Second quarter GDP was below our expectations despite strength in consumer spending and in business investment. Other data that arrived during the past month were mostly positive until the employment data, which was weaker than expected. The result is that our forecast for economic activity over the next year is a little stronger than three months ago.
NIPA Data
The advance release of 2026Q2 of national income and product account data put real output growth at a disappointing 1.5%. (See chart.) This was below the rate for Q1 (2.1%) and even more for the average over the previous year (2.7%). It was also less than our 2.0% forecast in May (the first using Q1 data).
Beneath this top line, however, the situation is more complicated.
On the positive side, domestic final demand was strong. Both consumption and investment had growth that was significantly better than that over the previous year. For consumption this strength came primarily from spending on goods. Investment growth was in business spending on equipment and intellectual property, while spending on structure was significantly negative (data centers notwithstanding). Residential investment registered weak growth, which after eight quarters of negative results was a definite improvement. Government spending was a little short of flat. In total, consumption, investment and government added 3.4% to Q2 GDP.
But, on the negative side more than half of this was offset by inventory shrinkage (-0.6%) and especially by a larger trade deficit (-1.2%). The latter came from a surge in imports of goods, which swamped a solid rise in exports.
To summarize, the second quarter had strong growth from the private sector – significantly better than the average over the previous four quarters. In both periods government was basically flat, while inventories were a negative. And imports shifted from contraction to expansion – a negative for domestic production.
Recent Monthly Data
The mixture of positives and negatives in the NIPA data also extends to higher frequency data released during July and early this month.
Starting with data for June: Industrial production had weak growth as in May. Housing starts rose significantly, from a very weak May level, but only into the stagnant range that goes back to 2022. And real household income and consumption both had solid gains. Moreso for the latter, causing saving rate to decline to just 2.7%
Turning to July/August data. The University of Michigan consumer sentiment index rose in July as it had in June but then fell back to its June level in their preliminary August report.
And then there is the employment report for July. The household survey had its headline unemployment rate declining another tick to 4.1%. And the number of unemployed workers decreased by 178 thousand. Sounds good. Except that the labor force fell by 264 thousand. So, as was also the case in June, the drop in the unemployment rate came from unemployed workers who stopped looking for work, rather than unemployed workers who found jobs. The household measure of employment decreased by 87 thousand in July.
But this was nothing compared with the establishment survey, which had an employment decrease of 23 thousand in July plus -103 thousand in revisions to May and June. Looked at differently, a month ago the 3-month job growth (April – June) averaged 111 thousand per month. Now, with the revisions to May and June, and with July replacing April, the average is just 20 thousand.
Grasping for a silver lining, a lot of the decrease came in the government sector. Specifically, a July decrease of 53 thousand, and May/June revisions of -48 thousand, nearly all at the state and local level. The private sector thus added 30 thousand employees in June, offset by -55 thousand in revisions.
Overall, a month with a few numbers that are encouraging, a couple that are neutral, and a labor market report which is concerning at best.
Baseline Forecast
As shown in chart below, our revised forecast for output is stronger than in May although not much different from our update to that forecast a month ago. The update was motivated by our conclusion that the AI investment boom would be of some duration. We see nothing in the Q2 data to change that perception. In our current outlook we have real GDP growth averaging 2.5% over the upcoming year. This growth rests on continuing solid growth in consumer spending and AI investment, although in both cases somewhat below the rates in Q2. Growth in government spending turns barely positive. This expansion, together with some growth in inventories, about offset continuing increase in the trade deficit. The latter, however, is not as extreme as in Q2.
Our view of the labor market reflects the dramatic shift in labor force dynamics since the beginning of this year. Over this period the participation rate has decreased by a full point – to just 61.4% in July. We think this rapid plunge is an aberration (more on this below), but even at a much slower rate, it implies slower growth in employment.
We have introduced such a shift into our forecast. Over the next year we now have employment growth that averages about 56 thousand per month, down from about 69 thousand in our May outlook. Relative to the recent past, this still looks optimistic.
Discussion
We think the forecast outlined above is a plausible possibility of the average path for the economy over the next year. The likelihood that the economy will be close to our forecast on a quarter-by-quarter basis is vanishingly small. There are always unexpected events that hit the economy.
In addition, there are other plausible paths – lots of them. We think (hope?) that our specific path is about in the middle between more optimistic and more pessimistic alternatives. Even if that is true, it doesn’t mean that we think our forecast path necessarily has a higher probability than others on either side. Rather we think that the probability curve around our forecast is quite flat over a significant range of average growth outcomes – at least from 3%+ to 2%- for the upcoming year.
The obvious line of inquiry this view leads to is discussion of the forces that will determine which path the economy actually follows. That is, discussion of Rumsfeld’s “known unknowns.” The problem is that this is a long list – and getting longer. The top three on our current list are: 1. The current chaotic policy environment; 2. The AI revolution; 3. The labor supply situation.
Starting with the policy environment. This is an unknown that currently operates in the short-term, meaning that the situation today could change (a lot) a month from now – and then maybe to something more like where we are now a month after that. The two areas that are most relevant to the economy are probably tariffs and the horrible mess in the Middle East. The elevated uncertainty from either is a negative for the economy, probably one that we will live with for a long time. We have little hope that the upcoming election will improve the situation – more likely the opposite. Ditto for the end of the Trump era in a little over two years.
Second, artificial intelligence. In the short-run (that is, the next year) AI seems likely to be a positive for the economy, which is a significant element in our forecast. The major risk that we see to this optimistic evaluation is from the financial markets. Over the past year AI investment has shifted from being financed primarily from free cash flow to external funds. As J.P. Morgan said when asked what the stock market will do, “It will fluctuate.” A fluctuation down in AI tech stocks could be a very unpleasant event.
Over a more extended horizon (saying the next three or four years) we will likely learn if the much-hyped productivity upside of AI is a real thing. Assuming it is that could, over this time frame, be a mixed blessing. The upside is obvious – more and better goods at lower cost. The downside is less certain depending on how AI users implement the technology. Will it be used mostly for labor-replacement or labor-augmentation? Current evidence (mostly anecdotal) is mixed. So far, AI is a very large known unknown.
Third, the labor supply. The basic macroeconomic theory of aggregate supply is quite simple: Growth is result of increase in the supply of labor and growth in labor productivity. The latter comes from investment in labor-augmenting capital equipment (like AI) and from education (including OJT) that increases “human capital.”
In the relatively recent past, the growth of the labor supply was viewed as both stable and predictable, driven by past demographic developments: birth numbers 20+ years earlier minus the number of workers approaching retirement plus (for the U.S.) a stable flow of immigrants. But recently this is no longer the case. To begin with, the pandemic accelerated the orderly progression into retirement. This slowed the rebound in the labor force when the economy reopened, but then probably increased labor force growth for a couple of years as retirement reverted to a more normal pattern. During roughly the same time frame, the immigration situation turned to anything but stable – presumably increasing the labor force during much of President Biden’s term and then moving to the opposite extreme under President Trump (a third item in the chaotic policy list). Our model equation for the labor force includes demographic variables intended to capture factors like the aging out of baby-boomers. The equation currently implies labor force growth of about 60 thousand per month. It does not include any variables related to immigration.
One final problem. Rumsfeld’s full quote mentions known knowns, known unknowns, and unknown unknowns. This leaves out a category – unknown knowns. Specifically, things we think we know that in fact are wrong. This could easily be the case with the labor supply. Given the huge revisions, both month-to-month and in benchmarking, we are increasingly skeptical of the data from both BLS labor market surveys.