How to read the history of social interventions

From my latest for National Review:

All newcomers to social policy should receive a mandatory inoculation against easy answers and outsized expectations. Reading through decade after decade of disappointing results from randomized experiments — probably while attached to one of those torture devices that holds the eyelids open — is the best way to do it.

National Review doesn't usually allow illustrations, but the image above is what I had in mind. Read the whole piece there.

New Politico op-ed: "What happened to Warren the watchdog?"

Jason Delisle and I have an op-ed in Politico's magazine that expands on some of the themes we developed in our National Affairs article back in September. This time we contrast Elizabeth Warren’s former role as government watchdog—in which she pointed out that TARP imposed costly market risk on taxpayers—with her current role as evangelist for federal student loans, in which she ignores the cost of market risk. 

Here's a sample:

...[B]efore joining Congress, Elizabeth Warren was a tenacious government watchdog. She charged that the Troubled Asset Relief Program, created in the wake of the 2008 recession, was exposing taxpayers to risky investments and thereby imposing costs that the feds were low-balling. She wrote letters, published reports, testified before Congress and demanded answers from tight-lipped administration officials. It was an admirable performance.

But then she became a politician, herself. In advocating for enhancing the federal student loan program, the now-senator from Massachusetts has taken the other side of her own argument, completely disregarding the cost of market risk. And she’s not alone in her inconsistency. Government accounting has become a political tool with politicians selectively hiding or exposing costs, depending on whether they support the program in question.

Read the whole thing there. I also referenced the op-ed in a recent blog post for National Review.

A couple of days after our op-ed was posted, Jared Bernstein wrote a response piece titled, "Don't pick on Elizabeth Warren." At first he seems to mount a defense of her inconsistency by saying that she was merely following the law, which requires that TARP receive a different accounting treatment than student loans. But then he concedes that this defense really isn't valid, since Warren has been enthusiastic about her advocacy in both cases. His remaining points are standard critiques of fair-value accounting that we addressed in our National Affairs article.

There is room for some agreement, however. Bernstein writes: "Yes, we fight about some weird stuff in D.C." I'll second that one.

The value of a college major... in the American Community Survey

Administrators of the enormously useful Integrated Public Use Microdata Series (IPUMS) web site sent an email this week warning of changes to the American Community Survey (ACS), which is a mini-Census conducted each year. Specifically, the Census Bureau is proposing to eliminate ACS questions dealing with marital history and college majors. IPUMs administrators, who streamline the ACS (and other datasets) for easy use by researchers, are concerned about the loss of those questions.

It's not my place to declare what the Bureau should do with its limited budget. I obviously haven't performed the cost-benefit analyses needed to determine which ACS questions should stay and which should go.

But, cost aside, the ACS question about college majors is valuable to researchers. Labor economists in particular are forever looking for a robust measure of human capital. Knowing whether a person graduated from college is useful, but it's even more useful to know the field the person studied. After all, workers with engineering diplomas are likely to possess more valuable skills (in terms of both knowledge and raw ability) than workers who studied, say, nineteenth-century French poetry.

Without a strong measure of human capital, it's difficult to investigate questions about the returns to education. It's also difficult to match workers with similar skills. When Andrew Biggs and I analyzed the wages of public- versus private-sector workers, we did our best to generate an apples-to-apples comparison by controlling for human capital differences. We discovered that controlling for college major -- as opposed to controlling only for college graduation -- made a significant difference in our results.

The table above comes from our analysis of Gov. Scott Walker's public-sector reforms in Wisconsin. It shows the wage premium (a positive value) or wage penalty (a negative value) associated with working for the Wisconsin state government relative to the private sector. The premium/penalty varies depending on the dataset and the skills controlled for. But notice the difference between the last two rows. The premium is 1.0 percent when using the standard controls along with PUMAs (geographic areas). But the premium goes up to 5.6 percent when degree fields are included.

Apparently, Wisconsin state workers tend to hold college degrees in fields that are less valuable than the degrees held by college grads in the private sector, with obvious implications for a pay comparison. That's the kind of result that we would not be able to observe in future years if the ACS changes go through.

The problem with big tents

John Fund has a piece for NRO (where I’m a contributor) titled, “More Non-White Voters for the GOP.” I’m not as optimistic as he is. The improvement in the Asian vote -- 58-40 for Democrats in the 2010 midterm, but 50-49 for Republicans in 2014 – is dubious given the tiny exit poll samples. And if there was any significant change among Hispanic voters, it was away from the GOP: 60-38 for Democrats in 2010, now 62-36 in 2014.

Let’s also remember that the major challenge of coalition building is adding without subtracting. It would be great if Republicans could add more minority voters, ceteris paribus. But as the perennial debate over whether the GOP should emphasize economic or social issues demonstrates, appealing too much to one group can alienate another.

I’m not saying that ethnic voting blocs must necessarily line up against each other, but elections often seem to shake out that way. The Florida governor’s race might be one such example. Fund writes:

Republicans suffered a disappointment...in Florida, where incumbent GOP governor Rick Scott’s share of the Hispanic vote fell from 50 percent in 2010 to 38 percent this year. Scott won the election in both 2010 and 2014 by a single point, which makes demographic comparisons easy — and troubling for Republicans.

Left unstated, perhaps because it’s mathematically obvious, is that Gov. Scott was still able to win in 2014 because he increased his share of the white vote enough to wash out his losses among Hispanics. This is consistent with Sean Trende's observation that demographic change may not doom the GOP as long as working-class whites respond by abandoning the Democrats. Whether this is a sustainable or desirable strategy for Republicans, I don’t know. But the trade-off is evident.

"The case for fair-value accounting" in fall issue of National Affairs

I mentioned back in May that Jason Delisle and I were working on a long article about fair-value accounting. It's done, and you can find it in the fall issue of National Affairs--a periodical so serious, it doesn't even have pictures on its cover! The full text is here, but readers may find it easier to download the pdf and print out the whole thing.

Read pp. 95-111 !

I've written a lot about fair-value accounting--e.g., for Heritage, Forbes, and National Review--but up until now I had not given the issue the detailed treatment that it deserves. This article is our full statement on what fair-value accounting is, why it's needed, and how both parties have been inconsistent in their support for it.

I realize the topic seems dry, but all I can say is give it a shot, and you'll likely find it thought-provoking, and perhaps even entertaining. After putting together example after example of bogus "free lunch" accounting, even I was surprised at the extent of the government's malfeasance. And although most people realize that political advocates are not intellectually consistent, their flip-flopping on fair-value is one of the starkest (yet largely unrecognized) hypocrisies in all of Washington, DC.

Jason D. has a two-minute summary of the piece on New America's EdCentral site, and I'll update this post as we put out additional promotional material.

Update 9/26:

National Review has posted my own preview of the article, this one focusing on the "Understated Costs and Perverse Incentives" section.

Also, in an interesting development, the Center on Budget and Policy Priorities now opposes applying FVA to Social Security in response to our article:

Jason Delisle and Jason Richwine, writing in the latest issue of National Affairs, correctly note that the logic of our argument [against FVA] is inconsistent with a 2005 CBPP analysis of proposals to invest part of the Social Security trust funds in stocks instead of Treasury bonds.  We concur.  We have re-analyzed our assessment of investing a portion of the Social Security trust fund in equities and now come to a different conclusion than we did in 2005.

To understand what that all means, see the section of our article titled "Fair-Value's Fair-Weather Friends." More on this later.

Update 10/1:

Mentions of the National Affairs article by the Washington Post's Wonkblog, RealClearPolicy, Committee for a Responsible Federal Budget, and Arnold Kling.

Andrew Biggs criticizes the free-lunch Social Security accounting now espoused by CBPP.

I preview the "Fair-Value's Fair-Weather Friends" section for National Review.

Common Core's remarkably weak evidence base

Must this happen so often? The backers of a broad-based political movement tell you their cause is an evidentiary slam dunk, but then you look at the research yourself and find basically nothing there. The Common Core education standards (de facto national standards) are a great example. As I wrote for National Review earlier this week:

...[T]he Center for Education Policy at George Washington University has put together a compendium summarizing over 60 research papers related to Common Core design and implementation. If there is empirical evidence on the importance of strong standards, this is probably the place to find it. Unfortunately, only two papers in the entire compendium are devoted to measuring the impact of Common Core on test scores. Both papers employ the dubious correlation-across-states methodology, and both give mixed results at best.

Read the whole thing there.

Here's a radio interview I did to promote the piece:



And the Washington Free Beacon did a nice summary as well.

"Why we shouldn't raise teacher pay" published in The Federalist

I'd been meaning to write a piece on public school hiring practices, so when the editors of The Federalist offered me a chance to do something for them on education policy, I figured this was the time. Here's the thesis of the article:

Even without the tenure obstacle, putting the best teachers in the classroom is a more challenging problem than many reformers will admit. One of the most common reformist prescriptions is raising teacher pay to attract stronger applicants. The logic seems simple, even obvious. But raising teacher pay will not work. In fact, it could be counter-productive. The reason lies not just with the well-known difficulty in predicting who will be a good teacher, but also with the entrenched hiring system of public schools.

The article focuses on the latter obstacle--the teacher hiring system that often disfavors the brightest applicants--but an entire piece could be written on the difficulty in predicting who will become a good teacher. Yes, smarter teachers tend to be more effective, ceteris paribus, but the correlation is not strong. Even if public schools were eager to take in the best and brightest, doing so would be no panacea.

Here's the conclusion of the piece:

Public school systems need fundamental changes in how they operate to improve teacher quality, and abolishing tenure just scratches the surface. The creeping emphasis on credentials must be reversed. School administrators must be willing to hire promising applicants who never received the standard education-school training. Objective evaluation systems must be adopted and refined. All parties must become comfortable with a process that will increase teacher turnover. And, finally, the public must maintain sober expectations about the value of high-quality teachers, understanding that their effectiveness is naturally limited by the abilities and family situations of the students themselves. To effect all these changes, pundits and policymakers must move beyond their “pay teachers more” mantra. The idea is attractive for its simplicity, but in reality it is no solution at all.

 Read the whole thing here.

Paul Ryan's education plan: Good policy, wrong goal

The education portion of Paul Ryan’s “Expanding Opportunity in America” plan moves federal policy in the right direction, but part of the rationale is troubling.

First, some of the policy proposals. Ryan would turn federal funds for the ineffective Head Start program into a block grant for states to integrate into their early education services. Encouragingly, he would fund multiple randomized experiments to test claims made by preschool advocates. Let’s hope those are multi-site experiments conducted by outside investigators so that the results are most informative. As for K-12 education, Ryan would block-grant Title I-A spending and make it portable, part of a longstanding conservative goal. The block grant would lessen federal involvement in education, and the portability feature would encourage school choice.

These are all steps in the right direction, but I still feel uncomfortable reading through the education section. There is a troubling focus on the desire to “close the achievement gap.” In fact, based on the introduction, much of Ryan’s education plan appears premised on the need for gap-closing.

That’s a misguided and potentially counter-productive goal. Socioeconomic achievement gaps in school are inevitable for a couple of simple reasons: Smarter people tend to attain higher socioeconomic status, and smarter people also tend to have smarter kids. Those are generalities, of course, but in the aggregate it would be shocking if the children of rich kids did not do better in school than the children of poor kids, even if both groups enjoyed the same educational resources.

The gap-closing mindset leads to condemnation of schools serving low-income students as “failing” even when they might actually be doing a decent job. It generates layers of bureaucracy tasked with employing the latest pedagogical fads. It disregards the needs of gifted students looking for creative outlets. And it may ultimately undermine the case for school choice: When education reformers conclude that school choice does little to close test-score gaps, they will move on to alternative reforms that give less power to parents.

The success of an education policy should be measured not by how much it closes gaps, but by the degree to which it tailors instruction to individual student needs. School choice is probably the best way to pursue the latter goal, and I wish the Ryan plan—as positive as it is in most respects—had recognized that.

SPLC: Republican opposition to Common Core is “hate and extremism”

The editors of National Review (where I'm a contributor) warned last week that free speech rights are threatened by the pernicious notion that businesses should not enjoy First Amendment protections. The threat is real, but altering the Constitution like that is still a lot of work. Activists have an easier way to suppress views they don’t like: Just call those views “hate.”

The Southern Poverty Law Center (SPLC) runs a blog titled “Hatewatch: Keeping an Eye on the Radical Right.” It has a daily feature called Hatewatch Headlines, which is dedicated to "highlighting the best stories around the web on hate and extremism.”

On the same day as the NR editorial, Hatewatch Headlines offered a revealing selection. Amidst links to articles about secret KKK members, apologies for slavery, and anti-government terrorism, the SPLC listed a piece from the Washington Post titled: “Forget Obamacare. Common Core is the Republicans’ new big enemy.”



Common Core is the name for a set of K-12 education standards that supporters hope will become de facto national standards. (See here for my perspective.) Needless to say, what opposition to these standards has to do with Klansmen and terrorists is not clear. But the SPLC is happy to link any conservative cause to "hate." That shuts down debate without any need for direct censorship.

The Post article itself is a lazy attack on Common Core opponents, with the usual charge that opponents are against the standards simply because the president is for them. It’s a strange accusation, given that a number of liberal groups have stated their opposition as well. Nevertheless, Republican opponents are portrayed as extreme in the Post article, and the SPLC ups the ante by surrounding the link with variants of the word “hate.”

If even the most innocuous political position—say, preferring that school standards not be uniform nationwide—can be characterized in the media as hate and extremism, then weakening the First Amendment will not be necessary to control the discourse.

About half of teachers have education degrees

I received a request from Education Realist to detail how many and which types of teachers hold undergraduate education degrees, using the ACS data mentioned in the previous post.

The ACS provides five different occupational categories that pertain to public-school teachers. Here is their distribution in the combined 2010 through 2012 samples:

Right away we can see some limitations in the ACS data. It would be nice to separate elementary from middle school, and preschool from kindergarten, but we have to work with what's available. We also have to rely on the Census Bureau to accurately categorize respondents' descriptions of the work they do, and it's not clear that it has. The percentage of secondary teachers appears too low--I expected to see around 25 percent--which suggests some misclassification.

I created three categories for college majors: pure education, subject-specific education, and non-education. Included in the first category is any field that is primarily about pedagogy rather than an academic subject. They are the types of majors that people usually mean when they say "education":

General Education
Educational Administration and Supervision
School Student Counseling
Elementary Education
Early Childhood Education
Secondary Teacher Education
Special Needs Education
Teacher Education: Multiple Levels
Miscellaneous Education
Educational Psychology

By contrast, some teachers major in how to instruct a particular subject. Their courses of study are generally similar (and sometimes identical) to majoring in the subject itself, but with an added teaching component. These are the subject-specific education majors I found in the ACS:

Mathematics Teacher Education
Physical and Health Education Teaching
Science and Computer Teacher Education
Social Science or History Teacher Education
Language and Drama Education
Art and Music Education

The table below shows how pure education, subject-specific, and non-education degrees are distributed among teachers of different grade levels. (Click the table to see a larger version.) Overall, 49 percent of public-school teachers in the ACS have a pure education degree, another 11 percent have a degree in how to teach a particular subject, and the remaining 40 percent have a non-education degree.

The most common non-education job held by education majors is...

The American Community Survey, a mini-Census conducted each year, began asking college graduates for their degree field in 2009. The new information helps answer a number of important questions about both the labor market value of different degrees and the skills possessed by different workers.

One question that's interested me for the past several years is how the skills of public-school teachers compare to the skills of other college graduates. A common view is that teachers should be paid like the average college graduate, but not all four-year degrees are created equal.  Elementary school teachers typically major in education, which is often considered one of the least rigorous courses of study. (Secondary school teachers, being more specialized, usually major in an academic field or in how to teach a particular field.)

One way to better assess teacher skills is to examine what happens to the wage of the typical worker who switches between teaching and non-teaching jobs. It turns out that people generally receive a pay raise when they move into teaching and a small pay cut when they move out, which implies that teacher salaries are at least adequate on average.

Another way to gauge teacher skills is to look at the types of non-education jobs held by education majors. If trained teachers tend to work prestigious jobs outside of teaching, that would suggest their skills are broadly valued and transferable. The first table below shows that 60.6 percent of people with a degree in education (and who have worked within the last five years) have an education-related occupation. The next table lists the top 10 occupations of the 39.4 percent of education majors who have a non-education job. The most common non-education job is administrative assistant, followed by retail salesperson, miscellaneous manager, childcare worker, and first-line supervisor of retail salespersons.

New fair-value cost estimates from the CBO

Though it won't say so explicitly, the CBO has been on a long quest to convince Congress to change its accounting procedures. It's out with another study showing the budgetary cost of credit programs under current-law accounting (FCRA) versus a more comprehensive method known as fair-value accounting (FVA). As usual, credit programs that show a profit under FCRA actually show a cost under FVA.

The analysis of the Export-Import Bank is particularly interesting, since the agency has come under increasing fire for being, in President Obama's words, "little more than a fund for corporate welfare." Ex-Im supporters had cited the "profit" the bank earns for the government as a justification for its existence. No longer.

To read all about FVA, please see my previous post.

Roundup of student loan commentary

The New America Foundation's Jason Delisle and I are hard at work on a 5,000-word piece that discusses fair-value accounting (FVA) and how it applies to issues such as student loans, public pensions, and private retirement accounts. In the meantime, here is a selection of my previous work on student loans and FVA.

"The Unknown Cost of Student Loans" is the foundational document that describes what FVA is and how it changes the cost estimates of student loans and other credit programs. It contains a numerical example of how budgeting works under the current system versus FVA. Though consumed with immigration work last April, I put this report together after-hours, getting it out just in time for the student loan debate in Congress.

"How Government Uses Accounting Tricks to Hide the Student Loan Swindle" is the companion op-ed that I had published in Forbes while the Congressional debate was ongoing. It's a less quantitative explanation of FVA, with application to both student loans and public pensions.

"The Student-Loan Deal: A Mixed Bag" describes how the debate was resolved in Congress. I reiterate my opposition to any federal involvement in student loans but note that the legislation at least pegs the subsidy to market interest rates.

"Attention, Elizabeth Warren: Students Are Already Free to Refinance Their College Loans" debunks the popular talking point that allowing students to refinance their college loans with the government is a matter of leveling the playing field. In fact, students are as free to refinance with a private lender as homeowners and businesses are. The trouble is that students are already getting a great deal on their loans due to government subsidies, and no private lender wants to take full responsibility for unprofitable loans.

"What ‘Profits’? Rolling Stone’s Matt Taibbi Misunderstands Student Loans" and "Sorry Slate, There’s Still No Such Thing as Federal Student-Loan Profits" criticize articles in the popular press for being entirely premised on the idea that student loans earn a profit for the government, even though FVA indicates they come at a cost.

New AEI paper grades all 50 states on public-sector compensation

Public-sector compensation has been politically salient in some states but a non-issue in others. States such as New York, California, and Ohio always seem to be discussing pension shortfalls, "windfall" retirements, excessive job security, etc. But where are these issues in Kansas? Or Virginia?

As Andrew Biggs and I show in a new American Enterprise Institute working paper, there is tremendous variation in the generosity of public-sector compensation (wages plus benefits) across states. Some states offer a large premium to public employees relative to what they would earn in the private sector, while other states appear to pay less than market levels. And, not surprisingly, the political salience of public-sector pay is positively correlated with its generosity.

The following table ranks each state from least generous (relative to private-sector pay in that state) to most generous. Because we shouldn't make too much of small differences between states, the states are sorted into five broad categories:

Table 2. Total Compensation Categories
Category label	Range	States
“Modest penalty”	-6% or less	Virginia
“Market level”	-5% to +5%	Kansas, Indiana, Minnesota, Georgia, West Virginia, Mississippi, South Dakota, North Carolina, Vermont, Colorado, Washington, South Carolina, Kentucky, Idaho, Arizona, Nebraska, Tennessee, Utah
“Modest premium”	+6% to +10%	Alaska, Missouri, Florida, Arkansas, Texas, Oklahoma, Maryland, Iowa, Montana, North Dakota, New Hampshire, Delaware
“Large premium”	+11% to +20%	Alabama, Louisiana, Wisconsin, Oregon, Ohio, Hawaii, Massachusetts, Nevada, Maine, New Mexico, Michigan
“Very large premium”	+20%	New Jersey, California, Rhode Island, Illinois, New York, Pennsylvania, Connecticut
Source: Authors’ calculations

Clearly, studies that evaluate public-sector pay by lumping all state workers together are not helpful for policymakers. The goal of our new paper is to help each state individually decide whether it needs to consider pay reform.

All of the main take-aways come in the first 15 pages of the paper. For those interested in delving deeper--much deeper--we've included a lengthy methodological appendix. The appendix is probably the most comprehensive discussion of technical and corollary issues ever put together on this topic. Want to know more about taking logs in a wage regression? Controlling for firm size? Breaking down the premium by education level? Accounting for salary growth in pension valuation? Adjusting the discount rate on deferred compensation to include market risk? Valuing job security? Considering other job amenities?

It's all there, contained within 87 pages, 14 figures, 5 tables, and 120 end notes. The paper is meant to be as much a resource for academic students of the issue as it is for state policymakers.

The myth of time diversification

Investments become safer the longer you hold them, right? If you think so, your reasoning probably goes something like this: The stock market is cyclical--sometimes it's up, sometimes it's down--but it has a generally upward trajectory. So year-to-year deviations from the average (or "expected") return are likely to cancel out, and you'll end up with the expected return in the long run. In other words, the risk of under-performing expectations over a long period of time approaches zero.

You've fallen for the time diversification fallacy. It's generated by a confused focus on the average yearly rate of return as opposed to what really counts--the cumulative return. Put simply, the risk of a sub-par average yearly return goes down over time, but the effect of such a low return--should it occur--is far more harmful to a portfolio when it occurs over a long period compared to a short period. Therefore, the risk of a large cumulative loss actually goes up the longer an investment is held.

For a numerical example of what I mean, please read the third fallacy discussed in my paper, "Nine Fallacies Used to Defend Public-Sector Pensions."

Or consider what finance professor Zvi Bodie noted about long-term put options. A "put" is basically an insurance policy that an investor can buy. Say you invest $100 in a risky asset that you expect to grow by 10 percent each year. You can purchase a put option that guarantees you the right to sell the asset in one year for the $110 that you're hoping for.

Like all insurance, put options cost money. Bodie noted that if time diversification is valid, then the cost of put options should go down as the time horizon becomes longer. Take your $100 investment. Since its expected annual rate of return is 10 percent, then you should have $110 after the first year and $259 after 10 years. If investments get safer over time, it should cost less to guarantee the $259 after 10 years than to guarantee the $110 after one year.

It doesn't. Bodie pointed out that option pricing models (including the standard Black-Scholes model) show put options becoming more expensive as the applicable time period increases.

Source: Zvi Bodie, "On the Risk of Stocks in the Long Run"

Why does all this matter? Well, it suggests some important investment advice: If you consider an investment too risky over one year, don't invest in it for the long-term either! In a public policy context, the time diversification fallacy is the basis for claims that public pensions need not worry about long-run risk. (They actually should.) It also shows up in the case for "collective" defined-contribution retirement plans. I wrote about collective plans recently for the Agenda.