More Trump Failures

Posted on by

Chart 1

A. Introduction

The failure of Trump’s economic policies in terms of his own stated objectives is becoming increasingly clear.  This is not to say that those stated objectives always make much sense.  They often do not.  But they provide a metric to assess whether Trump is succeeding in terms of his own stated objectives.

The release on July 2 of the regular monthly BLS Employment Situation report provides figures that allow for an update on where some of these stand.  This short post will look at several of them.

B.  Trump’s Anti-Immigrant Policies Have Not Led to Improved Job Prospects for Native Born Labor

The Current Population Survey of households of the BLS (the basis for the reported unemployment rate and related measures) provides a breakdown of labor market participation, employment, and unemployment between immigrants (which the BLS refers to as the foreign born population) and those born in the US – the native born.  While the BLS does not provide seasonally adjusted figures for this breakdown, the nonseasonally adjusted figures can still provide a meaningful comparison, especially when taken over several years.

The chart at the top of this post shows the ratio of the unemployment rates of the native born population to that of immigrants, from July 2021 to June 2026.  An argument Trump has made against immigrants is that they have been “taking American jobs”.  If so, then the deportation of hundreds of thousands of them would lead – by this argument – to improved job prospects for the native born.  The unemployment rate of the native born should fall.

Under Trump it has not, while it did during the Biden presidency.  The ratio of the unemployment rate of the native born population to that of immigrants was on a downward trend during the Biden administration.  This is the opposite of what would have happened if Trump’s argument were correct.  Job prospects of the native born population (relative to immigrants) were improving during the Biden term.

It then reversed under Trump, despite of (or more likely because of) his anti-immigrant policies.  Immigrants have been deported in massive numbers, but this did not lead to a fall in the unemployment rate of the native born relative to that of immigrants.  Instead, it rose.

This is a relative measure – a comparison of the unemployment rate of one group (the native born) to that of another (immigrants).  While this is the type of measure that Trump’s view of the world would engender – of one group in opposition to another in a zero-sum world where there are only a fixed number of jobs – reality can be different.  What is good for one group is not necessarily – and indeed not normally – bad for another.

It is more appropriate to focus simply on what happened to the job prospects of the native born themselves.  Is there any evidence that the deportation of hundreds of thousands of immigrants since Trump took office in January 2025 led to lower unemployment among the native born?

There is not.  The unemployment rate of the native born in absolute terms – while very low under Biden and still relatively low under Trump – continued on the same path it was on before:

Chart 2

The unemployment rate had gone as low as about 3 1/2% during the Biden presidency – which is extremely low.  It is now about 4 1/2% under Trump – still low, but not as low as before the aggressive anti-immigrant campaign.  Indeed, the trend since late 2022 looks to be basically the same – whether when Biden was in office or when Trump was – with no shift evident despite the anti-immigrant policies.  More basic underlying factors have been driving the figures, and not Trump’s deportations.

Why then did the ratio of the unemployment rate of the native born to that of immigrants turn around and start to rise under Trump, as seen in the chart at the top of this post?  It was because while the unemployment rate of native born labor continued on its prior upward path, the unemployment rate for immigrant laborers leveled off:

Chart 3

While there is a good deal of noise in the data (as the sample size of immigrant labor is far less than that of native born labor), and the lack of seasonal adjustment makes it more difficult to see the trends, it does appear that the unemployment rate for immigrant labor leveled off after Trump took office.  It had been rising before, although again I would emphasize that all of these unemployment rates are low by historical standards.

But with the unemployment rate of immigrant labor leveling off after Trump took office, while the unemployment rate of native born labor continued to slowly rise, the ratio of the latter to the former rose.  That is, Trump’s policies appear not to have led to improved job prospects for native born labor, but rather did so for the immigrant labor still in the country.

This is, in fact, not surprising:  Reducing the immigrant labor force can be expected to most affect the group most similar to them, which is other immigrants.  But it failed in its stated objective of improving job market conditions for the native born.

C.  Employment in Manufacturing

Trump also campaigned on a promise to raise employment in manufacturing.  He said he would impose impossibly high tariffs on imported manufactures to force Americans to buy from domestic factories.  High tariffs have indeed been imposed, with an average rate as much as ten times higher than when Trump took office, and at a level not seen for the US since the 1940s.   But they have varied widely by country (with rates for China especially high for a period) and by commodity (with a rate of 50% for steel and aluminum, 25% for cars and trucks and their parts, and 100% for pharmaceuticals, among many others).  And there have been numerous exemptions and special exceptions benefiting specific firms, often announced by Trump via a post on his social media site.  It has been chaotic.

The stated aim has been to force manufacturers to produce their products in the US rather than import them.  This would lead – it was argued – to higher employment in manufacturing.

It has not:

Chart 4

Manufacturing employment recovered under Biden following the Covid lockdowns, and it recovered to a level higher than what it was before.  It is interesting to note, however, that employment in manufacturing had already begun to fall well before the Covid lockdowns.  It hit a peak of 12.79 million in July 2019 (and in fact had hit this level already in January 2019), and had fallen by 47,000 workers by February 2020 – before the Covid lockdowns.  It then plummeted.

Employment recovered as the lockdowns ended, and this continued under Biden to a level above the peak it had achieved before.  It then started to fall slowly over the last two years of the Biden administration.  That fall then continued under Trump.  As of June 2026 (and based on the most recent BLS estimates, which will be updated), there are now 75,000 fewer workers employed in manufacturing than when Trump took office.

But is this important?  While employment in manufacturing has been falling, the productivity of the labor employed in manufacturing (output per employee) has been rising:

Chart 5

The data are drawn from the data I downloaded from the BEA and BLS for the prior post on this blog.  It is quarterly as the BEA estimates for GDP are quarterly, and the data for 2025Q4 are still the most recent available for GDP at the sector level.  Manufacturing “output” is more formally called the value-added produced in the sector, and is shown here in real terms (at the prices of 2017).

Relative to the first quarter of 2012, manufacturing output as of the fourth quarter of 2025 was 21% higher in real terms.  Employment was just 6.3% higher.  The difference between the two reflects greater average labor productivity in the sector.  Note that this can be due not solely to higher productivity in the production of a particular good.  It can also reflect changes in the mix of goods.  I suspect (and this is speculation, as the data at the level of detail required is only issued on an annual basis, and that for 2024 is the latest available) the change in the mix of goods that are manufactured will account for much of this increase in average productivity in recent years.  In particular, production of semiconductors and related products was strongly supported by the Biden administration.  Some of the major plants that most focus on are now coming online, but there is also production of related products that receive less attention.  In the BEA data through 2024, production in the “semiconductor and other electronic component manufacturing” subsector was already 26.3% higher in real terms in 2024 than it was in 2019.  Manufacturing as a whole grew by 5.0% over this same period.

The higher productivity in manufacturing is a good thing.  It is what enables living standards to rise over time (although while a necessary condition, it is not sufficient in itself and requires supportive policies to ensure the gains are fairly shared).  Despite what politicians (of all parties) say, the aim should not be employment per se, but rather improvements in living standards.

D. Employment in Coal Mining

Employment in coal mining has also long been a priority for Trump.  Coal is a terribly dirty fuel (in all phases, from digging it out of the ground, to transporting it, to burning it), and coal burning power plants are also expensive.  Indeed, the marginal cost of keeping older coal-burning power plants active (to cover simply the cost of the coal that is burnt and the cost of operations and maintenance – these are high, especially for the older and hence less efficient coal burning plants) is higher than the full cost of newly-built solar and wind generation facilities at attractive sites, including the cost of storage.  And this is the case before taking into account the subsidies that are available for the clean generation of power.

But for whatever reason, Trump has pushed strongly to maintain or increase employment in the mining of coal.  He has failed:

Chart 6

Employment in the sector was 39,200 as of June 2026 (in the most recent estimate, which is subject to updating), versus 40,500 when Trump took office in January 2025.  That is a fall of 1,300, or 3.2%.  It was lower in March and April – at 38,400, a reduction of 5.2% – but has received a bit of a boost, probably because of the shortage of liquefied natural gas (LNG) resulting from Trump’s war on Iran (where LNG – natural gas – is a primary fuel for power plants).

There has been a decline, but all these figures are small.  There are simply not many workers employed in coal mining.  They account for only 0.025% of total employment in the US economy – i.e. 99.975% are employed elsewhere.  Indeed, the 39,200 in coal mining can be compared to the 280,200 employed in the solar energy sector in the US (counting those employed in the manufacture, installation, and related work on solar power systems) as of 2024 – more than seven times as many.

E.  Conclusion

Trump has made clear that he is seeking to achieve certain aims through his economic policies.  They do not always make a lot of sense in themselves, but Trump has been clear that they reflect what he is trying to do.  And he has regularly claimed that he has had great success.

The data indicate otherwise.

Some First Hints that the AI Boom May Be Having an Impact on Productivity in the GDP Accounts: And Not Just in the Way Most Imagine

Posted on by

Chart 1

A. Introduction

Employment hardly grew in 2025.  Total nonfarm employment increased by only 116,000 between December 2024 and December 2025 in the most recent BLS estimates. Employment in the Private Education and Health Services sector alone rose by 682,000, meaning that in the entire rest of the economy, employment fell by 566,000 – over half a million.

On the face of it, this appears to be inconsistent with figures on GDP growth.  GDP fell in the first quarter of 2025, rose at reasonably rapid rates in the second and third quarters, and then grew only slowly in the fourth quarter (slowly with or without an adjustment for the impact of the federal government shutdown during the quarter).  Most of the growth – such as it was – can be attributed on the GDP demand side to the boom in investments to provide AI services (data centers, software, and such).  See Section C of this earlier post.

An increase in GDP coupled with less of an increase in employment implies that labor productivity rose.  This is by definition, as labor productivity is simply GDP divided by employment.  Over time, growth in real income per person is only possible with growth in productivity, so this is not necessarily bad.  As long as high unemployment is not an issue (and it is not at this time – while the unemployment rate under Trump has been higher than what it was under Biden, it is still low by historical standards), employment is at the level that is possible given the size of the labor force.  Incomes can then increase only with an increase in productivity.  This is not what the Trump White House has been saying – with its stated focus on jobs, jobs, jobs – but the lack of coherence is not surprising.

But what lies behind this?  Section B of this post will first look at the aggregate figures, comparing what was observed in 2025 to the observed trend over the prior 12 years.  At the aggregate level, the rate of growth in GDP was a bit less in 2025 than what it was in the prior 12 years.  But employment growth was much less, so productivity growth in 2025 was necessarily higher than before.  The figures are shown in the chart at the top of this post.

This is at the aggregate level.  What is of interest is what happened in a few key sectors that may be leaders in and beneficiaries of the boom in AI investments.  This will be examined in Section C below.  The BEA has now released data that allows us to examine at the sectoral level where productivity grew on the supply side of the economy – and in particular in sectors that may be especially able to make use of the new AI systems that the recent investments made available.  While sectors as defined by the BEA in the NIPA accounts (matched with employment in those sectors from the BLS databases) are relatively broad, just two of them – Information and Finance (accounting for about one-quarter of the economy together) – had a disproportionate impact on the growth in GDP in 2025 as well as on the growth in labor productivity.  Outside of those two sectors, the growth in GDP and in productivity both slowed in 2025 compared to the years before.

Also, the impact on overall productivity in 2025 came not only from the observed growth in productivity in each sector of the economy taken individually.  In addition, there was a compositional effect arising from the especially rapid growth in sectors where labor productivity was relatively high – and sometimes exceptionally high – compared to the overall average.  These sectors included Information and Finance.  A shift in the sector composition of GDP – arising from relatively faster growth in a few sectors where labor productivity is high – will by itself increase average productivity in the economy.  This is separate from and in addition to any increase in productivity at the level of the individual sectors.  That impact has typically been ignored in the discussion of the impact AI may have on productivity in the economy as a whole, but was significant in 2025.  This will be discussed in Section D below.

The post will conclude with a short Summary and Conclusions.

B.  Growth in GDP, Employment, and Labor Productivity in 2025 Compared to the Prior Trend

The chart at the top of this post shows the growth rates – all in real terms – for the economy as a whole, for employment, and for labor productivity, in the twelve years from 2013 through 2024 and then in 2025.  The year 2013 is a good starting point as the economy had by then largely recovered from the 2008/09 economic and financial collapse.  GDP (output for the economy as a whole) is measured in quarterly terms, so the growth rates are over the period from the fourth quarter of 2012 to the fourth quarter of 2024, and then between the fourth quarter of 2024 and the fourth quarter of 2025.  Also, because the employment figures gathered by the BLS are for nonfarm payrolls, the figures for total output are for GDP excluding agriculture, to put this on the same basis as the nonfarm payroll figures.  However, since agriculture is such a small share of GDP (less than 1%), the growth rates shown are almost exactly the same and are well within rounding.

Overall output (real GDP) grew at a rate of 2.5% per year between 2013 and 2024.  Growth was lower in 2025 at 2.0%.  There is year-to-year volatility so the reduction in 2025 is not necessarily significant unless it is sustained (but it also does not support claims by Trump that the economy was booming in 2025).  Employment (nonfarm payrolls) grew at a 1.3% annual pace in the twelve years leading up to 2025, with this then falling to just 0.2% in 2025.

The growth in output was less in 2025, but the employment growth was far less, so labor productivity rose at a faster rate in 2025:  a rate of 1.8% compared to an annual rate of 1.2% in the years leading up to it.  A 1.8% rate of growth in labor productivity – if sustained – would be a good rate, and close to the long-term 1.9% rate the US enjoyed prior to the 2008/09 economic and financial collapse at the end of the Bush administration (a record dating back to 1870).

But what were the factors lying behind that 1.8% rate of growth in labor productivity in 2025?  Was it a result of productivity growing across the board in most sectors, or rather rapid growth in a few sectors and not much elsewhere?  As we will see in the next section, it was the latter.

C.  The Impact of Growth in the Information and Finance Sectors Alone on the Growth in Employment and Labor Productivity

As has been discussed in prior posts on this blog, GDP is a measure of the total output (i.e. product) of the domestic economy and can be estimated in three different ways:  1) by summing the demands for the product, i.e. how all of it is used (with inventory accumulation or decumulation acting as a balancing item to match up what is supplied with what is demanded); 2) by adding up all incomes accruing from that production as wages to labor and as profits; and 3) by estimating directly the net production (i.e. net of purchases of intermediate goods used in that production, and more properly referred to as value added) of every sector of the economy and adding it up.  In principle, all three measures should yield the same GDP figure, but due to statistical noise and other real-world factors, discrepancies can arise.

Most look at GDP from the demand side estimates – the first of the three above, and also the first the BEA releases (usually one month after the end of each calendar quarter).  The second estimation by adding up all incomes generated – and which the BEA refers to as Gross Domestic Income (GDI) to distinguish it from GDP even though it should in principle be the same value – is usually released two months after the end of each calendar quarter.  The third – of production by sector – is usually released three months after the end of each quarter, and sometimes later.

It is this third set of estimates that is of interest here.  They provide a breakdown of GDP by sector, and can be found in the “GDP-by-Industry” section of the online NIPA accounts.  Formally, the measure is of the value added produced in each sector (that is, the total or gross output of the sector less the purchases of intermediate products used in that production), where the sum of the value added across all sectors equals overall GDP.  That sector value added is often loosely referred to as sector output or even sector GDP.  I will generally refer to it here as output, and real output refers to the value added in terms of the prices of 2017.

The question of interest is whether sectors that may have benefited most from the boom in AI investments accounted for the acceleration in the growth in labor productivity observed in 2025.  It is important to be clear in distinguishing between the boom in AI investments being made – a demand side matter – from sectors that may have made use of those new AI systems – a supply side matter.  The prior posts on this blog that examined the impact of the boom in AI investments on GDP in 2025 looked at the demand side impacts.  Most of the demand side impetus to GDP in 2025 came from the massive investments being made in new data centers and other facilities – as well in software – to support making AI available.  The question now is whether making AI available and increasingly effective may have led to increases in productivity in sectors that could make use of those investments.

The data on sector outputs suggests that this may have been the case.  I should hasten to add this is not proof, as this is only data on what has happened at a broad sectoral level, and cannot identify what the specific micro-level mechanisms were that led to these overall outcomes.  It is also only one year of data.  But they may be providing an early hint that AI is having an impact on productivity in certain sectors.

An issue is that the BEA sectors are broad, and thus include sub-sector activities where AI could have a major impact as well as others where it would not.  But within the 14 major sectors of the economy that the BEA distinguishes (and where the BLS provides comparable employment data), Information and Finance are two sectors where AI might be expected to have a significant impact.  Information includes data processing activities as well as internet publishing, although it also includes traditional publishing, movie-making, and broadcasting.  Finance includes banking and other such financial activities, but also real estate and rental activities.  Information and Finance together accounted for 27.4% of GDP as of the fourth quarter of 2025, with the rest accounting for 72.6%.

The growth in labor productivity in those two sectors in 2025 was exceptional:

Chart 2

Real output (i.e. real value-added) in those sectors grew at a 4.9% rate in 2025, up from a 3.0% trend in the years before.  And employment in fact fell in 2025, at a rate of -0.2%.  With fast growth in sector output while employment declined, output per unit of labor (labor productivity) rose in 2025 at a 5.1% rate, up from 1.9% in the years before.

This may be a hint that AI is having an impact.  Information and Finance are sectors where one can envision AI allowing more to be produced while requiring less labor.  It is of course not proof, as these are simply observations on the changes in 2025 in the aggregates for the sectors.  But it is consistent with an interpretation that AI may be having an impact here.

For the rest of the economy other than Information and Finance, labor productivity grew at a slower pace in 2025 than in the years before:

Chart 3

Real output in the sectors other than Information and Finance (and accounting for almost three-quarters of GDP) grew at a pace of only 0.9% in 2025 – down from 2.3% in the years before.  Employment rose at a rate of 0.2%, down from 1.3% in the years before.  Together, this meant the growth in labor productivity fell from a rate of just below 1.0% in the years leading up to 2025, to 0.6% in 2025.

The increase in labor productivity in 2025 – shown in the chart at the top of this post – can therefore be attributed at least in part to the significant increase in productivity in the Information and Finance sectors in the year.  In the almost three-quarters of the economy other than Information and Finance, productivity grew at a slower pace than it had before.

Finally, the similar calculations for a more narrowly defined sector – Computer System Design – are of interest:

Chart 4

Computer System Design is a small sector – accounting for only 1.8% of GDP – and is a sub-sector within the Professional and Business Services major sector of the BEA.  The Professional and Business Services sector is broad, and includes services from such highly-paid occupations as legal, consulting, and managerial services, to more mundane services such as those from janitors, groundkeepers, and in waste management.  One can envision that AI may be having a strong impact on the services provided in computer system design, but not so much in janitorial and similar services.  Thus the focus here is only on the former.

Output in Computer System Design was rising at a fast pace even before 2025 – a rate of 9.0% per annum – and then grew even faster at a 10.8% rate in 2025.  Employment also grew at a relatively rapid pace of 3.2% in the years leading up to 2025 – a good deal faster than the 1.3% pace of employment growth in the economy as a whole in those years.  With output rising at a 9.0% rate before 2025, labor productivity was growing at a 5.6% rate – much faster than the 1.2% pace of productivity growth in the economy as a whole in those years.  The sector has seen rapid productivity growth for some time.

Productivity then rose by substantially more in 2025.  Real output rose by 10.8% in the year.  Despite this, employment in the sector fell by 2.1%.  The implication is that labor productivity rose by an exceptional 13.2% in 2025.  Those employed in computer system design and related services could produce a good deal more than they could before, which is consistent with AI-enabled productivity gains.  But as anecdotal evidence has suggested, it has become extremely hard to obtain a new job in the field.

D.  The Impact of Shifting Sector Compositional Effects on Productivity Growth in 2025

All of the discussion on AI-enabled productivity gains (at least all that I am aware of) has been on what AI might make possible for a given sector or occupation.  The discussion above was similar, with a focus on a few key sectors.  There may, however, be a different source of growth in the productivity figures for the economy as a whole, i.e. for overall GDP.  Specifically, some of the sectors that have seen an acceleration in their growth – possibly due to AI – may also be sectors where labor productivity is especially high.  With such sectors growing faster than overall GDP, they will account for an increasing share of GDP.  And labor productivity in the economy as a whole will then increase due to their increasing weight in GDP – a compositional effect separate from what may be happening to productivity in the individual sectors alone.

Labor productivity – real output per employee – differs markedly across the different sectors of the economy:

Chart 5

The figures are for the fourth quarter of 2024 in order to focus on the base levels before the growth in 2025 (with levels at end-2025 that will, in any case, differ only by a small amount as growth in a year is a matter of only a few percentage points).  The levels vary greatly, from a few sectors where real output per employee (in 2017 prices) was over $500,000 (Mining, Utilities, Information, and Finance), to as low as $53,000 (in Arts, Entertainment, Recreation, Accommodation, and Food Services) – a difference of almost a factor of ten.  Overall real output (GDP) per employee was just below $150,000.

The range across the different sectors of the economy is wide.  Some sectors do not employ many compared to the other investments they need:  Mining and Utilities are prime examples.  Other sectors are much more labor intensive – the leisure and hospitality fields, for example – where total output (value added) per employee is relatively far less.  With such wide variation across sectors, the growth in labor productivity in the economy as a whole will be sensitive not only to what might be happening to productivity in the individual sectors, but also to what is happening to the mix of sectors that make up GDP when some sectors are growing faster than others.

As noted above, output growth was substantially higher in 2025 in the Information and Finance sectors than the output growth in the rest of the economy:

Chart 6

While this repeats material from the charts above, showing them all on the same scale makes clear how very different the growth rates were.

Labor productivity in those sectors also differed markedly from each other:

Chart 7

One can isolate the impact on productivity of the changing sector composition of GDP by calculating what would have happened to overall labor productivity in a case where sectors grew as they had in 2025 but with no change in productivity in the individual sectors themselves.  Of interest here is what may have been due to growth in 2025 in the Information and Finance sectors in comparison to the rest of the economy.  Any change in overall productivity would then be due solely to the resulting changes in sector weights in overall GDP:

Impact from Compositional Effects on Overall Labor Productivity

Growth Rates 2025

Actual

Compositional Effect on Productivity
Real Output:
  All (GDP)

1.99%

1.99%
  Information + Finance

4.86%

4.86%
  All Other

0.88%

0.88%
Employment:
  All (GDP)

0.20%

1.18%
  Information + Finance

-0.23%

4.86%
  All Other

0.24%

0.88%
Labor Productivity:
  All (GDP)

1.79%

0.80%
  Information + Finance

5.11%

0.00%
  All Other

0.64%

0.00%

Compositional Effect:  Calculates the change in overall labor productivity arising from shifts in the sector composition of GDP only.

This table presents the figures for the simple case where the sectors have been aggregated to just the two discussed above – to Information and Finance as one and everything else as the other.  The first column presents the figures as they actually were in 2025.  The second column sets labor productivity growth in the two individual sectors at zero.  Employment in each would then need to grow at the same rate as output growth in each.  One can then add up total output (GDP – the same as the actual in 2025) and total employment (the sum of what would then be needed in each of the sectors) to find that the overall growth in labor productivity would have been 0.80% simply from the change in the sector composition of GDP over the course of the year.

Labor productivity in the economy as a whole grew in this calculation despite no productivity growth in the individual sectors since the Information and Finance sectors grew relatively fast in 2025 and labor productivity in those sectors is substantially higher than what it is in the rest of the economy.  In 2025, the 0.80% coming from this shift in sector composition accounted for 45% of the growth of 1.79% in overall labor productivity in the year – close to half.

The consequences of such shifts in sector composition on labor productivity are typically ignored in discussions of what has happened to productivity in the economy.  This is understandable, as one can easily calculate what happened to overall labor productivity by dividing the growth in GDP in the period by the growth in total employment.  The initial (Advance) estimate of GDP is provided by the BEA just one month after the end of each calendar quarter, while the employment figures for the period are provided by the BLS even earlier.  Sectoral output figures are available only several months later, and by that time interest has shifted to what will be published for GDP for the next quarter.

The effect may be large.  The impact will differ in different time periods depending on how much faster or slower the various sectors are growing at, coupled with whether the faster (or slower) growing sectors are sectors with relatively high or relatively low labor productivity levels.  In 2025, the Information and Finance sectors grew relatively fast (possibly due to their ability to make good use of the new AI systems), and they are also sectors with far higher labor productivity levels than on average in the rest of the economy.  Their growing share in GDP then led by itself to substantial growth in average labor productivity in the economy as a whole.

E.  Summary and Conclusion

GDP grew at a rate of 2.0% in 2025.  That is not especially high, but nor is it zero.  Yet employment grew hardly at all.  The question is why.

We now have data at the sectoral level that allows a deeper examination of what has been going on.  A few sectors – specifically in the groups of the BEA that cover Information Services and Financial Services – saw especially rapid labor productivity growth.  Their production per person employed was already high, and it then grew even faster in 2025 than it had in the years leading up to 2025.  The output of those sectors also grew in 2025 at a substantially faster pace than output in the rest of the economy, leading to their share in the economy growing.

Both the rising productivity in those sectors and their rising share in the economy led to an increase in average labor productivity in the economy as a whole.  The shift in sectoral composition – a factor that is typically ignored in these discussions – accounted for close to half (about 45%) of the increase in labor productivity in the economy as a whole.

An alternative hypothesis set out by some for why employment was close to stagnant in 2025 despite the modest GDP growth centers on the sharp decline and possible reversal of net immigration into the US due to Trump’s anti-immigrant policies.  Under this hypothesis, GDP still grew (by 2.0%) despite the reduction in immigrant workers because firms were able to increase production from (i.e. raise the productivity of) the workers they still had to offset this.  If this were true, then one would see a general increase in productivity broadly across the economy, and not just in the Information and Finance sectors.  But this was not the case.  Labor productivity outside of those two sectors rose more slowly in 2025 than it had in the years leading up to 2025 (i.e. at a 0.6% rate in 2025 versus 1.0% per annum in the years leading up to it).  See Chart 3 above.  There is no sign that a shortage of workers (if it indeed existed) led firms to adopt approaches that would accelerate the pace of productivity growth.

Labor productivity rose sharply, however, in Information and Finance in 2025.  The data on this are clear.  But while the data can point to where the increases in productivity arose, the aggregate figures cannot in themselves tell us what was behind this. The Information and Finance sectors are, however, ones where it is plausible that the availability of the new AI systems may be enabling workers in those sectors to produce substantially more than they could before.  It might also be a key underlying factor in why those sectors grew especially fast in 2025 (at a 4.9% rate) compared to growth in the rest of the economy (a 0.9% rate).

These may be early hints that AI is having an observable impact on productivity as measured at aggregate levels.  It is still early, of course, and one will need to see whether this is sustained over time.  But it does provide a plausible explanation for why employment grew so slowly in 2025 despite the growth in GDP in the year.

More Evidence on the Damage Trump’s Policies are Doing to the Economy

Posted on by

Chart 1

A.  Introduction

On May 28, the Bureau of Economic Analysis (BEA) of the US Department of Commerce released its Second Estimate of GDP for the first quarter of 2026.  Along with it, it released its estimates of Personal Income and Outlays for April 2026.  Together, they provide further evidence on the damage that Trump and his misguided (as well as erratic) policies have done to the US economy.

This note will review some of the figures that came out.  The chart above shows in a longer-term context what has happened to real per capita disposable personal income – perhaps the best measure in the GDP accounts of average real incomes of Americans.  It stagnated in the first year of Trump’s return to the presidency and is now falling in 2026.  It is also now well below what it would have been had it continued to follow the rising trend path of the last 13 years.  The figures will be discussed in the next section below, as well as figures on the divergent paths of what has happened to wages and salaries (stagnant in real terms) in contrast to corporate profits (up by 12.0% in the first quarter of 2026 over the year earlier in nominal terms, and by 8.7% in real terms).

The section that follows will then discuss a few points that can be found in the new GDP estimates.  GDP growth in the first quarter was weak, with a revised estimate that real GDP grew at a 1.6% annual rate in the quarter (down from 2.0% in BEA’s initial estimate released in April).  But this includes the effect of the return to normal levels for a full quarter of government production following the end of the federal government shutdown in the fourth quarter of 2025.  That recovery already happened in mid-November.  The previous post on this blog discussed that impact and how it is measured.  The bounce back to normal levels led to GDP as measured that was 0.6 percentage point higher in the first quarter than otherwise by my calculations (and 1.0 percentage point higher in figures cited by the BEA when discussing the negative impact of the shutdown in the fourth quarter).  Excluding this impact of government workers returning to their offices, GDP growth in the first quarter would have been only 1.0% (using the 0.6% figure) or just 0.6% (using the BEA figure).

Furthermore, more than all of this growth was a consequence of the AI boom.  The contribution to the growth in GDP in the first quarter from private investment in information processing equipment and software totaled 1.4 percentage points in the BEA figures.  That is, after taking into account the impact on measured GDP from government workers returning to their offices for the full quarter and private investments linked to the AI boom, production in the entire rest of the economy fell.  Production in the entire rest of the economy other than AI investments would have led to a fall in GDP at a rate of – 0.3% using the 0.6% figure for the impact of the government shutdown (or at a rate of – 0.7% using the 1.0% figure the BEA cited for the impact of the government shutdown).

On top of this, inflation is now high.  As discussed in Section D below, the upturn in inflation started already in late 2025 / early 2026, i.e. before Trump’s decision to start a war with Iran on February 28.  The resulting jump in fuel prices led to inflation being even higher.

The economy is doing poorly.  Living standards are falling.  Only investments linked to the AI boom are keeping GDP growth positive.

B.  The Impact on Living Standards

Per capita disposable personal income in real terms was stagnant in the first year of Trump’s second presidency and falling in 2026.  It is now well below where it would have been had it continued on the previous upward trend.  The figures are shown in the chart at the top of this post.

The BEA provides an estimate of personal income monthly, and it can be found with its underlying components in Table 2.6 of the NIPA Accounts.  Personal income includes all sources of income accruing to individuals, including from wages and salaries (along with supplements to wages, such as company contributions to health and pension plans), income from unincorporated businesses (sole proprietorships and partnerships – i.e. most small businesses), rental incomes accruing to persons, personal interest income and dividend income, and current transfer receipts (such as from Social Security and Medicare) net of taxes paid for such programs (e.g. Social Security and Medicare taxes).

Personal income minus personal taxes (primarily income taxes) will then be disposable personal income.  The BEA deflates these figures using its estimates of the personal consumption expenditures price index (often referred to – not quite correct technically, but close – as the PCE deflator) to put them in real terms, and divides them by current population levels (with estimates from the Census Bureau) to put them in per capita terms.

Per capita disposable personal income in real terms was close to its long-term trend in January 2025, as Trump took office, and continued close to that trend until April 2025.  But that was the month when Trump announced huge and essentially arbitrary tariffs would be charged on imports on almost every country and region in the world (including an island populated only by penguins and seals).  He called this “Liberation Day”.  Erratic changes in tariffs since then, as well as in other policies (such as the granting of special favors or special penalties to various firms depending on Trump’s whims), have since continued.  Real personal income then came down from its April 2025 peak, stagnated to the end of the year, and fell to just $52,330 in the BEA estimate for April 2026.  This is below where it was when Trump took office, and $750 below where it was in April 2025.  This is in 2017 prices.  In current prices and as of April 2026, real personal income (at an annual rate) is now $980 per person less than it was on “Liberation Day”.

But a more appropriate measure of performance would be relative to where it would have been had it continued to rise as it had under Biden and before.  Compared to what it would have been, the shortfall in living standards by April 2026 came to $1,700 per person in terms of 2017 prices, or $2,200 for every man, woman, and child in the country in current prices.  For a family of four, the reduction in living standards as of April 2026 was $8,800 at an annual rate.  This is not a small amount.  Households could make good use of the higher income they would have had, had it continued to grow as it had under Biden and before.

Furthermore, the gap between what it could have been and what it actually has been under Trump is widening over time.  It is also an average, and hence does not take into account the increases in inequality of recent years.  There has been much discussion of the so-called “K-shaped” economy, where higher-income individuals are doing increasingly well while lower-income individuals are doing poorly.  With growing inequality, the reduction in the overall average real personal incomes under Trump has been especially stark for the lower and middle income classes.

Defenders of Trump might well point out that there was also a substantial dip in real personal incomes in 2022 during the Biden administration.  This is true and is seen in the chart at the top of this post.  It was, however, temporary.  Real personal incomes returned to their previous growth path by the end of that year, and then continued on that path until Trump took office.  The dip was a consequence of the severe disruptions to the US (and indeed world) economy due to the sudden lockdowns due to Covid in 2020 that continued into 2021, and then the time needed to re-establish the regular functioning of supply chains once the production plants and transportation networks could be reopened.  The impact of this on disposable personal incomes in 2020 and 2021 was masked by the numerous (and massive) emergency government support programs under both Trump and Biden – as seen by the sharp upward spikes in personal incomes in those years.  Much of this was saved (stores were often still closed), and the drawdown on such savings could then support purchases in 2022 despite real incomes being temporarily low while supply chains were still not fully functioning.  Real personal income then rapidly recovered, and by late 2022 it was back to its prior trend.

Another indicator in the recently released BEA estimates of the increasing stress that American households are experiencing can be found in the estimates of the personal savings rate.  This is also provided in Table 2.6 of the NIPA accounts.  The personal savings rate is personal savings as a percentage of disposable personal income.  That rate has been falling during Trump’s second term to just 2.6% as of April 2026 – less than half the rate of 5.5% of April 2025.  It is also now well below its recent longer-term average.  Between January 2013 and February 2020 (before the Covid disruptions began), it varied between about 5% and as much as 8%, and averaged 5.9%.

The 2.6% rate is low, and the fact it has been falling is an indication that households are stressed.  Given urgent current needs, they are saving less for retirement and other future objectives.  As with personal income, the BEA can only estimate personal savings as an average over all households.  Thus the 2.6% rate is an average that includes both upper income households who are likely saving a relatively high share of their income and lower and middle income households, who may not now be saving much at all.

At the same time as personal income has been falling, corporate profits have been rising at a fast rate.  The BEA estimates corporate profits only on a quarterly basis, and the initial estimates of these profits are released only with the release of the second estimates of the GDP accounts each quarter (as in the estimates released on May 28).  See specifically Table 6.16D in the NIPA Accounts.  Between the first quarter of 2025 and the first quarter of 2026, corporate profits in all industries rose by 12.0% in nominal terms.  Using the PCE deflator to put this in real terms, the increase was 8.7%.  In contrast, wages and salaries rose by just 3.5% in nominal terms between those two quarters, or 0.4% in real terms using the PCE deflator.  Adjusting also for population growth, the increase was essentially zero (less than 0.1%).

Corporate profits have been going up, and at a rapid pace.  Wages have not.

C.  The Growth in GDP in the First Quarter of 2026

The BEA’s estimate of GDP growth in the first quarter of 2026 was revised down from 2.0% (at an annual rate) in the BEA’s initial (“Advance”) estimate released on April 30 to 1.6% in the Second Estimate released on May 28.  But as noted above, this 1.6% rate includes the impact of the bounce-back to normal levels of federal government production of services for a full calendar quarter.  It had been curtailed during the shutdown that spanned almost one-half of the fourth quarter of 2025, and GDP measures the flow of goods and services provided over a full quarter.  Taking this effect into account, growth in the first quarter of 2026 was even less.

The impact of the government shutdown was discussed in the previous post on this blog.  GDP is the sum total of a flow of goods produced and services provided during a period of time (a calendar quarter here), and the reduction in the provision of those government services in the first half of that quarter meant a reduction in GDP in the quarter.  As discussed in that blog post, the impact (by my calculations from the figures the BEA provided) reduced measured GDP by about 0.6 percentage points (at an annual rate) below what it otherwise would have been.  The BEA, in commentary it provided with its releases of the GDP estimates for the fourth quarter of 2025, indicated the impact was about 1.0 percentage point of GDP.  The reason for the discrepancy is not clear, but one guess would be that some higher official at the BEA or the Department of Commerce took the 0.6% figure and rounded it to 1%, and that someone else started to write this as 1.0%.

With either figure, GDP in the fourth quarter of 2025 was reduced by some amount.  By simple arithmetic, there would then be a bounce-back effect on GDP in the first quarter of 2026 of a similar magnitude, as the government returned to normal operations for the full quarter.  Taking this into account, the rate of growth in GDP in the quarter other than from this return to normal government operations would have been 1.0% rather than the 1.6% reported (or 0.6% rather than 1.6% based on the 1.0% figure for the impact of the shutdown that the BEA cited).

But in addition, GDP growth – such as it was – is more than fully accounted for by the continuing boom in private investments linked to building the data centers, developing the software, and supplying the other equipment needed for the new artificial intelligence (AI) systems.  This AI boom accounts for much of the growth in GDP in 2025, with this continuing into 2026.

While the NIPA sector categories will not match precisely the investments related to the AI boom, a reasonable approximation is the sum of private investments in information processing equipment and in software.  The NIPA accounts provide figures for private investment in these categories, and from this the BEA provides figures (in Table 1.5.2 of the NIPA accounts) of the contribution from the growth of each to the overall growth in real GDP.  For technical reasons (the use of chain-weighted price indices), the sum of the individual contributions to the growth in GDP may differ slightly from the estimated growth in real GDP, but they are well close enough for the purposes here.  Of greater importance is that investments in information processing equipment and in software will be for more than that just for AI, plus there will be AI-linked investments in other categories as well.  These will in part offset each other.

What is clear is that in 2025 and continuing into 2026, there has been a major increase in private investment in these AI-related categories.  Their contribution to the growth in GDP in the BEA calculations (Table 1.5.2 in the NIPA accounts) was an average of a 0.90% point contribution to the GDP growth rate each quarter (at annual rates).  This is triple the average contribution to GDP growth of investments in information processing equipment and in software between the first quarter of 2013 and the last quarter of 2024, when its contribution was on average 0.30% point.

Subtracting from overall GDP growth the contribution of the AI boom, as well as accounting for the impact of the federal government shutdown, yields the contribution to the growth in GDP of the entire rest of the economy:

Contributions to GDP Growth

GDP Growth Contribution of      Info Processing                   + Software Impact of Gov’t Shutdown Contribution of All Else
2025Q1  -0.65%        1.30%     -1.95%
2025Q2   3.84%        0.80%      3.04%
2025Q3   4.38%        0.26%      4.12%
2025Q4   0.48%        0.78%   -0.57%      0.27%
2026Q1   1.62%        1.36%    0.56%     -0.31%

Seasonally adjusted annual rates.

(The figures for the impact of the federal government shutdown (-0.57% of GDP and +0.56% of GDP) have been rounded in the text to 0.6%, and are shown here at two digits of accuracy to be consistent with the rest of the table.  Also, they differ very slightly between the two quarters – 0.57% vs. 0.56% – as the impact is taken as a share of GDP, and GDP is slightly higher in the first quarter of 2026 than what it was in the fourth quarter of 2025.)

Taking into account the impact of the government shutdown and of the boom in AI investments, growth in the rest of the economy was essentially zero over the past half year.  It was relatively high in the second and third quarters of 2025, but was substantially negative in the first quarter.  While the quarter to quarter figures will bounce around (due both to real changes and to statistical noise), the economy – other than for investments related to AI – is clearly weak.  This is consistent with the findings discussed above on the stagnation in real personal incomes in 2025 and its fall in 2026.

Another sign of weakness in the US economy has been a continued decline in private investment in business structures (e.g. office buildings, commercial structures, warehouses) and in residential housing.  See Table 1.1.1 of the NIPA accounts.  Each has declined in real terms in every quarter since Trump took office at the start of 2025, most recently with real investment in business structures falling at an annual rate of 5.4% in the first quarter of 2026 and real investment in residential housing falling at a 6.2% rate in the quarter.  Other than for AI, private investors are wary of committing to investments in the economy.

A proviso on the AI investments should, however, be noted.  The figures above are based on the BEA calculations of what it terms the “contributions to the percent change in real gross domestic product”.  It is, however, a calculation from the demand side measure of GDP, where all the components of demand for GDP (private consumption, private investment, government, and exports less imports) are added up.  This provides an estimate of domestic production during the period, as private investment includes investment in inventory accumulation and changes in inventories act as a balancing item.  Increases in imports are therefore a negative contribution to the growth in GDP in this framework, and the BEA is only able to make an estimate of the change in total imports during the period – not imports that in some way both directly and indirectly provided part of the supply to fill a specific demand.

With imports equal to only about 14% of GDP, the approach is not unreasonable, as 88% of what is used to fulfill the various demands will come from domestic production.  (With imports at 14% of GDP, total supply will be 100 + 14 = 114, and the share domestically supplied will be 100 / 114 = 88%.)

But while the average import share in total supply is 12% ( = 14 / 114), the share is likely substantially higher for the investments linked to the AI boom.  How much higher is not clear.  Many of the semiconductor chips and much of the specialized equipment are imported, but the investments in the data centers supporting AI and in the software used for this will be more than just imports.  The data centers need to be built, the equipment put together, and the centers then connected to power, water, and information networks.  And the software, in contrast to the chips, is primarily from domestic production.

The relatively high share that is imported will matter for the impact such AI investments will have on domestic production rather than direct imports, and GDP is a measure of domestic production.  It is impossible to say how much that impact will be, but it will reduce the “contribution” of such investments to the growth in GDP (as depicted in the table above).  However, even with this, the contribution of the “all else” category to the growth in GDP is likely still to be small – just not as small as the figures indicate.

D.  Inflation is Now High

Inflation is now also a concern.  Table 2.8.4 of the NIPA accounts provides monthly estimates of the price indices estimated by the BEA for personal consumption expenditures – both overall and for the major types of products making up personal consumption.  (Technically these are price indices rather than price deflators, but in practice they are almost always the same within round-off and the terms – price indices or deflators – are often used interchangeably.)  The Fed uses the core PCE deflator (the deflator excluding food and energy) as the primary indicator of inflation that it focuses on, with the objective of keeping it at around 2.0% on an annualized basis.

Monthly changes in the price indices are volatile and often not meaningful, while changes in the indices over year-earlier periods will miss turning points due to the long lag.  Changes over six-month periods are usually a good compromise to show when a turning point has been reached.  And it is clear from this that inflation turned decidedly higher in late 2025 / early 2026:

Chart 2

The overall PCE price index over the six months ending in April 2026 rose at a 4.8% annualized rate.  The core PCE price index rose at a 3.8% pace.  Both of these are now far above the Fed’s 2.0% goal.  And this is not just due to energy prices:  By April, the six-month core PCE price index had risen by a full percentage point from the 2.8% rate of the six-month periods ending in late 2025.  Furthermore, energy prices in the months of January and February 2026 were in fact relatively low and below their levels of the last several months of 2025.  Trump did not launch his war against Iran until February 28, after which energy prices skyrocketed.  This then compounded what was already becoming an inflation problem.

Inflation by itself will not necessarily lead to a reduction in average real personal incomes in the NIPA accounts – the topic of Section B above.  Higher prices mean that the loss of one party is a gain to another.  And the stagnation in real personal incomes began in 2025 well before the recent jump in inflation.  But to the extent the inflated prices end up benefiting corporate entities (such as the big oil companies), average real personal incomes will be reduced as corporate profits go up.  This has likely been an additional factor in the more recent fall in 2026 in the absolute levels of average real personal incomes.

The recent rise in inflation does not in itself account for the slump in living standards under Trump.  The stagnation in real personal incomes was already underway in 2025.  Trump’s misguided policies led to that.  High inflation is now compounding those difficulties.

E.  Conclusion

There is another figure in the recently released NIPA accounts that is of interest as an indicator of what has happened to the living standards of lower-income Americans.  It has in fact had a positive contribution to GDP as mechanically measured.  Included within the goods and services that add up to overall personal consumption expenditures, the BEA has the category labelled “Final consumption expenditures of nonprofit institutions serving households (NPISHs)”.  These are the net expenditures of nonprofit groups serving lower-income households, such as food banks, health clinics, and other providers of similar services.  The “net” is net of any payments they receive from those receiving those services.  Table 2.8.11 in the NIPA accounts shows the percentage change in real expenditures on this consumption category over the same month one year before.

The net consumption of these goods and services provided through nonprofits was 10.6% higher in real terms in April 2026 than what it was in April 2025.  This is major growth (and a contribution to GDP as measured), and is the highest percentage increase since 2022 (when the disruptions of the Covid crisis were being finally resolved).  This need to resort to food banks and other services provided through non-profits is another indication that lower-income households are stressed in this economy, and need to find support somewhere.

This indicator of stress among American households is consistent with the stagnation – and more recent decline – in real personal incomes shown in the chart at the top of this post.  It is also consistent with the fall in the average personal savings to just 2.6% – half of what it was when Trump took office.  When times are difficult, households set aside their savings plans.  It is also consistent with slow growth in GDP outside of investments in the booming AI sector.  And it is consistent with the more recent rise in inflation – affecting some households more than others – where the inflation rate was already going up before Trump chose to bomb Iran and drove up the price of fuels.

Trump’s policies are doing real damage to the economy and to living standards, that are evident in data that cover only a little over a year since he took office in his second term.  But there is no indication that Trump recognizes this and that he intends to change what he has been doing.