I’ve now posted individual-level responses to the 1978-2025 Michigan Consumer Surveys to Kaggle in CSV and Stata formats. The University of Michigan’s Consumer Surveys are a widely followed source for data on consumer confidence and inflation expectations:
Their official site is good if you just want summary tables or charts like this:
But what if you want detailed crosstabs to see how sentiment differs for different groups, or microdata so that you can run regressions? With enough clicks you can get this from what UMich calls their “cross-section archive“. But it is pretty hidden, my student looking into this thought they just didn’t offer individual-level data; and even once you get their data, it is in an unlabelled CSV file with hard-to-understand variable names and codes. So I wanted to make it clear that the full data with all responses for all years is available, and if you use my Stata version it is even reasonably easy to understand (the code I adapted for labelling it is on OSF). Then you can run your regressions, or make charts like this:
The College-Only Covid Recovery
If you’re new here, a reminder that you can find other cleaned-up versions of popular datasets on my data page.
Despite DOGE and the President partially stopping some payments for some federal agencies, the changes so far aren’t visible at all in federal payments data. The Brookings Institution has put together a new tool that tracks daily spending data from the US Treasury. (My co-blogger Zach wrote about this tool last week too.) Here’s a chart from that tool showing total federal outlays by calendar year. Notice that 2025 is right on track with the past two years, or just slightly above (dollars are in nominal terms):
Of course, given the massive amount of US federal spending and the large number of agencies, we might expect it to take more than a few months to get spending under control or significantly alter its course. But this way of tracking the data is definitely picking up any changes made so far. For example, notice the flat lining of USAID funding after Trump comes into office at the end of January:
So while we don’t see any big changes yet in the aggregate spending, the few small agencies that DOGE has frozen are showing up in this data. That tells this will be a useful tool to follow going forward.
The chart comes from the Atlanta Fed’s GDPNow model, which tries to estimate GDP growth each quarter as data becomes available. The sharp drops in their Q1 forecast for 2025, based on the last two data updates, look pretty shocking. Should we be worried?
First, it’s useful to ask: has this model been accurate recently? Yes, it has. For Q4 of 2025, the model forecast 2.27% growth — it was 2.25%. For Q3 of 2024, the model forecast 2.79% growth — it was 2.82%. Those are very accurate estimates. Of course, it’s not always right. It overestimated growth by 1 percentage point in Q1 of 2024, and it underestimated growth by 1 percentage point the quarter before that. So pretty good, but not perfect. Notable: during the massive decline in Q2 2020 at the start of the pandemic, it got pretty close even given the strange, uncertain data and times, predicting -32.08% when it was -32.90% (that’s off by almost 1 percentage point again, but given the highly unusual times, I would say “pretty good”).
OK, so what can we say about the current forecast of -2.8% for Q1 of 2025? First, almost all of the data in the model right now are for January 2025 only. We still have 2 full months in the quarter to go (in terms of data collection). Second, the biggest contributor to the negative reading is a massive increase in imports in January 2025.
To understand that part of the equation, you have to think about what GDP is measuring. It is trying to measure the total amount of production (or income) in the United States. One method of calculation is to add up total consumption in the US, including by final consumers, business investments, and government purchases and investments. But this method of calculation undercounts some US production (because exports don’t show up — they are consumed elsewhere) and overcounts some US production (because imports are consumed here, but not produced here). So to make GDP an accurate measure of domestic production, you need to add in exports, and subtract imports.
Keep in mind what we’re doing in this calculation: we aren’t saying “exports good, imports bad.” We are trying to accurately measure production, but in a roundabout way: by adding up consumption. So we need to take out the goods imported — not because they are bad, but because they aren’t produced in the US.
The Atlanta Fed GDPNow model is doing exactly that, subtracting imports. However, it’s likely they are doing it incorrectly. Those imports have to show up elsewhere in the GDP equation. They will either be current consumption, or added to business inventories (to be consumed in the future). My guess, without knowing the details of their model, is that it’s not picking up the change in either inventories or consumption that must result from the increased imports. It’s also just one month of data on imports.
As always, we’ll have to wait for more data and then, of course, the actual data from BEA (which won’t come until April 30th). More worrying in the current data, to me, is not the massive surge in imports — instead, it’s that real personal consumption expenditures and real private fixed investment are currently projected to be flat in Q1. If consumption growth is 0% in Q1, it will be a bad GDP report, regardless of everything else in the data.
If you follow libertarian media such as Reason Magazine or its ancillaries, then you are well acquainted with the humdrum of “it goes without saying that most US programs should be ended“. They kind of just say this and then continue with their news. One of the favorites is to say that we should get rid of the Department of Education (ED). After all, 90% of K-12 education is paid for by states and localities. Here I was thinking “what does the Department of Education even do”?
Agreement is different from trust. I trust the Brookings Institute. They have a nice explainer on what ED does. It’s a quick overview and has plenty of the appropriate citations. I learned that most of what ED does concerns K-12 and is achieved through grants that have strings attached. Funding primarily goes to serving “educationally disadvantaged” communities (that have a high poverty rate). Funding also goes to programs for disabled children, minority education programs (like Howard University), and Indian tribes. They also administer Pell Grants and fund & regulate college loans (which are privately administered).
ED’s appropriated budget is online for anyone to see and includes pretty good detail about costs. The total discretionary cost of FY 2024 was $79 billion. The “mandatory” spending, which does not need to be voted on by congress every year, was $45 billion. For context, the entire federal FY 2024 expenditure was $6.75 trillion. So, eliminating the department of education *and* it’s responsibilities (an unpopular position) would reduce federal expenditures by 1.8%. For even more context, the budget deficit is $1.83 trillion or 27.1% of total federal expenditures. Eliminating ED and consolidating its responsibilities to other departments would save $0.6 billion. That assumes eliminating program administration, the ED office of civil rights, and the ED office of the inspector general.
Whenever researchers are conducting studies using state- or county-level data, we usually want some standard demographic variables to serve as controls; things like the total population, average age, and gender and race breakdowns. If the dataset for our main variables of interest doesn’t already have this, we go looking for a new dataset of demographic controls to merge in; but it has always been surprisingly hard to find a clean, easy-to-use dataset for this. For states, I’ve found the University of Kentucky’s National Welfare Database to be the best bet. But what about counties?
I had no good answer, and the best suggestion I got from others was the CDC SEER data. As so often, the government collected this impressively comprehensive dataset, but only releases it in an unusable format- in this case only as txt files that look like this:
I cleaned and reformatted the CDC SEER data into a neat panel of county demographics that look like this:
I posted my code and data files (CSV, XLSX, and DTA) on OSF and my data page as usual. I also posted the data files on Kaggle, which seems to be more user-friendly and turns up better on searches; I welcome suggestions for any other data repositories or file formats you would like to see me post.
This chart shows the average number of hours worked in the US, by gender, for those in their prime working ages (25-54), from 1900 to 2023. It includes both paid market work and household production (which includes activities like cooking, cleaning, shopping, and taking care of children):
A few things to notice. First, there is no data for 2020, as the ATUS didn’t publish any tables due to incomplete data from the pandemic. And even if we had data, it would have been a huge outlier year.
More importantly, there is an obvious long-term trend of declining market work and rising household production for men, and the opposite for women. In 1900 women worked over 6 times as many hours in the household as they did in the market, but by 2023 they worked almost the exact same number of hours in each sector.
Male hours in market work declined by about 16 hours per week (using 10-year averages, as there is a slight business-cycle effect on hours), but the total number of hours they worked declined much more modestly, by about 3 hours per week (note: these numbers include all men, whether they are working or not). Women saw similar changes, but in the opposite direction, with total hours worked only falling by about 4 hours per week, even though hours working at home fell by almost 22 hours.
Americans do have more leisure time than in 1900, but not dramatically so: perhaps 3-4 hours per week. This is an improvement, but less of an improvement than you might suspect by looking at hours of market work alone.
Ramey and Francis do try to carefully distinguish between household production and leisure. For example, yardwork and changing diapers are household production, while gardening and playing with your children are leisure. For some respondents to surveys, they may feel differently about whether gardening is “really” work or not, and some may enjoy changing baby’s diapers, but in general their distinctions seem reasonable to me.
Finally, we can say pretty confidently with this data that women have almost always worked more hours than men — the one exception in the 20th century being WW2 — and the gender gap was about 4 hours per week in both the early 1900s and the most recent decade (though it did fluctuate in between).
I was on a panel of economists last night at an event titled “The Economic Consequences of President Trump”. We each gave a 5-minute summary from our area of expertise and then opened up the floor for questions. This is a truncated summary of my talk. Since the panel included an investor, two industry economists, and another macro economist, I wanted to discuss something that was distinct from their topics. I’ve published a paper and refereed many articles concerning economic policy uncertainty (EPU) and asset volatility. I wanted to look at the data concerning President Trump – especially in contrast to Presidents Obama and Biden.
EPU matters because uncertainty can cause firms and individuals to delay investment and hiring decisions. Greater uncertainty can also cause divergent views concerning forecasted firm profitability. The result is that asset prices tend to become more volatile when EPU rises. One difficulty is that uncertainty occurs in our heads and concerns our beliefs, making it hard to measure. We try to get at it by measuring how often news media articles include the terms related to uncertainty, policy, and the economy. Since news content tends to report what is interesting, relevant, or salient to customers, there’s good reason to think that the EPU index is a decent proxy.
Using the Obama years as a baseline, the figure below simply charts out EPU. It was relatively low during Trump’s first term and then it was higher during Biden’s term – even after accounting for the Covid spike. The sharp increase toward the end is after Trump won the 2024 election. The EPU series conflicts with my perception of social media and media generally. My experience was that the media was far more attentive to the uncertainty that Trump caused. But, it may just be that the media outlets had plenty to report on rather than it being particularly indicative of EPU. After all, if the president exercises his power, then there is a certain swift decisiveness to it.
But if we look at a couple of particular policy areas, Trump’s administration faired worse. Specifically, Trump caused a ruckus concerning trade policy and immigration. Remember when Biden continued the aggressive trade policy that Trump had adopted? That’s consistent with lower EPU. Similarly, Biden made the immigration process much easier and faster while Trump’s deportation haranguing results in a somewhat stochastic means by which people are deported. Again, that spike at the end is after Trump won the 2024 election.
The average hospital is now 3/4 full- more full than during much of the worst of the Covid pandemic, and well above the 2/3 occupancy rate that prevailed during the 2010s. This is according to a study out yesterday in JAMA Open:
This seems to be due to a reduction in bed supply, rather than an increase in demand:
The number of staffed hospital beds declined from a prepandemic steady state of 802 000 (2009-2019 mean) to a post-PHE steady state of 674 000, whereas the mean daily census steady state remained at approximately 510 000
To me this is one more reason to reform Certificate of Need laws that put barriers in the way of hospitals opening or adding beds. Luckily I see a lot of momentum for CON reform this legislative season, including the highest-occupancy state, Rhode Island:
After raising rates in 2022 to belatedly combat inflation, the FOMC was feeling successful in 2024. They were holding the line and remaining steadfast while many people were getting all in a tizzy about pushing us into a recession. People had been predicting a recession since 2022, and the Fed kept the federal funds rate steady at 5.33% for an entire year. Repeatedly, in the first half of 2024, betting markets were upset that the Fed wasn’t budging. I had friends saying that the time to cut was in 2023 once they saw that Silicon Valley Bank failed. I remained sanguine that rates should not be cut.
I thought that rates should have been higher still given that the labor market was strong. But, I also didn’t think that was going to happen. My forecasts were that the Fed would continue to keep rates unchanged. At 5.33%, inflation would slowly fall and there was plenty of wiggle room for unemployment.
Then, we had a few months of lower inflation. It even went slightly negative in June 2024. Some people were starting to talk about overshooting and the impending recession. I documented my position in August of 2024. Two weeks later, Jerome Powell gave a victory lap of a speech. He said that “The time has come for policy to adjust”. Instead of discerning whether the FOMC would cut rates, the betting markets switched to specifying whether the cut would be 0.25% or 0.5%. The Fed chose the latter, followed by two more cuts by the end of the year.
I was wrong about the Fed’s policy response function. But why? Was the FOMC worried about the downward employment revisions? That was big news. Did they think that they had inflation whipped? I’m not sure. There was a lot of buzz about having stuck the soft landing. In late 2024, I leaned toward the theory that the Fed was concerned about employment. Like, they thought that we had been doing better until then.
The NSCH is the latest casualty of the new administration taking down major datasets from government websites. Between Archive.org and what I had downloaded for old projects, I was able to get all the 2016-2023 topical NSCH files and post them on an Open Science Foundation page.
I took this as a chance to improve the data- the government previously only made the topical Public Use Files available in SAS and Stata formats one year at a time, so I added a merged version for all available years in both Stata and Excel formats.
I hope and expect that the National Survey Children’s Health will be back up at official websites soon. But I expect that other datasets will be taken down permanently, so now is the time to download what you think you might need and add it to your data hoard– especially if you want anything from the Department of Education.
Economist Writing Every Day 