College Major, Marriage, and Children

The American Community Survey began in 2000, and started asking about college majors in 2009, surveying over 3 million Americans per year. This has allowed all sorts of excellent research on how majors affect things like career prospects and income, like this chart from my PhD advisor Doug Webber:

See here for the interactive version of this image

But the ACS asks about all sorts of other outcomes, many of which have yet to be connected to college major. As far as I can tell this was true of marriage and children, though I haven’t searched exhaustively. I say “was true” because a student in my Economics Senior Capstone class at Providence College, Hannah Farrell, has now looked into it.

The overall answer is that those who finished college are much more likely to be married, and somewhat more likely to have children, than those with no college degree. But what if we regress the 39 broad major categories from the ACS (along with controls for age, sex, family income, and unemployment status) on marriage and children? Here’s what Hannah found:

Every major except “military technologies” is significantly more likely than non-college-grads to be married. The smallest effects are from pre-law, ethnic studies, and library science, which are about 7pp more likely to be married than non-grads. The largest effects are from agriculture, theology, and nuclear technology majors, each about 18pp more likely to be married.

For children the story is more mixed; library science majors have 0.18 fewer children on average than non-college-graduates, while many majors have no significant effect (communications, education, math, fine arts). Most majors have more significantly more children than non-college graduates, with the biggest effect coming from Theology and Construction (0.3 more children than non-grads).

In this categorization the ACS lumps lots of majors together, so that economics is classified as “Social Sciences”. When using the more detailed variable that separates it out, Hannah finds that economics majors are 9pp more likely than non-grads to be married, but don’t have significantly more children.

I love teaching the Capstone because I get to learn from the original empirical research the students do. In a typical class one or two students write a paper good enough that it could be published in an academic journal with a bit of polishing, and this was one of them. But its also amazing how many insights remain undiscovered even in heavily-used public datasets like the ACS. We’ve also just started to get good data on specific colleges, see this post on which schools’ graduates are the most and least likely to be married.

John List, Dramatist

As someone who has dabbled in lab experiments for over a decade, I’m familiar with complaints about external validity. If an experiment is run with only college students, then how can we know if the finding will generalize to other populations? It’s a question worth asking, but many questions are worth asking and it doesn’t mean that controlled experimentation can’t add value to the economics literature. In the age of general suspicion of small studies, people say that replications are needed. We should only trust a conclusion that is supported by multiple studies. The thing about replications is that the process has to start somewhere. Empirical work has to get read and published. Replications are composed of individual studies.

I just met John List at the Alabama stop on his epic national book tour. He directed me to his work of art: Ungated Link. He wrote a play in response to the attacks on his work concerning external validity. He employs a rhetorical strategy of making your critics look obtuse. Even though the play is absolutely silly (thoroughly entertaining), he builds a strong defense for doing experiments. It is literally presented as the arguments of a defense lawyer. Before the trial begins, a “reporter” summarizes the conflict that has created the need for a formal trial:

Court Reporter Clifton Hillegass: Thank you Judge Learner. While it is never easy to convey succinctly the key points of a debate, this dispute has crystallized in a manner that leaves no middle ground. The prosecution, led by Mr. Naiv Ete, argues that all empirical work in economics must pass a set of necessary external validity conditions before being published in academic journals or used by policymakers. To date, in this courtroom no empirical work has passed his conditions, effectively rendering the question of generalizability beyond dispute, or as Livius Andronicus reminded us, Non est Disputandum de Generalizability. Ms. Minerva, Lead Defense, has argued that this line of reasoning leaves only theoretical exercises and thought experiments to advance science and guide policymaking, an approach that she fears will return us to the dark ages.

The paper is called “NON EST DISPUTANDUM DE GENERALIZABILITY?” It’s a good refresher on the history of science, not just economics.

Maybe the first best is for you to spend your weekend reading dense technical papers. But if you aren’t feeling up to that, then this play will make you feel like you learned something without even trying.

I’ll link this up to some of the posts I wrote last year about experiments and critics:

Calling Behavioral Economics a Fad

Behavioral Economist at Work

Health Insurance Benefit Mandates and Health Care Affordability

My article on benefit mandates was published today at the Journal of Risk and Financial Management. It begins:

Every US state requires private health insurers to cover certain conditions, treatments, and providers. These benefit mandates were rare as recently as the 1960s, but the average state now has more than forty. These mandates are intended to promote the affordability of necessary health care. This study aims to determine the extent to which benefit mandates succeed at this goal

I began my research career by writing about these mandates, and my goal with this article was to tie up that whole chapter. I realized that all my articles on benefit mandates, as well as most of what other economists write about them, simply try to measure their costs- how much they raise health insurance premiums, raise employee contributions to premiums, lower wages, lower employment, or harm smaller businesses. Its good to know their costs, but to really evaluate a policy we should learn about its benefits too so that we can compare costs and benefits.

One key benefit that had yet to be measured was how much a typical mandate lowers out-of-pocket health care costs. In this article, I estimate that the average benefit mandate lowers costs by 0.8%-1%. I argue that combining this with a measure of how mandates affect total health spending by households could provide a sufficient statistic for the net benefits of mandates for households. I’m not totally confident this works in theory though, and it has a big challenge in practice- one of my empirical strategies finds that mandates reduce total spending, but the other finds they don’t. So I think the main contribution of the article ends up being the first estimate of how the average state health insurance benefit mandate affects out-of-pocket costs.

I’m currently planning to move on from writing about mandates- other topics are catching my eye, state policymakers don’t seem to particularly care what the research says about mandates, and changes in how economists use difference-in-difference methods are making it harder to publish articles like this that study continuous treatments. But I think there are still big opportunities here for anyone who wants to take up the torch. First, the ACA Essential Health Benefits provision changed the game for state mandates in a way that I have yet to see the empirical literature grapple with. Second, there are more than a hundred separate types of state benefit mandates; in most of my articles I aggregate them but they should really be studied separately. A handful have been, such as mandates for autism treatments, infertility treatments, and telemedicine. But the vast majority appear to be completely unstudied.

P.S. Writing this article gave me two wildly varying opinions of our federal bureaucracy. I tried to get both data and funding from the Agency for Healthcare Research and Quality for this article. The data side worked well- they were surprisingly fast, efficient and reasonable about the process of accessing restricted data. On the other hand, I applied for funding from AHRQ in March 2019 and still have yet to officially hear back about it (it is “pending council review” in NIH Commons). This sort of thing is why nimble organizations like Fast Grants can do so much good despite having much smaller budgets.

P.P.S. This article is part of a special issue on Health Economics and Insurance that is still accepting submissions. I’m the guest editor and would handle your submission, though my own got handled by other editors and put though multiple rounds of revisions.

Mises’s Bureaucracy, a Recap

My favorite two economists are Ludwig Von Mises and Milton Friedman. They might consider one another from very different schools of thought, though there is reason to think that they are not so different. As an undergraduate student, I liked them both, but I became more empirics-minded in graduate school and as a young assistant professor.

As I progressed through graduate school and conducted empirical research, my opinions and policy prescriptions changed and were refined from what they once were. In graduate school, I didn’t study Austrian Economics, though it was certainly in the water at George Mason University. Recently, as an assistant professor with a few years under my belt, I picked up Bureaucracy (1944) and read it as a matter of leisure.

One word:

Continue reading

Day care and new pre-K findings

There was a buzz over a new study showing that pre-K is not necessarily good for children. It’s amazing how experts can be completely surprised by the results of a major study on an issue like pre-K education.* Noah Smith summarized the literature and thought through some policy implications. Emily Oster also just summarized the paper and points out that it provides almost no help for parents making decisions. **

I’ll offer some “amateur astronomer” observations about preschool and childcare.

What to call the daycare I patronize, since it offers all of the pre-K functions? I’ll call it Day-K. My kid comes home from Day-K with worksheets difficult enough for a kindergartener, but it was handed to a 3-year-old and the kid just scrawled a few lines of crayon across it. Most little kids aren’t going to retain material that is beyond their developmental level. Why bother printing these nice worksheets at all instead of just letting them color a bear?

Something that surprised me was how early kids can learn the alphabet and yet how disconnected that is from anything useful such as being able to read words. If a 2-year-old can do it (e.g. recognize “A”) then a 4-year-old can probably pick it up easily anyway.

Good private daycares in desirable urban areas are expensive but have unbelievable waitlists. Donald Shoup advocates that cities should charge more for parking. He reasoned that every city block should have an open parking space. Instead of spending valuable time circling like a vulture, you should just pay a lot of convenient parking or else know you will have to go somewhere else. Would the same logic apply to the good daycares? Should they not charge so much that there is always an open slot for the next parent who can pay? One issue with this from the daycare owner’s perspective is that they don’t want new kids cycling through constantly. A brand-new kid who does not trust the staff and has not learned the routine is a temporary disaster. I believe that the waitlists work because the owners want a predictable flow of great committed customers. By keeping fees low enough to have a long waitlist, they get good families to stay and they can easily fill any holes left by departures or dismissals.

If the program was free, I suspect that would change the dynamic inside compared to high-fee Day-K. Daycare kids are on a regimented schedule. Everyone thrives on the routine. The staff are happy when the kids know the rules. If people were coming and going unpredictably, that might make it harder for kids to learn.

Even under optimal conditions, there are scuffles at daycare. Being pushed down on the playground is often the only thing a kid will remember from a full day of “instruction”. How could pre-K actually negatively affect some kids, as the new study shows? One way I can think of is that the experience a good teacher tries to provide could be ruined by one kid who is loud or violent. If half of the classes are functioning as day care and having no impact at all on future outcomes and half of the classes have a kid hitting, then the average effect for all pre-K classes could be negative. The social environment of pre-K is probably highly variable. Sometimes you could get a great social atmosphere in which kids learn to share and sing. Sometimes the chaos level could make things difficult, I imagine. This is speculative. But I think it’s ok to speculate in the brainstorming period that should follow a surprising result.

Daycare centers have a fantastic physical environment. When I think of the returns to scale, the low table and chairs that fits the 3-year-olds perfectly comes to mind. A preschool classroom has a perfect bathroom with low toilets and sturdy step stools at the sinks. There is no heirloom China or nice upholstery in the room to worry about. There are dozens of age-appropriate toys and craft supplies can be bought in bulk. This physical environment allows kids to be creative and have fun. Adults don’t have to hover over them, afraid that they’ll hurt themselves or break something at any moment. By contrast, having a 2-year-old child roam my house was terrible. I kick myself for not making more up-front investments in kid-proofing and creating safe play areas. But it’s expensive and difficult for a parent to outfit their own home perfectly for each stage of development. The great thing about a daycare classroom for 3-year-olds is that it is perfectly fitted for 3-year-olds, because 3-year-olds will be cycling through it for the next decade. The physical scale factor makes me a daycare optimist for urban areas. However, as I wrote earlier, things could be trickier for low-density population areas.

The study has given us a lot to think about. I hope the research community can be helpful in continuing to figure out the puzzle.

One thing we can conclude, as Noah says in his blog, is that a compulsory university pre-K would be bad. Forcing families to send 4-year-olds to an institutional program (the way 5-16 kids are regulated) would be an expensive “own goal” policy. I don’t know of anyone seriously considering that, which hopefully means that nobody is.

* As a lab experimentalist, I’m used to being surprised by data. Check out this podcast just recorded with John List. He talks about surprising findings from field experiments. You never know until you run the experiment. Hence, my post in September about a rant about behavioral economics.

** Yesterday, Emily Oster announced that she is leaving Twitter because it had become a toxic place for her. You can still find her at substack, instagram, and other traditional publishing outlets (e.g. her books).

Teaching Price Controls (Poorly)

Economics textbooks differ in their treatment of price controls. None of them does a great job, in my opinion. The reason is mostly due to the purpose of textbooks. Despite what you might suspect, most undergraduate textbooks are not used primarily to give students an understanding of the world. They are often used as a bound list of things to know and to create easy test questions. If a textbook has to change the assumptions of a model too much from what the balance of the chapter assumes, then the book fails to make clear what students are supposed to know for the test.

I think that this is the most charitable reason for books’ poor treatment of price controls – even graduate level books. The less charitable reasons include sloppy exposition due to author ignorance or an over-reliance on math. I honestly would have trouble believing these less charitable reasons.

I picked up 5 microeconomics text books and the below graph is typical of how they treat a price ceiling.

The books say that the price ceiling is perfectly enforced. They identify producer surplus (PS) as area C and consumer surplus (CS) as areas A & B. There are very good reasons to differ with these welfare conclusions.

Problem #1

Continue reading

South Carolina Certificate of Need Repeal

The South Carolina Senate just voted 35-6 to repeal its Certificate of Need laws, which required hospitals and many other health care providers to get the permission of a state board before opening or expanding. The bill still needs to make it through the house, and these sorts of legislative fights often turn into a years-long slog, but the vote count in the senate makes me wonder if it might simply pass this year. That would make South Carolina the first state in the Southeast to fully repeal their CON laws, although Florida dramatically shrunk their CON requirements in 2019.

Source: Mercatus Center at George Mason University

This seems like good news; here at EWED we’re previously written about some of the costs of CON. I’ve written several academic papers measuring the effects of CON, finding for instance that it leads to higher health care spending. I aimed to summarize the academic literature on CON in an accessible way in this article focused on CON in North Carolina.

CON makes for strange bedfellows. Generally the main supporter of CON is the state hospital association, while the laws are opposed by economists, libertarians, Federal antitrust regulators, doctors trying to grow their practices, and most normal people who actually know they exist. CON has persisted in most states because the hospitals are especially powerful in state politics and because CON is a bigger issue for them than for most groups that oppose it. But whenever the issue becomes salient, the widespread desire for change has a real chance to overcome one special interest group fighting for the status quo. Covid may have provided that spark, as people saw full hospitals and wondered why state governments were making it harder to add hospital beds.

Remittances Eye-tracking Experiment: Meet the authors and paper

I am pleased to have been asked to discuss a paper in an ASHE (American Society of Hispanic Economists) session at the 2022 AEA meeting. Our session is “Hispanics and Finance” on Sunday January 9 at 12:15pm Eastern Time.

The paper is “Neuroeconomics for Development: Eye-Tracking to Understand Migrant Remittances”. Here is a bit about each author. Meeting in person is a benefit that I miss this time, since the meeting is virtual.

Eduardo Nakasone of Michigan State University has several papers on information and communication technologies and agricultural markets. I pondered this sentence from one of his abstracts, “Under certain situations, ICTs can improve rural households’ agricultural production, farm profitability, job opportunities, adoption of healthier practices, and risk management. All these effects have the potential to increase wellbeing and food security in rural areas of developing countries. Several challenges to effectively scaling up the use of ICTs for development remain, however.” His prior work on ICTs is relevant to the paper at hand, which is about how migrants utilize information about remittance tools.

Máximo Torero is the Chief Economist of the Food and Agriculture Organization (FAO). He has worked on development and poverty in many capacities including at the World Bank.

Angelino Viceisza, an associate professor at Spelman College, is doing interesting work at the intersection of Development and Experimental Economics. Here is his 2022 paper (Happy New Year!) published in the Journal of Development Economics.  

I am discussing their paper on how migrants choose financial services. The pre-analysis plan is public. Remittance sending is important for migrants and for the entire world economy. The authors remind us that a significant chunk of what migrants earn is “lost” to service fees. The authors are examining how migrants incorporate new information about competitive alternative services.

Some neat aspects of their work:

  • Their subject pool is migrants who send remittances, recruited in the DC area.
  • Like most experiments I am used to, the stakes are real and significant.
  • Not only can they observe which service is selected, but by using eye-tracking they can get a sense of what information was salient or persuasive.

It is potentially a big deal for migrants to compare services more rigorously and switch providers more readily. The internet, as least in theory, makes it easy to find information on transaction fees. Policy makers have even proposed subsidizing websites that compare the fees of money transfer operators (MTOs). The authors are trying to understand how such a website might impact behavior. A basic question is: does information in this format affect behavior? A small change in behavior could have a huge impact on the world economy and recipient countries. Imagine if a country currently receiving a billion dollars in remittances had 1% more next year because migrants switched to a more efficient service. Might it be cheaper to nudge people toward low-fee services than to send foreign aid?

Their experiment will reveal whether people make switches based on new information, and it also helps us start to understand which attributes of MTOs migrants consider. Their design includes a treatment manipulation that sometimes emphasizes either transfer speed or user reviews.

If you have read this far hoping for a summary of their results, I will disappoint. Their paper is not public yet and data is still being analyzed. I can say that migrant subjects do sometimes switch their choice of MTO, based on information, in some circumstances. They are more likely to make a switch when the induced stakes are higher. If you tune into the session tomorrow, you will get to hear a summary of preliminary results by the author (not free to public, requires conference registration).

Certificate of Need and Mental Health

Most US states require hospitals and other healthcare providers to obtain a “Certificate of Need” (CON) from a state board before they are allowed to open or expand. These laws seem to be one reason why healthcare is often so expensive and hard to find. I’ve written a lot about them, partly because I think they are bad policies that could get repealed if more people knew about them, and partly because so many aspects of them are unstudied.

States vary widely in the specific services or equipment their CON laws target- nursing homes, dialysis clinics, MRIs, et c. One of the most important types of CON law that remained unstudied was CON for psychiatric services. I set out to change this and, with Eleanor Lewin, wrote an article on them just published in the Journal of Mental Health Policy and Economics.

We compare the state of psychiatric care in states with and without CON, and find that psychiatric CON is associated with fewer psychiatric hospitals and beds, and a lower likelihood of those hospitals accepting Medicare.

Together with the existing evidence on CON (which I tried to sum up recently here), this suggests that more states should consider repealing their CON laws and letting doctors and patients, rather than state boards, decide what facilities are “economically necessary”.

PSNE: No More, No Less

Today marks the 27th anniversary of John Nash winning The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel for his contributions to game theory.

Opinions on game theory differ. To most of the public, it’s probably behind a shroud of mystery. To another set of the specialists, it is a natural offshoot of economics. And, finally a 3rd non-exclusive set find it silly and largely useless for real-world applications.

Regardless of the camp to which you claim membership, the Pure Strategy Nash Equilibrium (PSNE) is often misunderstood by students. In short, the PSNE is the set of all player strategy combinations that would cause no player to want to engage in a different strategy. In lay terms, it’s the list of possible choices people can make and find no benefit to changing their mind.

In class, I emphasize to my students that a Nash Equilibrium assumes that a player can control only their own actions and not those of the other players. It takes the opposing player strategies as ‘given’.

This seems simple enough. But students often implicitly suppose that a PSNE does more legwork than it can do. Below is an example of an extensive form game that illustrates a common point of student confusion. There are 2 players who play sequentially. The meaning of the letters is unimportant. If it helps, imagine that you’re playing Mortal Kombat and that Player 1 can jump or crouch. Depending on which he chooses, Player 2 will choose uppercut, block, approach, or distance. Each of the numbers that are listed at the bottom reflect the payoffs for each player that occur with each strategy combination.

Again, a PSNE is any combination of player strategies from which no player wants to deviate, given the strategies of the other players.

Students will often proceed with the following logic:

  1. Player 2 would choose B over U because 3>2.
  2. Player 2 would choose A over D because 4>1.
  3. Player 1 is faced with earning 4 if he chooses J and 3 if he chooses C. So, the PSNE is that player 1 would choose J.
  4. Therefore, the PSNE set of strategies is (J,B).

While students are entirely reasonable in their thinking, what they are doing is not finding a PSNE. First of all, (J,B) doesn’t include all of the possible strategies – it omits the entire right side of the game. How can Player 1 know whether he should change his mind if he doesn’t know what Player 2 is doing? Bottom line: A PSNE requires that *all* strategy combinations are listed.

The mistaken student says ‘Fine’ and writes that the PSNE strategies are (J, BA) and that the payoff is (4,3)*.  And it is true that they have found a PSNE. When asked why, they’ll often reiterate their logic that I enumerate above. But, their answer is woefully incomplete. In the logic above, they only identify what Player 2 would choose on the right side of the tree when Player 1 chose C. They entirely neglected whether Player 2 would be willing to choose A or D when Player 1 chooses J. Yes, it is true that neither Player 1 nor Player 2 wants to deviate from (J, BA). But it is also true that neither player wants to deviate from (J, BD). In either case the payoff is (4, 3).

This is where students get upset. “Why would Player 2 be willing to choose D?! That’s irrational. They’d never do that!” But the student is mistaken. Player 2 is willing to choose D – just not when Player 1 chooses C. In other words, Player 2 is indifferent to A or D so long as Player 1 chooses J. In order for each player to decide whether they’d want to deviate strategies given what the other player is doing, we need to identify what the other player is doing! The bottom line: A PSNE requires that neither player wants to deviate given what the other player is doing –  Not what the other player would do if one did choose to deviate.

What about when Player 1 chooses C? Then, Player 2 would choose A because 4 is a better payoff than 1. Player 2 doesn’t care whether he chooses U or B because (C, UA) and (C, BA) both provide him the same payoff of 4. We might be tempted to believe that both are PSNE. But they’re not! It’s correct that Player 2 wouldn’t deviate from (C, BA) to become better off. But we must also consider Player 1. Given (C, UA), Player 1 won’t switch to J because his payoff would be 1 rather than 3.  Given (C, BA), Player 1 would absolutely deviate from C to J in order to earn 4 rather than 3. So, (C, UA) is a PSNE and (C, BA) is not. The bottom line: Both players must have no incentive to deviate strategies in a PSNE.

There are reasons that game theory as a discipline developed beyond the idea of Nash Equilibria and Pure Strategy Nash Equilibria. Simple PSNE identify possible equilibria, but don’t narrow it down from there. PSNE are strong in that they identify the possible equilibria and firmly exclude several other possible strategy combinations and outcomes. But PSNE are weak insofar as they identify equilibria that may not be particularly likely or believable. With PSNE alone, we are left with an uneasy feeling that we are identifying too many possible strategies that we don’t quite think are relevant to real life.

These features motivated the later development of Subgame Perfect Nash Equilibria (SGPNE). Students have a good intuition that something feels not quite right about PSNE. Students anticipate SGPNE as a concept that they think is better at predicting reality. But, in so doing, they try to mistakenly attribute too much to PSNE. They want it to tell them which strategies the players would choose. They’re frustrated that it only tells them when players won’t change their mind.

Regardless of whether you get frustrated by game theory, be sure to have a drink and make toast to John Nash.

*Below is the normal form for anyone who is interested.