Relative to the initial Q2 data that were released in late July, the much-delayed Q3 numbers were clearly better than expected. The strength came from both private sector and government spending but did not produce a parallel boost for the labor market. Our revised forecast is stronger than three months ago, both near-term and longer-run.
NIPA Data
The “initial” GDP data for Q3 produced late last month combined the advance and 2nd releases for a normal schedule. As shown in the chart, top line growth was a very strong 4.3%, an increase from Q2 and well above our August expectation of 2.4%.
Moreover, this unexpected strength was across most of the private sector and some of government. The primary contributor was very robust consumer spending. Some, especially for goods purchases, may have been on items imported prior to tariff implementation, and some induced by equity market wealth effects. There are several caveats (see next section), but still an impressive result.
Investment and government were more complex. For investment, purchases of equipment and intellectual property products added 0.6% to GDP, due in large part to AI spending and far better than we expected. Investment in business structures and housing, on the other hand, reduced GDP by almost 0.4%, which was in line with our August forecast. For government, both federal nondefense purchases and state/local spending were weak – as we had forecast. But defense was stronger than we anticipated, with growth at a 2.8% rate.
Turning (reluctantly) to trade: Imports were less weak than we expected (a negative for GDP) while exports were stronger than we had forecast (a positive for GDP). The result was an improvement in the trade balance of $101 billion, almost exactly in line with our August forecast of $102 billion. [Sorry about these two sentences. The impact of trade in GDP accounting is inherently incomprehensible.]
Finally, inventories retreated for the second quarter in a row. This is consistent with some purchases being made from previous imports.
To summarize: A strong quarter – significantly stronger than expected – led by household spending. But also with solid business investment in goods and tail winds from the defense sector and exports.
Recent Monthly Data
Monthly data availability has improved significantly, but in many cases is still behind a normal schedule.
Consumer spending measured at a quarterly frequency grew a very strong 3.5% in Q3. At a monthly frequency the situation looks less good with annualized growth decelerating over July through September from 6.3% to 3.6% to 1.3%. Further, consumption growth has exceeded growth in income since May, a period that has seen steady decline in the saving rate. During this same period and beyond consumer sentiment declined significantly (although with some improvement in December). Finally, the labor market situation shows signs of weakening (see below).
Industrial production numbers, which extend through November, are a little confusing, especially when compared with those in our previous report. The reason traces to a set of annual revisions that generally lowered the level of production, but more for data in late 2024 and early 2025 than for more recent months. The result is that the year-over-year number for November looks very healthy (total IP = 2.5%; manufacturing = 1.9%). But most of that growth was in early 2025. Over the past four months total IP is flat and manufacturing has slightly negative growth.
For the residential construction sector, we finally have some new data – for September and October. Still nearly two months behind schedule, but better than nothing. The new housing starts numbers were disappointing. September came in at a 1.306 million annual rate, barely within the 1.3-1.5m range that has held for the past three years. October – at 1.246m – was below the range and the weakest value since May 2020 during the pandemic shutdown. Building permits were just above 1.4 million in both September and October. This is barely back in their recent 1.4-1.6m range after several month below.
And finally, there is new on schedule labor market data for November and December. The job openings number (for November) was 7.15 million down 300k from October. This is the lowest since September 2024 (which was the lowest since 2020) and is indicative of a continuing decline in labor demand.
Two days later came the December employment report. The headline number in the household survey –the unemployment rate – was encouraging. The November value was lowered from 4.6% to 4.5% due to updated seasonal adjustment. And the new number for December came in down a tick at 4.4%. Less positive, while most of this decline was workers who moved from unemployed to employed, part was due to unemployed workers dropping out of the labor force. An implication of this was that the labor force declined for the month.
In the establishment survey December employment rose by 50 thousand, about the same as in November (modestly revised lower from 64k to 56k). The private sector also showed modest growth (November at 50k; December at 37k). In both months all the private increase (and then some) was accounted for by the health care and social assistance category.
All of this implies that the labor market weakening, which we have been worried about for most of the past year, is continuing to slowly develop. But a slowdown in the overall economy – also a central feature of our baseline scenario – is not in evidence, except perhaps in the housing sector.
Baseline Forecast
The result is a more optimistic outlook for the overall economy than in August. (See chart.) Over the first three forecast quarters (25Q4-26Q2) we now expect GDP growth at 2.5%, up from 1.5% in August.
The stronger performance results primarily from stronger business investment spending on equipment and intellectual property. We have gradually become convinced that the AI boom is real and will be for at least this year. Beyond that we have some concerns that it will prove to also be partly a bubble, implying downside risk later in this decade. For now, over the three quarters through 2026Q2, we expect these two components to have a combined growth rate of 5.0%. In our August forecast that number was just 0.6%.
Other components of GDP are also stronger, but much less so. Growth in consumption is 1.5% vs. 1.2%; investment in structures -2.4% vs. -3.1%; government spending 0.8% vs. 0.2%.
A different view is to compare our near-term forecast with recent data. From this perspective the forecast looks less optimistic.
Overall, the GDP growth rate is the same in the two periods – both at 2.5%. But these numbers incorporate a large inventory swing. Abstracting from this (that is, looking at final sales) the data period had a 2.9% growth rate, while the forecast period is just 1.8%. As shown in the chart, this is due to significant reductions in the growth contributions from consumption, investment (combining business and residential spending) and net exports (a smaller decrease in the trade deficit), slightly offset by stronger government spending.
This less optimistic interpretation is supported by our expectation for the labor market. We expect the deceleration in job creation that characterized 2025 to continue through most of this year. For 2026 as a whole, we expect a job loss of about 80k. In our August forecast there was a small increase (+100k)
Discussion
To summarize, we have become more optimistic about the outlook for the economy as a whole, although with an employment picture that is bleak compared to the recent past.
Two factors underlie our GDP optimism:
First, we think that our forecasts for most of the past year we have overstated the (negative) effects of the Trump tariff regime. It is certainly true that the chaotic tariff changes have induced considerable quarter-to-quarter volatility, especially in the first half of 2025. But the 3rd quarter’s strength (which seems to have continued in Q4) is not at all consistent with the widespread “recession is imminent” talk that greeted “Independence Day.” There is a risk in this shift in perspective: It could be that our earlier pessimism was warranted, but that our timing expectation was wrong. If so, 2026 will be weaker than our revised forecast.
Second, as mentioned earlier, we think that the AI phenomenon is having a substantial impact on growth that will endure. One obvious question is: How long? Our current answer is at least into next year. At the most basic level AI is a positive supply-side shock to productivity in the production of many kinds of information-based products. The potential value of these productivity gains underlies the surge in AI investment that is ongoing and that is a basic element of our near-term optimism. It has also produced a surge in stock market wealth that is a support for upper-K consumption.
But beyond the near-term, the storyline gets complicated. As AI is implemented many firms will be able to accomplish current tasks with fewer workers. Ideally, the employees who are freed up will be repurposed to other tasks (perhaps requiring additional training). Or they could be “released to find other uses for their skills.” Stated more bluntly, they could be fired. This latter course would probably increase unemployment (structural unemployment not cyclical), leading to adverse effects on consumption. We think this possibility is a significant risk to our forecast in 2027 or beyond.
Farther out, AI has the potential to enable a surge in economic activities as entrepreneurs devise entirely new businesses based on AI. This could have enormous upside potential, but likely not until late in this decade and into the 2030s.
Returning to the near-term, our forecast of continuing solid growth resting on strong investment and continuing adequate consumer spending puts the Federal Reserve in a complicated position. Inflation is currently well above their 2% target – a condition we expect to continue through this year and to recede only in 2027. This would argue against further easing beyond the 1.75% implemented over the past year and a half, especially if growth remains healthy, as we expect.
But suppose that AI related layoffs become a problem. Or that the housing sector remains slow still further. In this situation the political cries for lower rates would become deafening.
And very likely in error. Nearly all the forces that have driven economic fluctuations in this decade, including those mentioned above, have been disturbances on the supply-side of the economy. And these are ill-suited to monetary remedy.
For example, the pandemic “recession” of 2020 was due to a government mandated shutdowns of vast sections of production in the economy. Fiscal stimulus, while politically perhaps inevitable, met this reduced supply with increased demand. This was the perfect recipe for a surge of inflation, without doing anything to correct the underlying economic problem. Only reopening the economy could do that. Fed easing, both via drastic interest rate reductions and through quantitative easing ratified this outcome.
The current underlying forces driving the economy (tariffs and AI) raise similar policy issues, although at a much-reduced scale. Standard fiscal and monetary policies do not address the basic problems and could make things worse. For the Fed the appropriate policy is probably to do nothing. Set the federal funds rate at a level judged consistent with long-run trends and leave it there. Unfortunately, this is a path the would be almost impossible to follow in the real world. Especially if the underlying shocks start to generate adverse effects – for tariffs, stagflation; for AI, adverse employment developments.
The leading candidates to replace Jerome Powell as Fed Chair (the two Kevins and Chris Waller) might want to think carefully about what they wish for.