Fiscal Health & Divided Government

For the past two years, we’ve had a unified government in which the Republican party has populated the White House and had majorities in both chambers of congress. Once the new legislators are in office we’ll have a divided government. That means that the political party controling the White House won’t also control the senate and house of representatives. In the spirit of saying something of substance, I want to know what the implications are for our fiscal picture.

One can measure fiscal health in many ways. Does a divided government bode well for shrinking the deficit? Rather than spit-ball about politics, I take the data since 1960 in one hand and a model in the other. See the end notes for model details.

Since 1960, the most popular form of government split has been a Democratic president and unified government, followed by Republican presidents and a divided government. Democrats have controlled at least one chamber of congress for 62% of the years. Do these splits matter for federal outlays, receipts, and the deficit? Since a unified democratic government was most common, I use that as the baseline fiscal case and compare the cases to it.

Below are the average impacts of government-party types on federal spending, revenues, and the deficit. The blanks are insignificant differences. On average, the only party split that results in less spending is a Democrat president with divided government. Those and unified Republican governments raise more revenue and reduce the budget deficit from its trend (there was a small decline in in discretionary spending at the start of Trump’s 2nd term with a unified government).  The bad news for deficit hawks is that Republican presidents with a divided government – the type that we’ll have for the next two years – perform poorly. On average, they reduce government revenue and increase the deficit. Again, all of this is ‘on average’.

Continue reading →

Random Sample Vs Sobol Sequence

Say that you have a bachelor’s in finance, but that you’re not exactly a ‘math’ person. You know that diversification is good for reducing the volatility of returns, but the calculations are a bit mysterious to you. You know what the efficient frontier is, but its calculation is elusive. All is not lost! There are alternatives to the math: Sampling methods. By testing a few possible portfolios, we can get an idea of which one is the best fit for you. Here I’ll provide an example using 1) a simple random sample, 2) stratified random sample, 3) Sobol sequence, & 4) Scrambled Sobol.

I’ll use a two-asset portfolio here for ease. I assume that the portfolio includes no short or leveraged positions and that the asset weights sum to one. This is known as ‘long only’. The challenge is that we want a certain return-volatility profile for our portfolio, but it’s a bit hard to figure out what our options are. Luckily, both return and standard deviation (SD) of a portfolio are determined by the weights on the individual assets.

With a long only portfolio, all possible weights are uniformly distributed between one asset having zero weight and the same asset having 100% of the weight. Since the weights sum to one, whatever weight remains after the first asset is attributable to the 2nd asset. See the figure below. Unlike many distributions, we know exactly what’s possible for the weights. The portfolio expected return is simply the linearly weighted average and the portfolio variance is quadratically weighted. Return and SD equations are below.

The portfolio return is mathematically easy, since it’s just a linear average: more weight on asset A implies a rate of return that is closer to that of asset A. But the same is not true for SD, due to the covariance component. We can graph how both portfolio returns and SD are related to the asset weights below.

We already know that the possible weights are distributed uniformly (and so are the returns), but the possible SD are not distributed uniformly. Without the math, we want to know what our possible portfolio SD are so that we can determine what mix of assets is right for us. This example is a bit trite, but the gains occur when generalizing to many-asset portfolios. We can choose from a few methods for describing the correspondence between weights and the SD.

Method 1: Random Sample

The easiest thing to do is take a random draw of a few weights and calculate their associated returns and SDs. However, randomness also implies that we may get a set of points which are clustered together or that don’t sample the various nooks and crannies very well.  An example of a random sample is below, along with a line that connects the sampled options (I assume that we know the extrema weights: 0% & 100% values). Random sampling does an ‘ok’ job of choosing relevant portfolio allocations. But due to the random nature, substantial pieces of the actual SD menu options may be omitted. Sometimes the omitted portions are no big deal and other times they are – and which occurs entirely at random.  In the image below, the random portfolios left out the possible allocations that yielded the lowest volatility.

Method 2: Stratified Random Sample

The fact that the possible weights are distributed uniformly in weight-space is a big advantage. One way to constrain our sample of weights is to use a stratified random sample. In this case, we split the uniform weights into several equal intervals and then take a random sample from within each interval. This ensures that each interval’s characteristics are reflected, even partially, in the overall sample. The various shaded regions below reflect the regions from which a sample was drawn. While the regions are uniformly distributed, the random sample(s) within each region can occur anywhere within the region.

The green line implied by the sample is closer to the minimum SD, and the portions that have less weight on asset A are now less accurate (further from the dotted line). So, we gained some precision in some parts of the SD correspondence, but lost some in others. Loss is especially likely to occur near more jagged or irregular possible values. Since we don’t necessarily want the minimum SD allocation, using a stratified random sample is not always better than a simple random sample. But at least the maximum gaps between the observations are constrained. A simple random sample can detect local irregular shape sometimes, whereas a stratified random sample might always lack adequate resolution in a narrow interval.

Method 3: Sobol Sequence

Continue reading →

What Will the GDP Report Say Tomorrow?

Based on all of the good predictors that I track, GDP growth for the 3rd quarter of 2026 will likely be right around 3.2%:

Table 1

Kalshi, the Atlanta Fed’s GDPNow model, and the WSJ survey of economists are typically the best predictors, and they are all right around 3.2%. I’ve also added a new column to the table, a Nowcast from the St. Louis Fed. This one tends to be a pretty poor predictor, so you might wonder why I’m adding it to my list. Read on below to see why, but for now it’s interesting to note that it is very close to the three other good predictors (the NY Fed model is generally not very good).

When I have tracked these predictors in the past, I have been comparing them to the advanced GDP estimate, the first one the BEA puts out. Being able to predict this release is a good and hard enough task, but ultimately we don’t actually care about the number that will come out tomorrow. This number will be revised many times with more data, not only the 2 months after it, but through subsequent benchmarking and revisions in later years. Ultimately, we care whether these predictors are getting the true number, not BEA’s initial release (even though they are calibrated to predict that initial release, for the most part as I understand them).

What does it look like if we compare those predictions to the current best data on GDP growth going back to 2023? On this measure, the St. Louis Fed Nowcast looks pretty good recently:

Table 2

While the St. Louis Fed Nowcast looks pretty bad over the entire time period — the worst of these 5 predictors — it has done very well in the past 3 quarters, including nailing the 1st quarter of 2026 exactly (though in one quarter, they got the sign wrong). So I’m now tracking it in my table to see if it will continue to perform well — maybe they improved the model? The 3rd quarter data won’t actually tell us much about the St. Louis Fed model relative to the others, since they all are saying the same thing, but it will be worth watching in the future.

If we tally up the best predictors using the final or most recent GDP growth estimate, the best predictor is Kalshi’s betting markets. Kalshi has the lowest average error, which is just slightly better than the WSJ survey of economists. But Kalshi was also the best predictor in 6 of the 14 quarters in Table 2. The WSJ survey was only the best in 3 quarters — and in two of those it had the same prediction as Kalshi. It was only best as a standalone in one quarter (the oldest one in these 14 quarters). The St. Louis Fed was the best predictor in 5 quarters, all of these by itself, but its average error (as well as standard deviation and variance) is the highest of these 5 predictors.

Based on different combinations of the predictors, the best pair of predictors is Kalshi and WSJ. Averaging those together gets you an average error of 0.68, better than any single predictor. That means my prediction based on all this information is 3.15% growth in the third quarter for the final figure, once we eventually have it. The average of the Atlanta Fed and Kalshi is the best pair for the advanced figure, which gives a slightly different prediction of 3.24%.

Disability: Income, Unemployment, & Education

Let’s talk about disability. The American Community Survey (ACS) identifies seven broad categories, not including the baseline case of ‘No Disability’. I’ve written before about how disability designations are difficult because many qualifying conditions are a matter of degree. Surely you’ve met a person who is legally blind but can see *some*. How much sight ability can remain while still being considered blind? If a person can identify the sun’s location, but nothing else, then is that person still blind? Similarly, who’s to say whether some physical or cognitive characteristic is disabling? Why not enabling – depending on the task?

Wisely, the ACS side-steps this entire discussion and simply asks respondents whether they have “difficulty”.  We can be agnostic about the cause and degree while still trying to get at the heart of the matter. One downside is that people tend to have more difficulty as they age. Since that’s not relevant for earlier life economic outcomes, I only examine ages 24-45. Difficulty with “self care” is too small of a category in the 2024 ACS for the analysis here. Below are the other six difficulties identified:

  • Hearing Difficulty (n=6,609)
  • Seeing Difficulty (n=8,307)
  • Cognitive Difficulty (n=26,792)
  • Ambulatory Difficulty (physically getting around) (n=4,568)
  • Independent Living Difficulty (n=5,997)
  • No Disability (n=885,302)

We should say early on that many of the people who participate in “Deaf” (capitalized) culture do not consider deafness to be a disability and that any such suggestion is motivated by bias and hearing culture. I won’t engage with that here. But I can say that there is a long tradition going back through the 1800s that rejected alms-giving for deafness and instead supported independent living for deaf people. Historically, this manifested as outlawing panhandling while featuring deafness and rejecting government programs that provide simple transfers due merely to deafness. More recently, the push to reject the ‘disability’ label and its baggage has even been the motivation for some deaf people to oppose cochlear implants being covered by medical insurance. For ease, I’ll be using the colloquial terms for all of the above difficulties. Now that you have the greater context, let’s jump into the data (unweighted).

The baseline income chart is above. Given the above description, it should come as no surprise that deaf people have the highest average income among disability groups, conditional on employment. However, it’s also surprising that most of the above disabilities have incomes that I wouldn’t exactly call destitute. Many people who have a disability look much like everyone else. From an economic point of view, we all have a comparative advantage and it’s perfectly consistent with gainful employment and flourishing otherwise.

Continue reading →

Has Mexico’s “Canasta Básica” Policy Made Food More Affordable?

A recent article in The Guardian profiles economist Isabella Weber. She doesn’t seem to like other economists very much, supposedly because we “obsess over ideological purity and are inflexible when searching for solutions.” Weber wants economists to come up with solutions that work even if… I guess, even if they don’t work well in theory? Maybe they will work well in practice?

Anyway, so far as I can tell, the article only gets into details on two concrete solutions to affordability. I’m sure she has more in her book, but let’s see what these are. The first for governments to create “publicly owned entities that can buy and sell grain into its market,” presumably to smooth out fluctuations in grain prices. The second is “the Mexican government’s canasta básica policy, through which retailers voluntarily cap the total price of a basket of two-dozen essential goods.”

Let’s look at the second one, since I think for developed countries like the US, fluctuating grain prices aren’t the primary source of affordability concerns. But perhaps American consumers would like a basket of grocery essentials to be cheaper.

Has this policy in Mexico worked? It was first implemented in 2022, and food inflation has certainly fallen in Mexico since then. But it has also fallen everywhere else in the world — 2022 was peak inflation not just in the US, but in most countries — so we can’t judge the success of the program based on that alone.

Instead, we can look at internationally comparable data on food spending, which comes from the USDA’s Economic Research Service. Our World in Data has a nice chart based on this data, and even though it is now two years out of date (I will provide 2025 data below), the 2023 data makes a case against the success of the “price controls” in Mexico (I put “price controls” in quotes because it is not a true price control, but an attempt to control prices via incentives):

Figure 1

It will probably not surprise you that the US spends much less on food as a percent of total spending than Mexico, given that we are a high-income country. But it is still worth pointing out just how low food spending is in the US. However, this chart still paints Mexico in a very poor light: for countries with about the same income level in the Americas (as measured by consumer expenditures), Mexicans spend much more on food than Argentina, Chile, or Panama.

If we use the 2025 USDA data, and focus on the 10 countries with consumption expenditure closest to Mexico (between about $8,000 and $12,000 of consumption), we see that Mexico is very much on the high end: only Argentina was clearly higher:

Table 1

Finally, we can compare Mexico against itself in this data, looking at food spending going back to 2017:

Figure 2

Food spending actually peaked in 2020, though be cautious of any data from 2020 as the pandemic made data collection difficult. After declining in 2021 and then into 2022 (the policy seems to have been put in place in mid-2022), there haven’t been any further declines. And food spending is still well above pre-pandemic levels (unlike the US, where the 2025 figure of 6.4% is the same as 2017-18).

Ultimately, we would want a more sophisticated analysis than my simple tables and charts here. But it doesn’t appear that Mexico’s 2022 policy really made food much more affordable, either relative to peer countries with similar income levels, or relative to Mexico’s own recent history. And that’s even before we get into any unintended consequences of the “price controls.” Of course, the policy may have made that basic basket of two-dozen items more affordable than it otherwise might be — for that we would need to compare individual item prices to peer countries — but it hasn’t obviously made food overall more affordable in Mexico.

Marriages By College Major

[Edited to include divorces]

The American Community Survey (ACS) collects information on how many times people have been married.  Using the 2024 ACS, we can see how that changes across ages. Getting married is what we call an absorbing state. Once you get married once, you’ll never be married zero times again. So long as people get married across the age distribution, the proportion of never married people falls.

The same is true for getting married a 2nd time. Then, you’ll never have been married once ever again. While the proportion of people never married must shrink with age, the proportion of people married once can rise. If a few people get divorced or married a 2nd time, but an even bigger number of people get marred for the first time, then the group of people married only once can grow. The ACS stops measuring once someone has been married three times – then it’s “three or more” times. A subsequent divorce the universal absorbing state. Those people never go into another category.

The chart shows the proportion of people by age and the number of times that they’ve been married. Darker hues reflect divorces.* Importantly, this is the snapshot in 2024. So each age group is a cohort snapshot in time. That’s why the proportion of people in their 80s who’ve been married zero or one time is greater than people in their 70s. They’re a different cohort. When the now 80-year-olds were in their 70s, the sum of the blue and light yellow bars were larger than for more recent 70-year-olds.

Continue reading →

Health Insurance Benefit Mandates and Premiums

The Center for Modern Health just released a report “Paring Down Coverage Mandates To Reduce Health Insurance Premiums“. It builds on one of my papers to estimate how much states could drive down insurance premiums by reducing the number of things health insurers are required to cover:

Estimated savings for Family Plans (Table 2 of CMH Report)

It’s nice to see that someone in the health policy world remembers that health insurance benefit mandates exist and can be a significant cost driver. I’m also glad to see that someone still cares about my 2014 paper (the first one I ever got accepted at a journal), though I wish there were more up to date estimates available for them to cite.

Why has no one developed more recent estimates? Partly for data reasons- the organizations that had been putting out counts of mandates by state each year (the Council for Affordable Health Insurance & Blue Cross Blue Shield Association) stopped doing so. Partly because of the ACA, which changed how mandates worked in complex ways.

That said, there there seems to be some more recent data from 2017 that could be used for a slight update, and this post and my ideas page detail the many things we still don’t know about how mandates operated prior to 2014:

State Health Insurance Mandates: Most of my early work was on these laws, but many questions remain unanswered. States have passed over a hundred different types of mandated benefits, but the vast majority have zero papers focused on them. Many likely effects of the laws have also never been studied for any mandate or combination of mandates. Do they actually reduce uncompensated hospital care, as Summers (1989) predicts? Do mandates cause higher deductibles and copays, less coverage of non-mandated care, or narrower networks? How do mandates affect the income and employment of relevant providers? Can mandates be used as an instrument to determine the effectiveness of a treatment? How does all this change in light of the ACA? On the identification side, redoing older papers using a dataset like MEPS-IC where self-insured firms can be used as a control would be a major advance.

Wealth of Generations in 2025

Two weeks ago I provided an update on estimates of wealth for recent generations in the US. But just this week, the Federal Reserve gave us the once-every-three-years complete update to this data, which allows us to calculate medians rather than average, and to more precisely separate age groups.

I will have a lot more to say about this data, both on this blog and a book I am working on, but for now here is the chart you have all been waiting for:

Degrees, Higher Degrees, & Higher Income

I’ve discussed multiple times now how economics degrees are associated with higher income. One reason may be that it’s not the bachelor’s degree that’s doing the legwork, but rather the graduate degree. I examine that hypothesis here for the 25 most popular degrees in the 2024 American Community Survey.

The data identifies who has a bachelor’s degree and in what field. We also know whether an individual has a master’s or professional degree, or a doctorate. Unfortunately, we don’t know the field for the higher degrees. So, while we can say that 45% of Economics majors have a graduate degree, we can’t say that the graduate degree is in Economics. After all, plenty of people major in one thing and pursue a graduate degree in another thing.

Degree Types by College Major

Below is a chart that ranks the 25 most popular college majors and whether those graduates also hold a graduate degree. There are several patterns. First, the tier near the top of the chart shows more business majors tend to be terminal. Students majoring in communications, business, and marketing tend not to proceed to get higher degrees.

That contrasts starkly with the bottom tier of the graph where majors like Chemistry, Biology, Political Science, Math, and History all tend to have high rates of graduate degrees.  I’m not sure why this is. Maybe it’s because the skills are more research-oriented in nature and prepare those students well for graduate school. Other alternatives may be that the bachelor’s degree alone doesn’t provide much income or that a graduate degree results in a large proportional increase in income for those majors.

Economics majors are on the lower end in terms of the proportion with only a bachelor’s degree (55.5%), near the middle of the pack for proportion with a master’s or professional degree (39.7%), and have an average proportion holding doctorates (5.1%).

Because the numbers make it easy, I’ll pick on education majors a little. Given the high proportion of education and elementary education majors who have graduate degrees, there are relatively few of them who have doctorates. Other majors that have a lot of graduate degree holders tend to have more doctorates, but education majors in higher education are overwhelmingly people with master’s degrees. Given that many of them work for governments, I suspect that they often get procedural pay raises merely for holding a higher degree.

Income by Degree Type and College Major

Below is a chart that shows the average income among employed people by college major and higher degree. The highest-paying terminal bachelor’s and master’s degrees are both among Economics majors ($144k & $203k). A previous post showed that Economics majors had the highest overall income and here we can see that it comes primarily from the bachelor’s and master’s degree holders, while economics majors with doctorates earn the 3rd highest income ($208k) behind computer science ($247k) and electrical engineering ($238k).

Continue reading →

Economic Freedom Is Enough

Adam Smith kicked off modern economics by asking what causes nations to become wealthy. We’ve spent centuries improving our answers to the question and we still have lots to learn, but I think Smith’s original answer holds up remarkably well:

Little else is requisite to carry a state to the highest degree of opulence from the lowest barbarism, but peace, easy taxes, and a tolerable administration of justice; all the rest being brought about by the natural course of things. All governments which thwart this natural course, which force things into another channel, or which endeavour to arrest the progress of society at a particular point, are unnatural, and to support themselves are obliged to be oppressive and tyrannical

I was reminded of this by the latest Economic Freedom of the World report just released:

The overall correlation of GDP per capital and economic freedom is strong, but I find the extremes even more striking. Economic freedom is necessary for wealth: every country with a GDP per capita above $50,000 has above-average economic freedom (getting even close without high economic freedom requires having lots of oil). Economic freedom is sufficient: every country with an economic freedom rating over 8 has a GDP per capita over $50,000.

I periodically see someone claim that a country “did everything right” in terms of policy but still wound up poor, usually implying that the country is doomed by their geography or genes, or arguing that “neoliberal policy” failed. Whenever I check in on these countries, I inevitably find them to be far from “doing everything right”, with economic freedom scores typically far below 8.

It’s true that these are “just correlations” and that economic freedom alone isn’t a perfect predictor. It’s worth looking into the outliers and wondering- if a country is above the trend line without oil wealth (like China), what else are they doing right? If a country is well below the overall trend line (like Guatemala), does it mean they are have something else working against them to counteract the benefits of economic freedom? Or does it mean they are about to turn the corner and see high economic growth?

Overall though I don’t think it’s a big oversimplification to say that Adam Smith was right, and economic freedom is enough to bring prosperity. If you’re looking for economic growth, start with Smith’s peace, easy taxes, and tolerable administration of justice, or the Economic Freedom of the World’s small government, strong property rights and legal system, sound money, and freedom to trade at home and abroad.