the sectors would have been in 2015 (relative to 1947) if one had somehow forced net imports in the sectors to zero in 2015, with productivity remaining the same. There would have been an essentially zero change for agriculture (while the US is the world’s largest food exporter, it also imports a lot, including items like bananas which would be pretty stupid to try to produce here).  There would have been somewhat more of an impact on manufacturing, although employment in the sector would still have been well below what it had been decades ago.  And employment in services would have been a bit less. While most production in the services sector cannot be traded internationally, the sector includes businesses such as banking and other finance, movie making, professional services, and other areas where the US is in fact a strong exporter.  Overall, the US is a net exporter of services, and an abandonment of trade that forced all net imports (and hence net exports) to zero would lead to less employment in the sector.  But the impact would be relatively minor.

Labor productivity is then simply production per unit of labor.  Dividing one by the other leads to the following chart:

Productivity in agriculture grew at a strong pace, and by more than in either of the other two sectors over the period.  With higher productivity per worker, fewer workers will be needed to produce a given level of output.  Hence one can find that employment in agriculture declined over the decades, even though agricultural production rose strongly. Productivity in manufacturing similarly grew strongly, although not as strongly as in agriculture.

In contrast, productivity in the services sector grew at only a modest pace.  Most of the activities in services (including construction) are relatively labor intensive, and it is difficult to substitute machinery and new technology for the core work that they do.  Hence it is not surprising to find a slower pace of productivity growth in services.  But productivity in services still grew, at a positive 0.9% annual pace over the 1947 to 2015 period, as compared to a 2.8% annual pace for manufacturing and a 3.3% annual pace in agriculture.

Finally, and for those readers more technically inclined, one can convert this chart of productivity growth onto a logarithmic scale.  As some may recall from their high school math, a straight line path on a logarithmic scale implies a constant rate of growth.  One finds:

While one should not claim too much due to the break in the series in 1998, the path for productivity in agriculture on a logarithmic scale is remarkably flat over the full period (once one abstracts from the substantial year to year variation – short term fluctuations that one would expect from dependence on weather conditions).  That is, the chart indicates that productivity in agriculture grew at a similar pace in the early decades of the period, in the middle decades, and in the later decades.

In contrast, it appears that productivity in manufacturing grew at a certain pace in the early decades up to the early 1970s, that it then leveled off for about a decade until the early 1980s, and that it then moved to a rate of growth that was faster than it had been in the first few decades.  Furthermore, the pace of productivity growth in manufacturing following this turn in the early 1980s was then broadly similar to the pace seen in agriculture in this period (the paths are then parallel so the slope is the same).  The causes of the acceleration in the 1980s would require an analysis beyond the scope of this blog post. But it is likely that the corporate restructuring that became widespread in the 1980s would be a factor.  Some would also attribute the acceleration in productivity growth to the policies of the Reagan administration in those years.  However, one would also then need to note that the pace of productivity growth was similar in the 1990s, during the years of the Clinton administration, when conservatives complained that Clinton introduced regulations that undid many of the changes launched under Reagan.

Finally, and as noted before, the pace of productivity growth in services was substantially less than in the other sectors.  From the chart in logarithms, it appears the pace of productivity growth was relatively robust in the initial years, up to the mid-1960s.  While slower than the pace in manufacturing or in agriculture, it was not that much slower.  But from the mid-1960s, the pace of growth of productivity in services fell to a slower, albeit still positive, pace.  Furthermore, that pace appears to have been relatively steady since then.

One can summarize the results of this section with the following table:

Growth Rates:

1947 to 2015

Employment

Productivity

Output

Total (GDP)

1.5%

1.4%

2.9%

Agriculture

-0.7%

3.3%

2.6%

Manufacturing

-0.3%

2.8%

2.5%

Services

2.1%

0.9%

3.0%

The growth rate of output will be the simple sum of the growth rate of employment in a sector and the growth rate of its productivity (output per worker).  The figures here do indeed add up as they should.  They do not tell us what causes what, however, and that will be addressed next.

D.  Pulling It Together:  The Impact on Employment, Prices, and Sector Shares

Productivity is driven primarily by technological change.  While management skills and a willingness to invest to take advantage of what new technologies permit will matter over shorter periods, over the long term the primary driver will be technology.

And as seen in the chart above, technological progress, and the resulting growth in productivity, has proceeded at a different pace in the different sectors.  Productivity (real output per worker) has grown fastest over the last 68 years in agriculture (a pace of 3.3% a year), and fast as well in manufacturing (2.8% a year).  In contrast, the rate of growth of productivity in services, while positive, has been relatively modest (0.9% a year).

But as average incomes have grown, there has been an increased domestic demand in what the services sector produces, not only in absolute level but also as a share of rising incomes.  Since services largely cannot be traded internationally (with a few exceptions), the increased demand for services will need to be met by domestic production.  With overall production (GDP) matching overall incomes, and with demand for services growing faster than overall incomes, the growth of services (in real terms) will be greater than the growth of real GDP, and therefore also greater than growth in the rest of the economy (agriculture and manufacturing; see Chart 5).  The share of services in real GDP will then rise (Chart 3).

To produce this, the services sector needed more labor.  With productivity in the services sector growing at a slower pace (in relative terms) than that seen in agriculture and in manufacturing, the only way to obtain the labor input needed was to increase the share of workers in the economy employed in services (Chart 1).  And depending on the overall rate of labor growth as well as the size of the differences in the rates of productivity growth between the sectors, one could indeed find that the shift in workers out of agriculture and out of manufacturing would not only lead to a lower relative share of workers in those sectors, but also even to a lower absolute number of workers in those sectors.  And this is indeed precisely what happened, with the absolute number of workers in agriculture falling throughout the period, and falling in manufacturing since the late 1970s (Chart 6).

Finally, the differential rates of productivity growth account for the relative price changes seen between the sectors.  To be able to hire additional workers into services and out of agriculture and out of manufacturing, despite a lower rate of productivity growth in services, the price of services had to rise relative to agriculture as well as manufacturing. Services became more expensive to produce relative to the costs of agriculture or manufacturing production.  And this is precisely what is seen in Chart 4 above on prices.

To summarize, productivity growth allowed all sectors to grow.  With the higher incomes, there was a shift in demand towards services, which led it to grow at a faster pace than overall incomes (GDP).  But for this to be possible, particularly as its pace of productivity growth was slower than the pace in agriculture and in manufacturing, workers had to shift to services from the other sectors.  The effect was so great (due to the differing rates of growth of productivity) that employment in services rose to the point where services now employs close to 90% of all workers.

To be able to hire those workers, the price of services had to grow relative to the prices of the other sectors.  As a consequence, while there was only a modest shift in sector shares over time when measured in real terms (constant prices of 1947), there was a much larger shift in sector shares when measured in current prices.

The decline in the number of workers in manufacturing should not then be seen as surprising nor as a reflection of some defective policy.  Nor was it a consequence of free trade agreements.  Rather, it was the outcome one should expect from the relatively rapid pace of productivity growth in manufacturing, coupled with an economy that has grown over the decades with this leading to a shift in domestic demand towards services.  The resulting path for manufacturing was then the same basic path as had been followed by agriculture, although it has been underway longer in agriculture.  As a result, fewer than 1% of American workers are now employed in agriculture, with this possible because American agriculture is so highly productive.  One should expect, and indeed hope, that the same eventually becomes true for manufacturing as well.

Delusional: Is This What We Are to Expect from the New Trump Administration?

Definition of delusional in English:

delusional

ADJECTIVE

Characterized by or holding idiosyncratic beliefs or impressions that are contradicted by reality or rational argument, typically as a symptom of mental disorder:

‘hospitalization for schizophrenia and delusional paranoia’

‘he was diagnosed with a delusional disorder’

 Based on or having faulty judgement; mistaken:

‘their delusional belief in the project’s merits never wavers’

‘I think the guy is being a bit delusional here’

 

Donald J. Trump was inaugurated as President of the United States at 12:00 noon on January 20.  A day later, his new White House Press Secretary and Communications Director Sean Spicer in his very first press briefing of the new administration, launched a tirade against the press, for reporting (falsely he claimed) that attendance at the inauguration was less than the number who had attended Obama’s inauguration in 2009 (or indeed any prior inauguration). And he was visibly angry about this, as can be seen both in the transcript of the press briefing, and in a video of it.  He charged that “some members of the media were engaged in deliberately false reporting” and claimed that “This was the largest audience to ever witness an inauguration — period — both in person and around the globe.”

Furthermore, after many reports challenged Spicer’s assertions, the new administration doubled down on the charges.  Reince Priebus, the new White House Chief of Staff, vowed on Sunday that the new administration will fight the media “tooth and nail every day and twice on Sunday” over what they see as unfair attacks on Trump (by claiming, falsely they say, that the crowds had been larger at Obama’s inauguration).  And Kellyanne Conway, a spokesman for the White House and Counselor to the President, said on Sunday that what Press Secretary Spicer had asserted was not wrong but rather “alternative facts”.

Finally, one has Donald Trump himself, who claimed that he saw what “looked like a million, a million and a half people” present at his inauguration as he took the oath of office. One does not know how he was able to make such a count, and perhaps he should not be taken too seriously, but his administration’s senior staff appear to be obliged to back him up.

What do we know on the size of the crowds?  One first has to acknowledge that any crowd count is difficult, and that we will never know the precise numbers.  Unless each person has been forced to pass through a turnstile, all we can have are estimates.  But we can have estimates, and they can give some sense as to the size.  Most importantly, while we might not know the absolute size, we can have a pretty good indication from photos and other sources of data what the relative sizes of two crowds likely were.

So what do we know from photos?  Here we have a side-by-side photo (taken at Obama’s first inauguration and then at Trump’s) from the top of the Washington Monument, of the crowd on the Mall witnessing the event.  They were both taken at about the same time prior to the noon swearing-in of the new president, where the ceremony starts at 11:30:


inaugeration-attendance-2017-vs-2009

 

 

 

 

 

 

 

The crowd in 2017 is clearly far smaller.  This has nothing to do with the white mats laid down to protect the grass (which was also done in 2013 for Obama’s second inauguration).  There are simply far fewer attendees.

There is also indirect evidence from the number of Metrorail riders that day.  Spicer said in his press briefing “We know that 420,000 people used the D.C. Metro public transit yesterday, which actually compares to 317,000 that used it for President Obama’s last inaugural.”  Actually these numbers are wrong, as well as misleading (since the comparison at issue is to Obama’s first inauguration in 2009, not to his second in 2013). As the Washington Post noted (with this confirmed by CNN) the correct numbers from the Washington Metropolitan Area Transit Authority (which operates the Metro system) are that there were 570,500 riders on Metro on Trump’s inauguration day, 1.1 million riders in 2009 on Obama’s first inauguration day, and 782,000 riders in 2013 on Obama’s second inauguration day.  What Trump’s press secretary said “we know” was simply wrong.

It is also simply not true that Trump drew a larger estimated TV audience than any president before.  Nielsen, the TV ratings agency, estimated that Trump drew 30.6 million viewers, while Obama drew 38 million viewers at his first inauguration.  And Reagan drew more, at 42 million viewers, for his first inauguration.  Furthermore, both Nixon (in 1973) and Carter (in 1977) drew more viewers than Trump, at 33 million and 34 million respectively. The Trump figure was far from a record.

So how many people attended Trump’s inauguration, and how does that figure compare to the number that Obama drew for his first inauguration?  A widely cited figure is that Obama drew an estimated 1.8 million for his first inauguration, but, as noted above, any such estimate must be taken as approximate.  But based on a comparison of the photos, experts estimate that Trump drew at most one-third of the Obama draw in terms of the number in attendance just on the Mall.  There were in addition many others at the Obama inaugural who were not on the Mall because they could not fit due to the crowding.

Why does this matter?  It matters only because the new Trump administration has made it into an issue, and in doing so, has made assertions that are clearly factually wrong.  Trump did not draw a record number to his inauguration, nor a record number of viewers, nor were there a record number of riders on the Washington Metro system.  These are all numbers, and they can be checked.  While we may not be able to know the precise number of those who attended, we can come to a clear conclusion on the relative size of those who attended this year versus previous recent inaugurations.  And Trump’s attendance was not at all close to the number who attended Obama’s first inaugural.

What is disconcerting is that Trump, his new Press Secretary, his Chief of Staff and others in his administration, should feel compelled to make assertions that are clearly and verifiably wrong, and then to attack the press aggressively for pointing out what we know. And this on his second day in office.  While this is not inconsistent with what the Trump team did during his campaign for the presidency, one would have hoped for more mature behavior once he took office.  And especially so for an issue which is fundamentally minor. It really does not matter much whether the number attending Trump’s inauguration was more or less than the number who had attended prior inaugurations.

Presumably (and assuming thought was given to this) they are setting a marker for what they intend to do during the course of the presidential term, with aggressive attacks on the press for reporting errors in their assertions or on contradictions with earlier statements.  If so, such a strategy, including denial of facts that can readily be verified, is truly worrisome. Facts should matter.  Not all that we will hear from the new administration will be so easy to check, and the question then is what can be believed.

Perhaps, and more worrying, they really believe their assertions on the numbers attending. If so, they are truly delusional.

Tax Cuts Do Not Spur Growth – There Are Income as well as Substitution Effects, and Much More Besides: Econ 101

gdp-growth-and-top-marg-tax-rate-1930-to-2015

A.   Introduction, and a Brief Aside on the Macro Issues

While there is much we do not yet know on what economic policies Donald Trump will pursue (he said many things in his campaign, but they were often contradictory), one thing we can be sure of is that there will be a major tax cut.  Republicans in Congress (led by Paul Ryan) and in the Senator want the same.  And they along with Trump insist that the cuts in tax rates will spur a sharp jump in GDP growth, with the result that net tax revenues in the end will not fall by all that much.

But do tax cuts spur growth?  The chart above suggests not.  Marginal tax rates of those in the top income brackets have come down sharply since the 1950s and early 1960s, when they exceeded 90%.  They reached as low as 28% during the later Reagan years and 35% during the administration of George W. Bush.  But GDP growth did not jump to some higher rate as a result.

This Econ 101 post will discuss the economics on why this is actually what one should expect.  It will focus on the microeconomics behind this, as the case for income tax cuts is normally presented by the so-called “supply siders” as a micro story of incentives.  The macro case for tax cuts is different.  Briefly, in times of high unemployment when the economy is suffering from insufficient demand in the aggregate to purchase all that could be produced if more labor were employed, a cut in income taxes might spur demand by households, as they would then have higher post-tax incomes to spend on consumption items.  This increase in demand could then spur production and hence GDP.

Critically, this macro story depends on allowing the fiscal deficit to rise by there not being simultaneously a cut in government expenditures along with the tax cuts.  If there is such a cut in government expenditures, demand may be reduced by as much as or even more than demand would be increased by households.  But the economic plans of both Trump and Congressman (and Speaker) Paul Ryan do also call for large cuts in government expenditures.  While both Trump and Ryan have called for government expenditures to increase on certain items, such as for defense, they still want a net overall reduction.

The net impact on demand will then depend on how large the government expenditure cuts would be relative to the tax cuts, and on the design of the income tax cuts.  As was discussed in an earlier post on this blog on the size of the fiscal multiplier, If most of the income tax cuts go to those who are relatively well off, who will then save most or perhaps all of their tax windfall, there will be little or no macro stimulus from the tax cuts.  Any government expenditure cuts on top of this would then lead not to a spur in growth, but rather to output growing more slowly or contracting.  And the tax plan offered by Donald Trump in his campaign would indeed direct the bulk of the tax cuts to the extremely well off.  A careful analysis by the non-partisan Tax Policy Center found that 71% of the tax cuts (in dollar value) from the overall plan (which includes cuts in corporate and other taxes as well) would go to the richest 5% of households (those earning $299,500 or more), 51% would go to the top 1% (those earning $774,300 or more), and fully 25% would go to the richest 0.1% (those earning $4.8 million or more).

[A side note:  To give some perspective on how large these tax cuts for the rich would be, the 25% going to the richest 0.1% under Trump’s plan would total $1.5 trillion over the next ten years, under the Tax Policy Center estimates.  By comparison, the total that the Congressional Budget Office projects would be spent on the food stamp program (now officially called SNAP) for the poor over this period would come to a bit below $700 billion (see the August 2016 CBO 10-year budget projections).  That is, the tax breaks to be given under Trump’s tax plan to the top 0.1% (who have earnings of $4.8 million or more in a year) would be more than twice as large as would be spent on the entire food stamp program over the period.  Yet the Republican position is that we have to cut the food stamp program because we do not have sufficient government revenues to support it.]

The macro consequences of tax cuts that mostly go to the already well off, accompanied by government expenditure cuts to try to offset the deficit impact, are likely therefore to lead not to a spur in growth but to the opposite.

The microeconomic story is separate, and the rest of this blog post will focus on the arguments there.  Those who argue that cuts in income taxes will act as a spur to growth base their argument on what they see as the incentive effects.  Income taxes are a tax on working, they argue, and if you tax income less, people will work longer hours.  More will be produced, the economy will grow faster, and people will have higher incomes.

This micro argument is mistaken in numerous ways, however.  This Econ 101 post will discuss why.  There is the textbook economics, where it appears these “supply siders” forgot some of the basic economics they were taught in their introductory micro courses. But we should also recognize that the decision on how many hours to work each week goes beyond simply the economics.  There are important common social practices (which can vary by the nature of the job, i.e. what is a normal work day, and what do you do to get promoted) and institutional structures (the 40 hour work week) which play an important and I suspect dominant role. This blog post will review some of them.

But first, what do we know from the data, and what does standard textbook economics say?

B.  Start with the Data

It is always good first to look at what the data is telling us.  There have been many sharp cuts in income tax rates over the last several decades, and also some increases.  Did the economy grow faster after the tax cuts, and slower following the tax increases?

The chart at the top of this post indicates not.  The chart shows what GDP growth was year by year since 1930 along with the top marginal income tax rate of each year.  The top marginal income tax rate is the rate of tax that would be paid on an additional dollar of income by those in the highest income tax bracket.  The top marginal income tax rate is taken by those favoring tax cuts as the most important tax rate to focus on.  It is paid by the richest, and these individuals are seen as the “job creators” and hence play an especially important role under this point of view.  But changes in the top rates also mark the times when there were normally more general tax cuts for the rest of the population as well, as cuts (or increases) in the top marginal rates were generally accompanied by cuts (or increases) in the other rates also.  It can thus be taken as a good indicator of when tax rates changed and in what direction.  Note also that the chart combines on one scale the annual GDP percentage growth rates and the marginal tax rate as a percentage of an extra dollar of income, which are two different percentage concepts.  But the point is to compare the two.

As the chart shows, the top marginal income tax rate exceeded 90% in the 1950s and early 1960s.  The top rate then came down sharply, to generally 70% until the Reagan tax cuts of the early 1980s, when they fell to 50% and ultimately to just 28%.  They then rose under Clinton to almost 40%, fell under the Bush II tax cuts to 35%, and then returned under Obama to the rate of almost 40%.

Were GDP growth rates faster in the periods when the marginal tax rates were lower, and slower when the tax rates were higher?  One cannot see any indication of it in the chart. Indeed, even though the highest marginal tax rates are now far below what they were in the 1950s and early 1960s, GDP growth over the last decade and a half has been less than it what was when tax rates were not just a little bit, but much much higher.  If cuts in the marginal tax rates are supposed to spur growth, one would have expected to see a significant increase in growth between when the top rate exceeded 90% and where it is now at about 40%.

Indeed, while I would not argue that higher tax rates necessarily lead to faster growth, the data do in fact show higher tax rates being positively correlated with faster growth.  That is, the economy grew faster in years when the tax rates were higher, not lower.  A simple statistical regression of the GDP growth rate on the top marginal income tax rate of the year found that if the top marginal tax rate were 10% points higher, GDP growth was 0.57% points higher.  Furthermore, the t-statistic (of 2.48) indicates that the correlation was statistically significant.

Again, I would not argue that higher tax rates lead to faster GDP growth.  Rather, much more was going on with the economy over this period which likely explains the correlation. But the data do indicate that very high top marginal income tax rates, even over 90%, were not a hindrance to growth.  And there is clearly no support in the evidence that lower tax rates lead to faster growth.

The chart above focuses on the long-term impacts, and does not find any indication that tax cuts have led to faster growth.  An earlier post on this blog looked at the more immediate impacts of such tax rates cuts or increases, focussing on the impacts over the next several years following major tax rate changes.  It compared what happened to output and employment (as well as what happened to tax revenues and to the fiscal deficit) in the immediate years following the Reagan and Bush II tax cuts, and following the Clinton and Obama tax increases.  What it found was that growth in output and employment, and in fiscal revenues, were faster following the Clinton and Obama tax increases than following the Reagan and Bush II tax cuts.  And not surprisingly given this, the fiscal deficit got worse under Reagan and Bush II following their tax cuts, and improved following the Clinton and Obama tax increases.

C.  The Economics of the Impact of Tax Rates on Work Effort

The “supply siders” who argue that cuts in income taxes will lead to faster growth base their case on what might seem (at least to them) simple common sense.  They say that if you tax something, you will produce less of it.  Tax it less, and you will produce more of it. And they say this applies to work effort.  Income taxes are a tax on work.  Lower income tax rates will then lead to greater work effort, they argue, and hence to more production and hence to more growth.  GDP growth rates will rise.

But this is wrong, at several levels.  One can start with some simple math.  The argument confuses what would be (by their argument) a one-time step-up in production, with an increase in growth rates.  Suppose that tax rates are cut and that as a result, everyone decides that at the new tax rates they will choose to work 42 hours a week rather than 40 hours a week before.  Assuming productivity is unchanged (actually it would likely fall a bit), this would lead to a 5% increase in production.  But this would be a one time increase. GDP would jump 5% in the first year, but would then grow at the same rate as it had before.  There would be no permanent increase in the rate of growth, as the supply siders assert.  This is just simple high school math.  A one time increase is not the same as a permanent increase in the rate of growth.

But even leaving this aside, the supply sider argument ignores some basic economics taught in introductory microeconomics classes.  Focussing just on the economics, what would be expected to happen if marginal income tax rates are cut?  It is true that there will be what economists call “substitution effects”, where workers may well wish to work longer hours if their after-tax income from work rises due to a cut in marginal tax rates. But the changes will also be accompanied by what economists call “income effects”.  Worker after-tax incomes will change both because of the tax rate changes and because of any differences in the hours they work.  And these income effects will lead workers to want to work fewer hours.  The income and substitution effects will work in opposite directions, and the net impact of the two is not clear.  They could cancel each other out.

What are the income effects, and why would they lead to less of an incentive to work greater hours if the tax rate falls?:

a)  First, one must keep in mind that the aim of working is to earn an income, and that hours spent working has a cost:  One will have fewer hours at home each day to enjoy with your wife and kids, or for whatever other purposes you spend your non-working time. Economists lump this all under what they call “leisure”.  Leisure is something desirable, and with all else equal, one would prefer more of it.  Economists call this a “normal good”.  With a higher income, you would want to buy more of it. And the way you buy more of it is by working fewer hours each day (at the cost of giving up the wages you would earn in those hours).

Hence, if taxes on income go down, so that your after-tax income at the original number of hours you work each day goes up, you will want to use at least some portion of this extra income to buy more time to spend at home.  This is an income effect, and will go in the opposite direction of the substitution effect of higher after-tax wages leading to an incentive to work longer hours.  We cannot say, a priori, whether the income effect or the substitution effect will dominate.  It will vary by individual, based on their individual preferences, what their incomes are, and how many hours they were already working.  It could go either way, and can only be addressed by looking at the data.

b)  One should also recognize that one works to earn income for a reason, and one reason among many is to earn and save enough so that one can enjoy a comfortable retirement. But in standard economic theory, there is no reason to work obsessively before retirement so that one will then have such a large retirement “nest egg” as to enjoy a luxurious life style when one retires.  Rather, the aim is to smooth out your consumption profile over both periods in your life.

Hence if income tax rates are cut, so that your after-tax incomes are higher, one will be able to save whatever one is aiming for for retirement, sooner.  Hence it would be rational to reduce by some amount the hours one seeks to work each day, and enjoy them with your wife and kids at home, as your savings goals for retirement can still be met with those fewer hours of work.  This is an income effect, and acts in the direction of reducing, rather than increasing, the number of hours one will choose to work if there is a general tax cut.

c)  More generally, one should recognize that incomes are earned to achieve various aims. Some of these might be to cover fixed obligations, such as to pay on a mortgage or for student debt, and some might be quasi-fixed, such as to provide for a “comfortable” living standard for one’s family.  If those aims are being met, then time spent at leisure (time spent at home with the family) may be especially attractive.  In such circumstances, the income effect from tax cuts might be especially large, and sufficient to more than offset the substitution effects resulting from the change in the after-tax wage.

Income effects are real, and it is mistake to ignore them.  They act in the opposite direction of the substitution effect, and will act to offset them.  The offset might be partial, full, or even more than full.  We cannot say simply by looking at the theory.  Rather, one needs to look at the data.  And as noted above, the data provdes no support to the suppostion that lower tax rates will lead to higher growth.  Once one recognizes that there will be income effects as well as substitution effects, one can see that this should not be a surprise.  It is fully consistent with the theory.

One can also show how the income and substitution effects work via some standard diagrams, involving indifference curves and budget constraints.  These are used in most standard economics textbooks.  However, I suspect that most readers will find such diagrams to be more confusing than enlightening.  A verbal description, such as that above, will likely be more easy to follow.  But for those who prefer such diagrams, the standard ones can be found at this web posting.  Note, however, that there is a mistake (a typo I assume) in the key Figures 2A and 2B.  The horizontal arrows (along the “leisure” axis) are pointed in the opposite direction of what they should (left instead of right in 2A and right instead of left in 2B).  These errors indeed serve to emphasize how even the experts with such diagrams can get confused and miss simple typos.

D.  But There is More to the Hours of Work Decision than Textbook Economics

The analysis above shows that the supply-siders, who stress microeconomic incentives as key, have forgotten half of the basic analysis taught in their introductory microeconomics classes.  There are substitution effects resulting from a change in income tax rates, as the supply-siders argue, but there are also income effects which act in the opposite direction. The net effect is then not clear.

However, there is more to the working hours decision than the simple economics of income and substitution effects.  There are social as well as institutional factors.  It the real world, these other factors matter.  And I suspect they matter a good deal more than the standard economics in explaining the observation that we do not see growth rates jumping upwards after the several rounds of major tax cuts of the last half century.

Such factors include the following:

a)  For most jobs, a 40 hour work week is, at least formally, standard.  For those earning hourly wages, any overtime above 40 hours is, by law, supposed to be compensated at 50% above their normal hourly wage.  For workers in such jobs, one cannot generally go to your boss and tell him, in the event of an income tax increase say, that you now want to work only 39 1/2 hours each week.  The hours are pretty much set for such workers.

b)  There are of course other workers compensated by the hour who might work a variable number of hours each week at a job.  These normally total well less than 40 hours a week.  These would include many low wage occupations such as at fast food places, coffee shops, retail outlets, and similarly.  But for many such workers, the number of hours they work each week is constrained not by the number of hours they want to work, but by the number of hours their employer will call them in for.  A lower income tax rate might lead them to want to work even more hours, but when they are constrained already by the number of hours their employer will call them in for, there will be no change.

c)  For salaried workers and professionals such as doctors, the number of hours they work each week is defined primarily by custom for their particular profession.  They work the hours that others in that profession work, with this evolving over time for the profession as a whole.  The hours worked are in general not determined by some individual negotiation between the professional and his or her supervisor, with this changing when income tax rates are changed.  And many professionals indeed already work long hours (including medical doctors, where I worry whether they suffer from sleep deprivation given their often incredibly long hours).

d)  The reason why one sees many professionals, including managers and others in office jobs, working such long hours probably has little to do with marginal income tax rates.  Rather, they try to work longer than their co-workers, or at least not less, in order to get promoted.  Promotion is a competition, where the individual seen as the best is the one who gets promoted.  And the one seen as the best is often the one who works the longest each day.  With the workers competing against each other, possibly only implicitly and not overtly recognized as such, there will be an upward spiral in the hours worked as each tries to out-do the other.  This is ultimately constrained by social norms.  Higher or lower income tax rates are not central here.

e)  Finally, and not least, most of us do take pride in our work.  We want to do it well, and this requires a certain amount of work effort.  Taxes are not the central determinant in this.

E.  Summary and Conclusion

I fully expect there to be a push to cut income tax rates early in the Trump presidency.  The tax plan Trump set out during his campaign was similar to that proposed by House Speaker Paul Ryan, and both would cut rates sharply, especially for those who are already well off. They will argue that the cuts in tax rates will spur growth in GDP, and that as a consequence, the fiscal deficit will not increase much if at all.

There is, however, no evidence in the historical data that this will be the case.  Income tax rates have been cut sharply since the Eisenhower years, when the top marginal income tax rate topped 90%, but growth rates did not jump higher following the successive rounds of cuts.

Tax cuts, if they are focused on those of lower to middle income, might serve as a macro stimulus if unemployment is significant.  Such households would be likely to spend their extra income on consumption items rather than save it, and this extra household consumption demand can serve to spur production.  But tax cuts that go primarily to the rich (as the tax cuts that have been proposed by Trump and Ryan would do), that are also accompanied by significant government expenditure cuts, will likely have a depressive rather than stimulative effect.

The supply-siders base their argument, however, for why tax cuts should lead to an increase in the growth rate of GDP, not on the macro effects but rather on what they believe will be the impact on microeconomic incentives.  They argue that income taxes are a tax on work, and a reduction in the tax on work will lead to greater work effort.

They are, however, confused.  What they describe is what economists call the substitution effect.  That may well exist.  But there are also income effects resulting from the changes in the tax rates, and these income effects will work in the opposite direction.  The net impact is not clear, even if one keeps just to standard microeconomics.  The net impact could be a wash.  Indeed, the net impact could even be negative, leading to fewer hours worked when there is a cut in income taxes.  One does not know a priori, and you need to look at the data.  And there is no indication in the data that the sharp cuts in marginal tax rates over the last half century have led to higher rates of growth.

There is also more to the working hours decision than just textbook microeconomics. There are important social and institutional factors, which I suspect will dominate.  And they do not depend on the marginal rates of income taxes.

But if you are making an economic argument, you should at least get the economics right.

Why It Is Important to Vote This November 8

trump-and-clinton-picture2-001

There is little need to repeat here the many reasons why the election of Donald Trump to the presidency (or indeed any position of authority) would be a disaster.  He has eminently disqualified himself by both his words and his actions, and I have little to add.  And there are many reasons why Hillary Clinton should be elected, not simply in order not to elect Trump.  Even her critics admit that she has the background and experience in both the executive and legislative branches of government – as First Lady (with an active role in policy discussions), as Senator from New York, and as Secretary of State for four years under President Obama –  that few candidates for the presidency could match.

Furthermore, even Donald Trump has said she is a fighter, and that is precisely what is needed if the policies that Obama has championed are to continue to move forward.  A Clinton administration will fight for action to address global warming, to moderate health care costs, to improve education, to reform immigration, to re-build our infrastructure, and more, just in the domain of domestic policy.  A Trump administration would move us backwards on each.  And I far prefer an administration that takes pride in making decisions based on what is in their head, as Obama has done, rather than based on what is in their gut, as Obama’s predecessor was proud to brag of.

As I write this, the polls indicate Hillary Clinton holds a substantial lead.  That may unfortunately have the effect of leading some share of Clinton supporters (and Trump opposers) not to bother to vote on November 8.  They may feel it would not matter, so why bother.  But there are important reasons why all those supporting Clinton, who want the country to move forward rather than backward, need to make the effort to vote.  This blog post will outline a few.

a)  Trump’s share in the vote might well be higher than what the polls indicate:  As of October 23, an average of recent polls indicates that Clinton leads Trump by about 7 percentage points nationally.  While in the US system the candidate receiving the most votes nationally is not necessarily the one elected (due to the electoral college system, so only the votes in a limited number of swing states decide the outcome, as discussed in this earlier blog post), a national margin of 7 percentage points is substantial and reflective of what is happening in the key states.

But the final vote may well be different.  First, it is common that there is a tightening in the race in the last few weeks of most American elections.  There is a good chance this might happen again here.  But second, and more fundamentally, it is important to recognize that the polls may not be assessing accurately the extent of Trump’s support.  This is not due to any kind of conspiracy, or incompetence, but rather because polling this year is particularly difficult to do well.  Trump is an especially controversial candidate, known for his racist as well as misogynist remarks in this campaign.  Some Trump supporters might not admit to a pollster that they support him.  His true support might be several percentage points higher than what the polls indicate, and there are indications that this may have been an issue during the polling for the primaries in at least some of the states. I am not saying that it necessarily is now, but rather that we just do not know.

b)  A focus by Trump on high turnout of his base, instead of a broadening of his base, is not an unreasonable strategy:  Most major party candidates for the presidency seek to broaden their base of support as the election approaches by appealing to the middle.  Trump has not done this.  His focus has been and continues to be on energizing his base, with a continued use of extremist remarks to stoke concerns (the election is rigged, Hillary is a crook whom I will throw into jail, I won’t necessarily accept the results of the election unless they show I won, and so on).

With a base of support that is well less than 50% (even if one discounts the polls to a significant extent; see above), such a strategy might be seen as making it impossible to win.  The moderate middle is not attracted, but indeed repelled.  But it is not necessarily an unreasonable strategy.

The key is to recognize that a very high share of eligible Americans do not vote.  In the 2012 presidential election, only 58% of the population that were eligible to vote in fact cast a ballot for the presidency.  If Trump is able to energize his base and get a high share of them to vote, they can end up winning.

This can be illustrated with some numbers.  Using the polling averages as worked out by the Huffington Post, and rescaling to remove the undecideds, then as of October 23, polling indicated that Clinton would receive 48% of the vote and Trump 41% (with others receiving 11%, primarily Gary Johnson of the Libertarians and Jill Stein of the Greens).  To arrive at these numbers, pollsters used various methods to try to take into account the likelihood that those being polled would actually vote.  But none of these methods are very good.  Some pollsters ask the individual whether they voted in the previous election. However, the share saying they voted is always substantially higher than the share we know actually did vote.  Or some pollsters adjust the figures based on patterns for the share of those who voted in the past who have a similar income or education level, or are of the same ethnic group, or some other such grouping (using exit polling).  But this also does not work very well since the share of different groups who vote changes from election to election depending on the candidates and other issues.

For the purposes here, which are simply illustrative, let’s assume that these polling numbers reflect accurately the share of the population who prefer each of the candidates, but not necessarily the shares of those who actually will vote.  Furthermore, let’s assume that 53% of Clinton’s supporters will actually vote while 63% of Trump’s supporters will (recall the actual average in 2012 was 58%).  Multiplying out the numbers to get those who actually will vote, one finds that Trump in such a scenario would receive a higher share of the vote than Clinton:

Supporters

Turnout

Voters

Share of Vote

Clinton

48%

53%

25.44

44.1%

Trump

41%

63%

25.83

44.8%

Other

11%

58%

6.38

11.1%

All

100%

 

57.65

100.0%

Turnout matters.  A strategy focussed on turning out a high share of your base supporters, by energizing them through extremist rhetoric with no suggestion of compromise, is not necessarily an irrational one, even if it means losing the more moderate voters.  You could end up with more votes than your opponent.

c)  The winning margin matters for Trump to accept the result of the election:  If Hillary Clinton wins the election, but by a relatively narrow margin, Trump has said that he will not necessarily accept the result.  Trump made this clear in the third presidential debate, and has repeated his remarks since then despite of, and in the face of, strong criticism.  An important strength of American democracy, which distinguishes it from what is seen in a number of other countries around the world, is that the loser of the election concedes and accepts the result.  It might take some time (and court challenges) to determine the winner, but in the end the loser has always graciously accepted the decision (as Al Gore did in 2000).

Trump has been intentionally ambiguous on whether he will.  But the larger the margin by which he loses, the more difficult it will be for him to contest the results.

d)  The winning margin matters for the Republicans to move on:  Trump has upended the national Republican Party by capturing a base, primarily of angry white males with less than a college education, who have said they are willing to take extreme measures to get what they want.  If Trump loses, but by a relatively narrow margin, one can be sure that there will be Trump-like candidates seeking the Republican nomination in 2020, and perhaps even Trump himself.

Strong supporters of the Democrats might feel that this may not be so bad.  Such a candidate would likely lose again.  But that would be short-sighted.  Democracies need a multi-party system, with at least two responsible parties that can each govern responsibly. One-party states, whether in Japan or elsewhere, end up in difficulty.  And one-party states are indeed rare.  Eventually, an opposition party wins, as the electorate tires of those in power and as those long in power become increasingly ineffective.

American democracy needs a responsible opposition party.  Republicans at the national level are not providing that now, and that is a problem for all of us.

e)  The winning margin matters for Clinton to govern effectively:  Everyone agrees that there is much that needs to be done.  But opponents of the measures a Clinton administration would promote to move the  country forward would be emboldened in their opposition should Clinton win by a relatively narrow margin.  The larger the margin, the more difficult it will be for her opponents to block her proposals.

f)  There is an innate inconsistency to be opposed to Washington gridlock, but also to be in favor of divided government:  Everyone agrees that gridlock in Washington is bad.  The country needs to move forward in numerous areas, but gridlock is blocking it.  At the same time, political scientists have long observed (and backed up in their research) that voters often prefer “balanced” government, where the executive branch is controlled by one party with the legislature by the other.

This arrangement may have worked well in periods in the past.  With the system of checks and balances built in to the US Constitution, one branch of government cannot change much alone, but must also receive the support of the other branches (with the judiciary playing an essential, but separate, role as well).

This changed, however, over the last two decades.  Rather than seek common ground on measures, with compromises in order to move things forward, Republicans in Congress decided to adopt a position of opposition.  As documented in the excellent book of Thomas Mann and Norman Ornstein, It’s Even Worse Than It Looks, Republicans decided that if the administration supported something, they would be opposed.  This applied even on measures that they themselves had originally proposed.  The authors, one based at the left-of-center Brookings Institution and one at the right-of-center American Enterprise Institute, provide numerous examples.

Such opposition continues.  Last week, Senator John McCain (who at one time was considered a relative moderate among Republicans) said on a radio talk show that he and his colleagues will oppose any Supreme Court nominee of Hillary Clinton.  He said “I promise you that we will be united against any Supreme Court nominee that Hillary Clinton, if she were president, would put up. … I promise you.”  While a spokeswoman later sought to moderate his position, it does not appear that his views had in fact changed.

Such an approach to government, of united opposition to any proposals put forward by the chief executive, can work in a different form of government.  In parliamentary systems (such as in the UK), the opposition party will typically oppose any measures put forward by the prime minister.  But the prime minister represents a majority in parliament, and hence with party line votes the measure will pass.

But the US Constitution did not establish a parliamentary form of government.  Rather, the system set up by the US Constitution has an independently elected president, along with certain powers assigned to the legislature (such as to make laws, pass a budget, provide “advice and consent” on judicial and senior executive branch appointments, and more).  It is a system of checks and balances, and does not work well when one party decides to act like the opposition in a parliamentary system and routinely oppose measures proposed by the chief executive.

A large winning margin by Hillary Clinton will make it more difficult for a Republican majority to continue to act in this way, at least at the start of the new administration.  And while it is conceivable that the Democrats might win control of the Senate (they need to pick up a net of four seats, assuming Clinton wins so that Vice President Tim Kaine will have the tie-breaking vote), it is doubtful they will pick up the net of 30 seats required to win control of the House.  Too many seats have been gerrymandered.

Voters can resolve this by not voting for divided government, but rather for one party.  And if that party is not to be the one with Trump as president, that means the Democrats. What will not resolve the issues would be to vote for Clinton, but then vote for Republican candidates for the House and the Senate, including those who have sought to keep their distance from Trump, with a number saying they will not themselves vote for Trump.  But it is not really that vote that matters.  What matters is the vote they will take for the leadership of the House or the Senate, and whether that leadership says that they will oppose anything being proposed by Clinton, as they have for Obama.  If so, then gridlock will continue.

Conclusion

It would be surprising if Hillary Clinton were not to win this election.  I do not expect her to lose.  But it should be recognized that it is possible.  While the polls put her comfortably ahead as I write this, polls can be wrong, for reasons discussed above.  And we have seen two major such cases already this year.  Most expected British voters would reject the proposal in the June referendum to leave the European Union (Brexit).  Most polls indicated the vote would be in favor of staying.  Instead, it lost, and by the substantial margin of 52% to leave and 48% to stay.

To be fair, the polls in the Brexit referendum were relatively close, especially just before the day of the vote.  A better example of how the polls can be wrong in a major way was the vote in Colombia on October 2 on whether to accept the peace accord the government had negotiated with the FARC rebel army.  The war had been going on for decades, and about 220,000 Colombians had died over the years.  Polls before the vote indicated that over 60% of Colombians would vote in favor of the accord.  But it narrowly lost, by 50.2% to 49.8%.  It is not clear why, although there are many theories.  But one important factor was turnout.  Only 37% of eligible Colombian voters actually voted, perhaps because they believed the peace accord would win easily.  Voter turnout was especially hurt along the country’s Caribbean coast, where a hurricane, while it remained off shore, nonetheless delivered heavy rains on the day of the vote.  Support for the peace accord was especially high in that region, but turnout was low.

I would not predict that the polls in the US presidential elections are wrong, but that there can be uncertainties.  This is especially so this year.  And, for reasons discussed above, the issue is not only who will win or lose, but also what the winning margin will be.  So vote this November 8, and vote for Hillary Clinton.

 

The Structural Factors Behind the Steady Fall in Labor Force Participation Rates of Prime Age Workers

labor-force-participation-rate-ages-25-to-54-all-male-female-jan-1948-to-sept-2016

I would like to acknowledge and thank Mr. Steve Hipple, Economist at the Bureau of Labor Statistics, for his generous assistance in assembling data on labor force participation rates used in the blog post below.  This post would not have been possible without his help.

A.  Introduction

Increasing attention has recently been directed to the decline in labor force participation rates observed for men over the last several decades, and for women since the late 1990s.  The chart above tracks this.  It has indeed been dubbed (for men) a “quiet catastrophe” in a new book by Nicholas Eberstadt titled “Men Without Work”.

The issue has been taken up by those both on the right and on the left.  Even President Obama, in one of the rare “By invitation” pieces that The Economist occasionally publishes, has highlighted the concern in an article under his name in last week’s issue (the issue of October 8).  President Obama treats it as one of “four crucial areas of unfinished business” his successor will need to address.  A chart similar to that above is shown.  President Obama notes that in 1953, just 3% of men between the ages of 25 and 54 were not working, while the figure today is 12% (that is, the labor force participation rate fell from 97% to 88%).  The share of women of the same age group not participating in the formal labor market has similarly been falling since 1999.

While Obama is careful in his wording not to say directly that all of this increase in those not working was due to “involuntary joblessness”, he does note that involuntary joblessness takes a devastating toll on those unable to find jobs.  This is certainly correct. The fundamental question, however, is to what degree do we know whether the rise has been involuntary, and to what degree has it risen due to possibly more benign factors with rational choices being made.

Nick Eberstadt, who is perhaps the person most responsible for raising the profile of this issue, is a senior researcher at the conservative American Enterprise Institute.  He sees it as a major problem.  And a recent piece by the conservative columnist George Will, which is essentially a review of Eberstadt’s new work, appeared in the journal of conservative opinion the National Review, where it was subtitled “American men who choose not to work are choosing lives of quiet self-emasculation”.

On the left, Larry Summers praised Eberstadt’s book in a review published in the Financial Times (see this blog post of his for a non-paywalled summary), as did Justin Fox in this post at Bloomberg View.  But both emphasize factors beyond the worker’s control, in particular that “good” jobs are disappearing because of technological change (as well as, perhaps, international trade), and that America’s high prison incarceration rates have made it impossible for many men to be hired for the jobs there are.

Alan Kreuger, a Professor of Economics at Princeton and a former chair of President Obama’s Council of Economic Advisers, suggests a different factor in a recent carefully done academic analysis.  (See also this piece by Peter Coy for a non-technical summary of Krueger’s work.)  He found that poor health may well explain the high level for men, and found from independent data that nearly half of prime age men not in the labor force are taking pain medication daily.  As with the others, Krueger stresses that the issue is an important one. He concludes his paper by noting that “The decline in labor force participation in the US over the past two decades is a macroeconomic and social concern”, and that addressing it for “prime age men should be a national priority”.

While both sides have praised Eberstadt’s work, it is probably not surprising that they stress different underlying causes.  Those on the right blame individuals for becoming increasingly unwilling to work, abetted by foolish government policies that have enabled them to stay at home rather than get a job.  Those on the left emphasize instead that “good” jobs are disappearing because of technological change (as well as, perhaps, international trade), and that America’s high prison incarceration rates have made it impossible for many men to find jobs.  Professor Krueger’s conclusions are somewhere in between:  Individual factors (health) may explain what is being seen, but the health issues may be getting worse due to factors beyond the individual’s control.

It is not clear to me that any of these explanations really suffice.  But to develop an understanding of what might be going on, it is important first to examine more closely the underlying data for those who are not in the labor force and the reasons they give for this.

A first step is to separate the male and female rates, such as in the chart above.  It is an update of a chart that appeared in a post on this blog from last March (which in turn updated a similar chart from an even earlier post, from August 2014).  It tracks the monthly rates since 1948, with male and female rates shown separately as well as for everyone together.  Since demographic factors will affect labor force participation rates, particularly as a consequence of the increasing share of the baby boom generation who are now moving into their normal retirement years, the chart controls for age distribution by including only those aged between 25 and 54, the prime working years.

As the chart shows, the overall labor force participation rate (for men and women together) has been falling since the late 1990s.  The overall rate rose prior to then, but solely because the female rate was then rising strongly, as women increasingly entered into the formal, paid, job market. This peaked for women in the late 1990s, after which their rate as well as the overall rate began a slow fall.

The male rate, in contrast, started from a high level, of around 97% in the mid-1950s, after which there has been a slow but more or less steady fall.  It is now around 88%. Interestingly, since 1999 the female rate has moved almost exactly parallel to the male rate (at 83 to 84% of it, as discussed in the earlier blog post), suggesting that the underlying causes of the declines in both since 1999 might be similar.

A critical question is why.  The breakdown into separate male and female rates is a first step, but only a first step.  One wants to go beyond this.  The purpose of this blog post will be to take that next step, using BLS data I recently became aware of which reports on the survey responses of individuals on the primary reasons they are not in the labor force. This post will first review those results, and will then discuss some of the reasons that might explain the declining rates, especially for men.

B.  Non-Participation in the Labor Force by Prime-Aged Males

We will look at the rates for males first.  The chart below provides the reasons given (as a share of the male population aged 25 to 54) for why they were not participating in the labor force over the years from 1991 to 2015:

males-reasons-for-not-participating-in-the-labor-force-1991-to-2015The data was assembled by Mr. Steve Hipple, an economist on the staff of the Bureau of Labor Statistics (BLS) of the US Department of Labor.  He authored a Beyond the Numbers article of the BLS in December 2015 titled “People who are not in the labor force:  Why aren’t they working?”, which provided a first look at the reasons given by respondents for why they are not participating in the labor force, focusing on data for 2004 and for 2014.  The chart above is based on data assembled by Mr. Hipple for the full period from 1991 to 2015.

The data is derived from responses to queries made in the Annual Social and Economic Supplement to the Current Population Survey (CPS-ASEC).  This is a joint effort by the BLS (which conducts the Current Population Survey monthly, from which the official unemployment rate, among many other measures, is derived) and the US Census Bureau. The CPS-ASEC survey is undertaken once a year, each Spring, and asks a larger national sample a broad range of questions focused on conditions (such as on employment and household incomes) in the previous calendar year.  Among the questions it asks is whether each adult member of the household was in the labor force (where the labor force is defined as all those employed and all those unemployed who were actively seeking employment at some point in the year), and if not, what the reason was.  The possible responses are those listed in the chart above.

Several points should be noted:

a)  The numbers ultimately come from a survey of individuals, and hence will have the shortcomings of any survey.  There will be statistical error simply from the size of the sample, but more importantly also non-statistical error from how people choose to respond.  The reasons why an individual may not be participating in the labor force may be interpreted differently by different individuals, and multiple factors may apply (for example, they may be somewhat ill, have had difficulty in finding a job, and are at an age where early retirement is possible).  While such inherent limitations should be recognized, they also may not so much affect the trends, and the trends are of most interest here.

b)  The CPS-ASEC survey asks the respondents on their status in the previous calendar year, and covers the status over the entire year.  Hence if they were employed for part of the year but not for the full year, they would still be counted as part of the labor force.  The CPS survey, in contrast, is monthly, and asks for the status at that point in time (or, to be more precise, in the preceding week).  For this reason, one should expect to find that the share of the population counted as in the labor force to be higher in the annual figures than in the monthly figures, since they will be included in the annual numbers if they were in the labor force at any point in the previous year and not simply at the point in time of the monthly survey.  And one does see this in the results reported.  Conversely, the share not in the labor force will then be lower in the annual figures than in the monthly estimates.

The annual and monthly figures will, however, move similarly.  But note that the chart at the top of this post is based on the monthly estimates from the CPS, while the charts here for the reasons for not participating in the labor force are drawn from the annual CPS-ASEC estimates. The totals for those not in the labor force will differ for this reason.

c)  Participation in the labor force is defined as anyone in a paid job for as little as one hour in a week, plus those unemployed (defined as those actively looking for a job but do not have one). Thus to be counted as not in the labor force but “retired” or “going to school” is quite strict.  If one is retired but working for pay a few hours a week, or in school but working in the school cafeteria for some extra income, one is counted as part of the labor force and hence will be excluded from those not in the labor force.

d)  One must similarly be careful in the interpretation of the “could not find work” category.  The CPS-ASEC questionnaire asks whether the respondent had spent “any time trying to find a job” in the previous year.  If they had, they would be included in the unemployed.  If not, they would then be asked why they were not in the labor force that year, from this list of possible reasons.  Those who responded that they “could not find work” would be saying that they could not find work during this period even though they also say they had not actively searched for a job in this period.  It is possible, however, that they had looked before, could not find anything suitable, and believed this still to be the case even though they had given up actively looking.

e)  Only data going back to 1991 could be readily assembled.  While this covers a significant period, 25 years, it would be interesting if results further back were available.  The downward trend in the male participation rate started in the mid-1950s, and it would be of interest to see whether the causes prior to 1991 were similar to those since then.  I suspect they probably were, but this is speculation and one would like to see if that is indeed true.

The CPS-ASEC goes back to 1959 I believe (although initially under a different name), the monthly CPS goes back further, and a CPS questionnaire I found from 1978 asks a similar question (although without “retired” as a separate category).  The older data is not easy to access, however, and requires special software as well as expertise that I do not have.

However, the 25 years of data from 1991 to 2015 do show some interesting trends. Specifically, for the male rates (the chart above):

a)  The share of males aged 25 to 54 not in the labor force almost doubled over this period, from 5.9% of the male population in this age group in 1991 to 11.5% in 2013 and 2014 before dipping to 11.1% in 2015.  This was a significant increase.  And as the chart at the top of this post shows, this was a continuation of a similar trend in the decades prior to 1991.

b)  The increase in the total was not due to just one or two causes, but rather to substantial increases in the shares for each of the given reasons other than, interestingly, “could not find work”.  About 0.6% of the prime working age male population responded “could not find work” in 1991, and 0.7% did in 2015.  But the share reporting “could not find work” did fluctuate over the period, bumping higher in those years when the labor market was weak and unemployment high (1992, 2002, and then especially in 2009/10), and being compressed in the mid to late 1990s (the Clinton years) when labor market conditions were strong.  It appears to be capturing that labor market participation rates of prime age males are less when labor markets are weak.  This would be hidden unemployment.  However, the extent is limited.  The difference between the peak rate (in 2010) and the low rates in the late 1990s is only about 1.0% point.

c)  In terms of shares among those prime age males not participating in the labor force, the most important reason given was “ill or disabled”.  Interestingly, this share fell from close to 60% of those not in the labor force in 1991, the first year for which we have data, to 50% in 2015.  There were larger relative increases in the other causes (other than “could not find work”).

d)  The shares that rose the most (in relative terms) were the share of prime age men going full-time to school (rising from 11.0% of those prime age males not in the labor force in 1991 to 15.5% in 2015), the share retired (rising from 5.9% to 9.6%), and the share taking on home responsibilities (rising from 4.6% to 10.8%).  The share of those who could not find work fell from 10.5% to 6.3%, and the share for “other reasons” fell from 8.5% to 7.9%.

We will discuss below some of the possible reasons for these changes.

C.  Non-Participation in the Labor Force by Prime-Aged Females

A similar chart can be drawn for the responses of women not in the formal, paid, labor force.  The huge post-World War II change was of course the entry of women into the paid labor force, almost doubling from 34% of prime working age women in 1948 to 77% in 1999.  The rate then slowly fell, in parallel with the male rate, to 74% in 2015.

What dominates in the share of women not in the paid labor force is the share with home responsibilities.  This came down sharply from 1991 to 1999, as the following chart shows, and almost certainly in the period before then as well.  Since 1999 it has fluctuated, but appears to be on an upward trend (as the male rate is as well, although starting from far lower levels):

females-home-responsibilities-as-reason-for-not-participating-in-the-labor-force-1991-to-2015

Interestingly, over the past two decades the rate fell when the labor market was strong in the mid to late 1990s, rose as the labor market weakened with the recession that began a few months after George W. Bush took office, fell once the labor market recovered (but with a lag), and then turned upward again after 2009 as the labor market weakened again. It then fell in 2015.  This pro-cyclicality may be implying that, for women, the “home responsibilities” reason is being given as the stated reason for not participating in the formal labor force, when in fact it may to some degree reflect hidden unemployment.  But we cannot know for sure.  It might also reflect what kind of jobs, and their wages, that women can get when labor market conditions are weak.  The impact of wage rates will be discussed below.

The share for “home responsibilities” remains high, however, and would dominate all else in a chart if left in.  To examine what is going on it is therefore best to first subtract out the home responsibilities cause, accounting for it separately, and then examining the break-down for all the other reasons given for not participating in the labor force.  The result then is a chart which is remarkably similar to the chart for males:

females2-reasons-for-not-participating-in-the-labor-force-excluding-home-responsibilties-1991-to-2015

One finds:

a)  The total rate for women not in the labor force, once one excludes those with home responsibilities, almost doubles between 1991 and 2015, as it did for the males.

b)  Once again, the largest share of women of ages 25 to 54 not in the labor force (and excluding also those with home responsibilities) are those recorded as ill or disabled.  But the share ill or disabled was largely flat between 1991 and 2015, accounting for 55% of the total in 1991 and 57% in 2015.

c)  The second highest share, as with the male rates, was the share going full time to school.  But it was largely flat for this group of women, at 20.2% of the total in 1991 and 20.5% in 2015.

d)  Women, similar to men, saw a relatively high increase in the share retired, rising from 5.7% in 1991 to 12.5% in 2015.  Also similar to men, the share recorded as could not find work fell sharply from 8.2% in 1991 to 4.1% in 2015.  And the “other reasons” share also fell, as it did for the males, from 10.4% to 5.9%.

There appears then to be a similar pattern in the female rates as in the male rates once one removes the effect of home responsibilities.  Furthermore, this similar pattern held prior to 1999 as it did after that; there was not a sharp break in that year.  Indeed, excluding full time home responsibilities for both men and women, one finds that the curves of the shares in the 25 to 54 age group not in the labor force basically lie on top of each other:

males-and-females-shares-not-in-labor-force-excluding-home-responsibilities-reason-1991-to-2015

 

For both men and women, the share not participating in the labor force (and excluding those with full time home responsibilities as well), both rose sharply over this period, basically doubling, with similar shares throughout (within a percentage point or less).

Finally, it is interesting to compare between men and women the stated underlying reasons for not participating in the labor force.  The chart above shows that the totals, once when excludes the big differences between men and women with home responsibilities, are similar, and has been over time (since at least 1991).  The chart below compares the male and female rates for the underlying reasons, for the year 2015:

share-not-participating-in-labor-force-in-2015-excl-home-responsibilties-males-and-females

The totals, excluding home responsibilities, are similar, as already seen.  But it is interesting that the male and female rates for the underlying causes, other than home responsibilities, are also mostly similar.  The shares of prime age men and women who are not in the labor force due to illness or disability are quite close, at 5.6% for men and 5.4% for women.  A conclusion that illness or disability is more of a problem for men than for women is not supported.  Also very similar are the rates in this age group who report being retired:  1.1% of men and 1.2% of women.  This is also the case for those reporting to be full time students, with rates of 1.7% for men and 1.9% for women.  The female rate reporting they could not find work is less than the male rate (0.7% for males and 0.4% for females), as is the rate reported under “other” (0.9% for males and 0.6% for females), but both of these are relatively small in absolute level.

It appears that similar factors for men and women (other than home responsibilities) might be underlying these rates, and their increase over time.  What might those be?

D.  Possible Reasons for These Changes in Shares Not in the Labor Force 

It was noted in the introduction that conservatives have interpreted the decline in the labor force participation rates, particularly of men, as reflecting an increasing unwillingness to work.  Liberals have focussed more on fewer “good jobs” or an inability to get and hold them due to conditions like previous incarceration or deteriorating health.  Can we conclude from the data reviewed above which, if any, of these interpretations might be correct?

To summarize some of the points already noted above:

a)  There appear to be multiple reasons given in the responses for each sex as to why they are not participating in the labor force.  That is, it is not just one factor that explains, at least directly, the increasing share not working.

b)  Furthermore, aside from changes in the shares taking on home responsibilities (which do differ, and differ greatly, between each sex), the multiple reasons appear to be broadly similar in the share of males and share of females not participating in the labor force.

c)  But one stated reason that is low in level and also has not grown over time is the share of the prime working age population who are not in the labor force because they say they could not find work.  The share was just 0.7% of all men of prime working age in 2015, and just 0.4% for all women of prime working age.

d)  The share saying they could not find work varies to a limited degree with the overall state of the labor market (somewhat higher when official unemployment is high, although not by a huge amount).  And while it was relatively high (in comparison to its level at other times) in 2009/2010, when official unemployment reached 10%, it is now back to levels seen previously.

e)  The “other reasons” factor is relatively low and without an obvious upward trend. This is fortunate, since we do not really know what lies behind it.  But being low, it does not appear to be important.

Can the increase be attributed then to an increased unwillingness to work, as the conservatives charge?  That is not so clear.  While this is now more speculative, one can also interpret the data as reflecting more positive developments.  Specifically:

a)  There was a substantial increase in the share of men and women aged 25 to 54 who were enrolled full time in school.  This is almost certainly a good thing.  This probably reflects an increasing share of students in their late 20s and perhaps later enrolled in post-graduate studies.  Medicine requires many years of study, for example, and business schools increasingly require students applying to their MBA programs to have worked for several years before being accepted.  One might similarly see students in post-graduate academic programs, in particular Ph.D. students, who are 25 or older.  Finally, there may now be an increasing number of students who have worked for 5 or 10 years or more who decide to go back to school to learn a new skill or profession.  This is not bad.  Finally, it is worth noting that the increasing shares of students in this 25 to 54 age group are similar for both men and women, suggesting that the underlying cause may well be due to developments in the system of education.

b)  One also sees increases, and similar increases for both men and women, in the share saying they are retired despite being age 54 or less.  Such early retirement is certainly unusual, but the shares are low (1.1% of the men in the full 25 to 54 age group in 2015, and 1.2% of the women).  But given how the retirement system has changed in recent decades, an increase over time should not be surprising. Traditional defined benefit pension systems typically required work to some age (perhaps 62 or 65) before they could be drawn.  With income in retirement now driven by individual accounts (401(k)s, IRAs, and even just normal savings), there is now more of an opportunity to retire earlier.  To the extent early retirement reflects what a person prefers, and is something that he or she can afford, this should be seen as a positive development.

c)  Home responsibilities have been the largest single reason for women not being part of the formal, paid, labor force, and came down sharply until 1999.  Since then it has risen, but with significant volatility.  As noted above, it may well reflect a degree of hidden unemployment, as it appears to rise and fall (although with a lag) with labor market conditions.  But if it is a personal preference, and something the family can afford, it is not necessarily a bad thing.

d)  The share of males reporting home responsibilities as the reason they are not participating in the formal labor force, while still well below the female rates, has trended upwards over this period.  This may reflect both changes in social acceptance of males staying home to take care of children or elderly parents, but also the increase in female participation in the labor force to the ceiling reached in 1999. With more women in the formal labor force, often in jobs that pay well, a married couple might well decide that they prefer that the husband take on home responsibilities rather than the wife.  The fact that such a choice can now be made is a good thing.

There may therefore be benign explanations for several of these developments leading to lower labor force participation rates.  From just the evidence here, we cannot be sure.  But similarly, we should not assume the development are necessarily negative.

Health issues are more complex, and also a much larger factor.  The highest single cause cited for not participating in the labor force was illness or disability, for both prime age men and for prime age women when one excludes home responsibilities.  In 2015, it accounted for 50% of the prime age men and 57% of the prime age women (excluding home responsibilities) not in the labor force.  The shares, while high, did not change that much over time.  They were 60% for men and 55% for these women in 1991.

But a roughly constant share of a total that has doubled implies a rough doubling due to illness or disability.  It is certainly possible that health conditions have deteriorated for a significant sub-set of the population in recent decades.  As earlier posts on this blog have documented, median real wages have been stagnant since around 1980.  And the distribution of income has also become remarkably worse, with all but the top 10% seeing their real incomes falling between 1980 and 2014.  Stagnant incomes do correlate with poor health status.

More directly, a study published last year by Professors Anne Case and Sir Angus Deaton (Nobel laureate) of Princeton found that mortality rates of middle-aged non-Hispanic white Americans have actually risen in the last decade and a half.  There are clearly issues with health, at least among a significant segment of the population.  This coincided with reports of increased incidence of severe pain and increased daily use of prescription strength pain killers, a factor highlighted by Professor Krueger in his paper on labor force participation rates.

It is difficult to say with certainty, however, whether such health conditions necessarily account for the rising share of prime working age Americans who report illness or disability as the reason for not participating in the labor force.  It may also be the case that the safety net that provides support for those who are ill or disabled (such as through disability insurance that is provided as part of the Social Security system in the US) has improved over time.  Workers who previously could not get such support may have remained in the official labor force, but in low wage or unproductive jobs and in great pain.  If support is now provided to such workers, which was not done before, this can be a good thing.

But it must be recognized that the level of support provided to disabled workers is extremely stingy in the US.  The average monthly benefit paid under Social Security Disability Insurance in the US is currently only $1,166 per disabled worker (as of August 2016), or $13,992 a year.  You do not take this if you can work.  The share of prime working age men not in the labor force is also higher in the US than in such countries as Canada, France, Germany, and Japan, all of which have far more generous safety nets for disabled workers than the US has.

Further work is needed to separate out these possible causes for the increase in the share reporting illness or disability.  And it is important, given the dominance of this stated reason for those not participating in the labor force.

E.  The Impact of Low and Stagnant Real Wages

Multiple factors appear, then, to underlie the rise in the share of prime working age males (for decades) and females (since 1999) who are not participating in the formal labor force. While certain analysts may emphasize one factor or another, often consistent with their particular political leanings, the truth is that the differing interpretations may well apply to different sets of individuals.  For example, those reporting that they have retired may have retired because, as noted above, they wanted to and they could afford to.  But it might well also reflect that at least some in this group could not find a good job, and hence decided (unhappily) to start to draw on their retirement savings.

More fundamentally, it is not an issue of a strict either/or.  What is missing from the various rationales given is the recognition that there are trade-offs, and a decision is made as to whether to participate or not in the formal labor force depending on a balancing of these considerations.  And this is really just basic economics.  Econ 101 teaches that decisions are made by a weighing of different factors, and that one needs to recognize that there are trade-offs.

Critical to this is the need to recognize that median real wages have been stagnant for decades, as discussed in the earlier post on this blog cited above.  The issue is not whether or not “good” jobs exist, but rather how much is being paid in wages for those jobs.  Real wages have been flat, and the minimum wage is now more than one-third less in real terms than what it was in 1968.

At such rates of pay, prime working age men or women may well find it better, for example, to go to school for a few more years in the hope of getting a better paying job later, rather than work now at relatively low wages on a job.  As standard economics teaches, an important cost of schooling (and typically far more important than the cost of tuition) is the cost of not working while one is in school.  But if wages are low, what is lost from not working is not so high, and at the margin it makes more sense to go to school. And this factor has become increasingly important over the past several decades, as real wages have stagnated.

Similarly, at the margin one might well decide to retire early, if one can afford it, than to work longer in a low paying job.  As wages have stagnated in recent decades, standard economics teaches that more workers, at the margin, will choose to switch over to early retirement.  This will also hold for those who are ill or disabled.  What matters is not just what might be available to the worker in disability payments (which are low), but how much this is relative to the wages they might earn.  As the real minimum wage has fallen over recent decades, taking disability payments (if one is eligible) becomes relatively more attractive than taking a minimum wage job.

And this of course very much holds for those taking on home responsibilities, in particular for those taking care of children or elderly parents.  If the wage of the job one can get is stagnant and low, while the cost of child care and elder care has been going up, the rational choice will increasingly be to stay at home to provide such care rather than work in a formal paid job.

The stagnation in real wages since around 1980 might then help explain, at least in part, the increase in the share in recent decades of those prime working age men and women choosing not to participate in the formal labor market.  And it is interesting to note that the pace at which prime age men have chosen to stay out of the formal labor force accelerated in the period after 1980 compared to the period before.  Over the 27 years between 1953, when the prime working age male labor force participation rate peaked at an average over the year of 97 1/2%, and 1980, when the average over the year was 94 1/2%, the average pace of change in the rate was 0.111% points a year.  Over the 36 years from 1980 to 2016, the pace picked up to an average of 0.167% points a year.  This is 50% faster.  It accelerated in the period that real wages have stagnated.

F.  Conclusion

The labor force participation rate for the prime working age population has been declining for men since the mid-1950s and for women since 1999.  This is significant.  Growth depends on the working population, and if fewer work, there will be less growth.  And not being able to find a job when one wants one, whether you are counted among the openly unemployed or the hidden unemployed, can be devastating to an individual.  It is not just the income that is lost, although this is of course hugely important, but also the impact on the psyche and sense of self-worth.

One has to be careful, however, in any attribution of the cause of this increasing share of the population not in the formal labor force.  Many factors are involved, and one should not jump to a conclusion such as that people are lazier now than they were before, or that jobs are simply not available now and were before.  One should rather recognize that choices are being made and that there are tradeoffs.  People may rationally and happily be choosing to enroll as a full time student, or to stay at home to take care of children or elderly parents, or to retire early.

But one should similarly not jump to the conclusion that these are necessarily happy choices.  This is especially clear for those who are not working due to illness or disability, who may obtain minimal or even no support from various disability insurance programs. Indeed, I would suspect that most of those who are not working due to illness or disability are depending on a working spouse for support.

Recognizing that these are choices that are being made is simple, basic, Economics 101. The choices may be happy ones or not, but they are all choices.  Basic to such choices is what one would obtain by working at a job, which is the opportunity cost of what one is giving up by deciding not to participate in the formal, paying, job market.  Central to this is the fact that wages have been stagnant for decades in the US, since around 1980.  At the margin, it might make more sense now than it had before not to seek a job but rather to enroll as a student, take care of home responsibilities, or retire early.  This stagnation in real wages may in part explain the acceleration of the pace of working age men dropping out of the formal labor force since around 1980.

This then suggests a further reason for why we need to be concerned with real wages that have remained stagnant, despite the significant productivity growth of recent decades. Real GDP per worker (i.e. productivity) is now 60% higher than what it was in 1980, but wages have been flat.  Prior to 1980, real wages and GDP per worker both rose at a similar rate. This then broke down after around 1980.

Returning to a more equitable growth process is not, however, a trivial task nor one which can be accomplished by fiat.  But this analysis suggests that should progress be made towards this, one would then expect to see higher labor force participation rates.  And there are indeed actions the government can take.  A number were discussed in this earlier blog post.  Ensuring workers are in a position to bargain for good wages by keeping the economy at close to full employment is probably the most basic.  And raising the minimum wage, which is now more than one-third below where it was in real terms in 1968, and indeed lower in real terms than what it was when Harry Truman was president, would be important for all low wage workers.

Low and stagnant real wages have had a number of adverse effects on the economy, including on productivity.  A lower rate of labor force participation is likely also one of them. If you want more people in paying jobs, pay them better.