For a better browsing experience, please upgrade your browser.


Last week, the NFL suspended Baltimore Ravens running back Ray Rice for two games over an incident in which Rice was charged with knocking his fiancee (now wife) unconscious during an altercation in a New Jersey casino. Many have criticized this suspension as too lenient, particularly compared to the usual four-game suspension handed down for violating the league’s substance abuse policy. Whether Rice’s punishment is fair touches on a number of delicate issues, including the NFL’s history and role in punishing off-field conduct, as well as its authority and obligations under the collective bargaining agreement.

But Rice’s offense — he pleaded not guilty and instead will participate in a pretrial intervention program — is indicative of a larger pattern in arrests of NFL players; they have been particularly prone to domestic violence arrests.

Although there seems to be an endless stream of stories about NFL player arrests and misconduct, this is largely because there are a lot of NFL players (and they’re famous). At the league’s peak (during training camps), there are about 2,560 players attached to NFL teams (limit 80 each). As I’ll show, arrest rates among NFL players are quite low compared to national averages for men in their age range — but there are some types of crimes that trail the pack significantly.

This data was tricky to work with. For NFL arrests, the most comprehensive source I could find was the USA Today NFL Arrests Database, which goes back to 2000 and is updated through the present (I calculated rates based on the 2,560 players per year estimate). For national arrest trends, I used the Bureau of Justice Statistics’ Arrest Data Analysis Tool to find the arrest rates per 100,000 for the male population in the 25-to-29 age group (this group is the most similar to the NFL as a whole, where the average team age varies from 25 to 27 years old). The difficulty is that the two sources code offenses differently, so I had to make several grouping choices. There’s a full explanation of my methodology at the end of this piece.

As you would expect from a much more affluent group (e.g. the poverty rate among NFL players is zero), NFL players have much lower arrest rates than average — basically across the board:


The most common arrests among the general public are for drug-related offenses and DUIs. The most common among NFL players is DUI, with assault a distant second. Overall, NFL players’ arrest rate is just 13 percent of the national average. But this isn’t distributed evenly across crimes in the slightest:


Note that murder scores relatively high, but the raw numbers are extremely low (there are two in the database, though a third case — domestic in nature — resulted in suicide). But there are 83 domestic violence arrests, making it by far the NFL’s worst category — with a relative arrest rate of 55.4 percent.

Although this is still lower than the national average, it’s extremely high relative to expectations. That 55.4 percent is more than four times worse than the league’s arrest rate for all offenses (13 percent), and domestic violence accounts for 48 percent of arrests for violent crimes among NFL players, compared to our estimated 21 percent nationally.

Moreover, relative to the income level (top 1 percent) and poverty rate (0 percent) of NFL players, the domestic violence arrest rate is downright extraordinary. According to a 2002 Bureau of Justice Statistics Report covering 1993 to 1998, the domestic victimization rate for women in households with income greater than $75,000 (3.3 per 100,000) was about 39 percent of the overall rate (8.4 per 100,000), and less than 20 percent of the rate for women ages 20 to 34. That report doesn’t include cross-tabs, and it’s a little out of date (more current data is harder to find because more recent BJS reports on the issue do not include income breakdowns), but that sub-20 percent relative victimization among high-income households is consistent with the NFL’s 13 percent relative arrest rate overall (arrest disparities between income levels are probably even greater than victimization rates).

Indeed, perhaps the question of how the NFL should “police” its players is the wrong analogy entirely. This situation may be more akin to tort law than criminal law: If the NFL is capable of reducing any harm its players are causing — whether through harsher suspensions or other policies targeting behavior — it may have a legal (or at least moral) duty to do so.

Methodology: Now, I’m not experienced working with crime data, so I apologize if I grouped things unconventionally:

  • Assault: The BJS statistics make a distinction between aggravated assault and regular assault (the NFL data does not), but they don’t break out battery, attempted murder or manslaughter (NFL data does). So I’ve grouped all of these under “assaults” on both sides.
  • Sex offenses: The BJS statistics break down sex crimes into rape and non-rape, while the NFL arrest data is broken down by assault and non-assault. Therefore, I’ve combined those categories on each side into “sex offenses.”
  • Gun-related: Because the BJS data specifically says its gun data is of the carrying/possession variety, I’ve combined other types of gun offenses in the NFL data with their respective crimes instead of in their own group.
  • Prostitution-related: The NFL data breaks out pimping and solicitation, and the BJS summary data doesn’t, so I’ve combined these into “prostitution-related.”
  • Disorderly conduct: In the NFL data, it appears that some “disorderly conduct”-type crimes are listed individually, and some aren’t. I took my best guess and included everything that sounded like disorderly conduct, including: alcohol-related crimes, resisting arrest, criminal mischief, disturbing the peace, evading police and public intoxication.
  • Drugs: There were 11 cases in which a player was arrested in connection with something nonviolent while in possession of both drugs and guns. Because it was unclear to me whether all of those guns were illegal or not, I grouped these as drug offenses.
  • Domestic violence: The BJS summary data bunches domestic crimes with their respective counterparts. Because the vast majority of these are assault cases, I’ve broken down its assault data into domestic and nondomestic, based on another BJS report which states that approximately 21 percent of violent crimes are domestic. (How exactly that translates into arrests is very tricky; I tried calculating it from the huge National Incident-Based Reporting System data set myself and got a tentative number closer to 15 percent, but to be conservative, I’ve stuck with the 21 percent baseline for purposes of this analysis).
  • Finally, note, of course, I’m not saying that all of the players arrested are guilty, and only a small number are ever prosecuted or disciplined (which is also true to varying degrees for the general public).

comments Add Comment

It wasn’t long after the Seattle Seahawks put the finishing touches on their 43-8 Super Bowl rout of the Denver Broncos that media people began throwing around the d-word, dynasty, as they peered into the franchise’s future.

Of course, this happens with just about every Super Bowl winner; squint hard enough, and even the most obvious one-and-done champ looks like a perennial powerhouse. (In some ways, talk of that nature gets even more far-fetched with each passing season — we haven’t seen a repeat Super Bowl winner since the 2004 New England Patriots.) But in Seattle’s case, it might not be totally implausible to expect an elevated probability of a full-blown dynasty.

Historically, teams that have won a title find themselves surprisingly well-positioned to win more of them. Since the 1970 NFL-AFL merger, 55.9 percent of Super Bowl winners won at least one more championship within the following 10 seasons. Even within that club, though, Seattle is starting from a better spot than most. Its schedule-adjusted pythagorean winning percentage during the 2013 regular season ranked sixth among all Super Bowl champions since the merger, trailing only the 1985 Chicago Bears, 1991 Washington Redskins, 1975 Pittsburgh Steelers, 1996 Green Bay Packers and 1973 Miami Dolphins.

More importantly, the Seattle’s core is incredibly young for an NFL champion. Weighted by the Approximate Value produced by each player on the roster, the Seahawks’ average age (26.0) was the second-youngest in the league a season ago and ranked third-youngest among Super Bowl winners since the merger. The two champs who had lower average ages? The 1974 Steelers and 1981 San Francisco 49ers, each of whom would go on to win three more Super Bowls apiece in their next decade of play. (Seattle also ranks as slightly younger than the 1992 Dallas Cowboys, who won two additional rings in a dynastic run.)

Looking at all Super Bowl winners from 1970 to 2003 (for which we have a “next decade” worth of data), there’s a relationship between the team’s AV-weighted age in its championship season and its chances of winning additional titles.


Among the aforementioned 55.9 percent of all Super Bowl champions who won another before a decade was up, a disproportionate number are clustered among the youngest teams on the list. Eleven of the 12 youngest champions in our 1970-2003 group went on to win at least one more Super Bowl in the following decade, while only four of the 12 oldest champs would go on to win another title.

Usually, talk of dynasty potential among freshly christened champions isn’t very predictive. But because of their youth, these Seahawks are in a situation where the odds of winning another championship are particularly heightened.

comments Add Comment

The Bureau of Labor Statistics released an interesting chart Wednesday on paid vacation, building off a larger report on employee benefits published last week.

While overall access to paid vacation has been essentially flat since 2010, there have been significant changes beneath the surface. Take service workers: In 2010, 59 percent of workers in service jobs had access to paid vacation time; today only 55 percent do. But workers in maintenance and repair jobs have seen gains over the same period: 93 percent of them get paid time off, up from 90 percent in 2010.


It’s dangerous to read too much into year-to-year changes, especially when they’re just a few percentage points up or down. But the data does seem to fit with other evidence that the job market is increasingly bifurcated. People with trade skills, a college degree or other in-demand qualities are seeing increased opportunities and, in some cases, faster wage growth. Those further down the skills spectrum are seeing little wage growth; many are also stuck in part-time jobs, which are much less likely to offer benefits. (In 2014, 87 percent of full-time workers had access to paid vacation, and 78 percent got paid sick leave; for part-timers, those figures were 34 percent and 25 percent, respectively.)

comments Add Comment

House Speaker John Boehner said Tuesday that Republicans have no plans to impeach President Obama, and that all the impeachment talk was driven by Democrats hoping to stir up their base.

Boehner’s statement isn’t literally true: There have been mentions of impeachment around the edges of the GOP and by some Republican members of Congress. But on the whole, Democrats are spending a lot more time talking about impeachment than Republicans.

Consider, for example, the Sunlight Foundation’s Capitol Words database, which tracks words spoken in the House and Senate. So far in July, there have been 10 mentions of the term “impeachment” in Congress and four others of the term “impeach.” Eleven of the 14 mentions have been made by Democratic rather than Republican members of Congress, however.

Impeachment chatter has also become common on cable news. On Fox News this month, Sarah Palin, the former Alaska governor, called for Obama’s impeachment, for instance. But for every mention of impeachment on Fox News in July, there have been five on liberal-leaning MSNBC.

This data comes from a Lexis-Nexis search of transcripts on each network. It counts each mention of the words “impeach” or “impeachment.” The terms were used 32 times in a single episode of MSNBC’s “The Ed Show” on Monday. (Ordinarily, I’d adjust for the overall volume of words spoken on each network, but I know from my previous research that MSNBC and Fox News have about the same number of words recorded in Lexis-Nexis.)

The scoreboard so far in July: Fox News has 95 mentions of impeachment, and MSNBC 448. That works out to about 2.7 mentions per hour of original programming on MSNBC, or once every 22 minutes. (This data is as of late Tuesday afternoon.)


MSNBC hasn’t become quite as obsessed with impeachment as CNN was with Malaysia Airlines Flight 370, but it may be getting there. Impeachment mentions on MSNBC increased sixfold from May to July. Overall, since Jan. 1, MSNBC has mentioned impeachment 905 times to Fox News’s 213.

Some of the impeachment discussion from Democratic politicians and liberal commentators has a kind of a Br’er Rabbit quality. “Only please, Br’er Republicans, don’t impeach President Obama!” they say. But Democrats know such a move would be highly unpopular with the public and might be one of the few things that would revive their long-shot chances of recapturing the House of Representatives in November. In the meantime, Democrats have raised a bundle of money — at least $2.1 million — through emails like these:

Nate Silver's inbox

The Democrats’ strategy has a parallel in 2006. That year, Republicans tried to rally their base around purported Democratic threats to impeach President George W. Bush. Back then, Fox News was more likely to invoke the specter of impeachment, with 374 mentions of the term from Jan. 1 to July 29, 2006, compared with MSNBC’s 206.

comments Add Comment

U.S. gross domestic product, the broadest measure of goods and services produced in the economy, grew at a 4 percent rate between April and June, the Bureau of Economic Analysis said Wednesday. That represents a sharp reversal from the first quarter of the year, when the economy shrank. Moreover, that dismal first quarter wasn’t quite as bad as previously believed: GDP shrank at a 2.1 percent rate rather than the 2.9 percent reported last month. All in all, the report should ease concerns — never widely held by economists — the economy was headed for a major slump. But it will take at least a couple of more quarters of strong growth before this starts to look like a genuine acceleration, not just a one-off rebound.

Here are a few more observations from Wednesday’s report:

Yes, this was good news: GDP reports are notoriously complex, and there’s always room for caveats and alternative interpretations. But make no mistake: This was a good report, especially in context. A 4 percent annual growth rate isn’t a boom, but it’s significantly better than the 3 percent economists had been expecting, and it more than offsets the first quarter’s contraction. The growth was broad-based, with consumers, businesses and even the government contributing to the rebound. Wednesday’s report provides strong evidence the first-quarter contraction was a one-off event that was probably due in part to the historically bad winter in much of the country: Consumer spending, homebuilding and exports all struggled in the first quarter and rebounded in the second. The revisions, too, were cause for optimism: The first quarter wasn’t quite as bad as it once looked, and the end of 2013 was stronger, suggesting the economy carried more momentum into the start of 2014 and lost less of it when winter weather hit. All of that gives hope for continued growth.

But watch out for revisions: When the government released its first estimate of first-quarter GDP in April, it said the economy grew ever so slightly, at a 0.1 percent rate. Now the government thinks the economy actually shrank significantly. And though that swing was particularly drastic, big revisions are far from unusual: Initial GDP estimates are based on incomplete data and are notoriously unreliable. There’s particularly good reason to be cautious this time around: Health care spending has been extremely hard to estimate recently because of the changes brought on by the Affordable Care Act. Health spending wasn’t a big factor in Wednesday’s report, but significant revisions are possible.

Mixed signals for the future: Wednesday’s report is backward-looking, covering economic growth from April through June. But it contained some hints about the future, both positive and negative. On the positive side, the government revised down its estimates for growth in 2011 and 2012, and revised up its estimate for 2013; that had the effect of making the overall trajectory of the recovery look more positive. The government also revised up its estimate of how much Americans saved over the past three years, which means consumers might be in a better position to spend going forward. But on the negative side of the ledger, a big piece of the second quarter’s economic growth (more than 40 percent) came from inventories; that is, from companies restocking their shelves and storerooms rather than selling products. That kind of growth is unlikely to continue next quarter.

Prices rise faster: Consumer prices rose at a 2.3 percent annual rate in the second quarter, up from 1.4 percent in the first. (The Bureau of Economic Analysis calculates prices slightly differently than the Bureau of Labor Statistics, which publishes the better-known consumer price index.) The Federal Reserve, which prefers the BEA methodology, targets a long-run inflation rate of 2 percent. The Fed is unlikely to become too concerned by a one-quarter uptick, but Wednesday’s figure will nonetheless lead critics to argue that the Fed needs to raise interest rates soon to keep prices in check. Those critics have been proven wrong repeatedly in recent years, but with the unemployment rate falling and job growth holding strong, their concerns are gaining more widespread support.

comments Add Comment

As the worst Ebola outbreak in history spreads through West Africa, media reports have repeatedly cited health officials who say the virus kills up to 90 percent of people it infects. CNN has skipped the “up to” in several stories this week, reporting that typically 90 percent of people infected with Ebola die from the infection — and that the 60 percent mortality rate during the current outbreak reflects success from early treatment.

Behind these numbers, though, lies considerable uncertainty: about the typical death rate for Ebola, about the death rate for the current outbreak and about what that rate will end up being when the infection is halted.

The World Health Organization, the United Nations’ health arm, has compiled the number of cases and deaths from two dozen prior outbreaks, going back to the first outbreak, in 1976. Technically, the mortality rate per outbreak has ranged from 0 percent to 100 percent: Once in the Ivory Coast two decades ago, one person was infected and that person survived; and three times, a single infected person died.

Ignore those single cases, and the mortality rate has ranged from 25 percent (among 149 infected people in Uganda in 2007) to 90 percent (among 143 infected people in Congo in 2003). Overall, of 2,387 people diagnosed with Ebola in prior outbreaks, 1,590 people died — almost exactly two out of three.

In the four months of the current outbreak in Sierra Leone, Liberia and Guinea (and most recently Nigeria), the death rate has been lower: 56 percent through last Wednesday, the latest available numbers from the WHO.

Some news articles and doctors have attributed the lower death rate to catching cases early and providing infected people with treatment. There is no cure for Ebola, but hydration, electrolytes and bed rest can help manage symptoms, especially diarrhea. “The mortality rate in some outbreaks can be as high as 90 percent, but in this outbreak, it is currently around 60 percent, indicating that some of our early treatment efforts may be having an impact,” Stephan Monroe, deputy director of the U.S. Centers for Disease Control and Prevention’s National Center for Emerging Zoonotic and Infectious Diseases, said in a press briefing on Monday.

But there are several reasons to interpret the reported death toll as very preliminary; some reasons mean it might rise from here, while others mean it might fall. “There are things that could cause the overall number to be underestimated or overestimated,” Gregory Härtl, a spokesman for the WHO, said in a telephone interview on Tuesday. “That’s why you really have to take these numbers with a grain of salt.”

First, here are reasons why the true mortality rate of the current outbreak might be higher than current figures suggest:

  • The outbreak isn’t over yet. The disease often takes weeks to progress once there are symptoms. The WHO added more than 300 people to its count of cases in the 11 days through July 23. Those people aren’t out of the woods. In the latest numbers, just 12 deaths but 108 new cases were added over three days, for a mortality rate of one in nine. Before that update, the mortality rate had been between 60 percent and 63 percent for each of the last eight updates, every few days over three weeks. “I’d still expect about another 200 to 250 people to die from this even if not another single case occurs,” Benjamin Neuman, a virologist at the University of Reading’s School of Biological Sciences in the U.K., said in a telephone interview on Tuesday. “That will play itself out over the next month. There will be a considerable amount more death, even if the health response is perfect.”
  • Some false positives are likely, and they artificially inflate case totals, suppressing the mortality rate. False negatives also are possible, but less likely. Confirming a case requires multiple tests, and each test creates risk of infection for medical workers, so better to err on the side of caution and quarantine and treat everyone who might be infected. Nearly one-third of the cases counted by the WHO aren’t confirmed. And less than one-quarter of the people whose diagnosis is labeled as suspect have died.

There are also reasons why the true mortality rate might be lower than reported:

  • People who exhibit symptoms common to both Ebola and to many other infectious diseases and then recover are less likely to be tested than are people who are severely ill.
  • And people who die from Ebola without being diagnosed likely will eventually be identified as victims of the virus, said the WHO’s Härtl, because traditional burial practices in the affected countries often lead to infections of people involved in the burials.

Together, those two factors mean the cases could be more undercounted than deaths are. Russian medical official Vladimir Nikiforov made a similar point when arguing in April that the mortality rate in West Africa is an overestimate.

Other factors could go either way, including the mixed accuracy record of labs involved in testing. “There have been some issues with this outbreak with labs having incorrect results,” Härtl said.

When this outbreak is done, it could include the majority of all Ebola cases in history, and perhaps help rewrite the conventional wisdom on the disease. But several factors will complicate the learning process. First, the strain is similar but not identical to prior strains: It is 97 percent identical, genetically, to a strain that struck Congo and Gabon in the past. Second, the same factors that call into question current mortality rates apply to prior outbreaks, meaning any differences could be the result of measurement rather than actual differences in the disease. And third, conducting research on the disease is highly dangerous and restricted to just a few dozen labs worldwide with Biosafety Level 4 clearance.

“It slows down research but keeps people safe,” Neuman said of the lab restrictions. “You have to balance those things.”

comments Add Comment

My colleague Allison McCann was dozing on the New York City subway with her hands in her pockets. When the train’s opening doors woke her, she quickly registered the cellphone and keys she was grasping as she leaped off the car. The relief at catching her stop lasted for half a second; the horror of realizing she had left her bag under the seat lasted longer. Allison spent her first week working at FiveThirtyEight making daily calls to New York’s Metropolitan Transportation Authority (MTA), describing the contents of her bag and hoping, hoping, hoping that some kind soul had turned them in.

The MTA’s lost and found system is vast. It has to be — whether by bus, train or subway, millions of people travel on the network each day, and they leave a bunch of stuff behind. The MTA publishes its lost and found inventory (spotted by Tim Wallace), so we’re able to explore the items on its shelves. The data is updated hourly, so bear in mind that some of the stuffed animals mentioned in the chart below might be claimed, and some extra umbrellas might have been handed in. As of 3 p.m. Monday, there were 168,478 items in the system.

Compared with some of the items abandoned by other commuters, the contents of Allison’s bag (wallet, running shoes, passport) were boring. These partial dentures were in the MTA’s care in 2007:

A set of incomplete dentures at the MTA's  storage area for lost items.

A set of incomplete dentures at the MTA’s storage area for lost items.

Bebeto Matthews / AP

And here are 20 of the weirdest items contained in the MTA’s hoard as of Monday:


It’s not just riders who leave the MTA system lighter than they entered; there’s a category in the lost and found data labeled “New York City Transit Equipment.” As of Monday, that category listed 306 items, including 34 badges, eight vests and seven radios.

But those numbers are nothing compared with the thousands of eyeglasses, keys and shirts (hang on, shirts?!) that have a temporary home in the MTA’s lost property offices (on the mezzanine below the A, C, E lines at Penn Station). Looking at the most commonly lost items, Allison has at least one source of potential solace: She did well to hang on to her cellphone.


The contents of Allison’s bag remain lost. But others who have filled out a lost property inquiry form have had better luck. About 42 percent of the items handed in to lost and found were claimed in 2013, according to Paul Fleuranges, executive director for corporate and community affairs at the MTA. As of Monday, the MTA had reunited 55,921 items with their owners — although the transit riders who left behind 1,813 “undergarments” might not be so eager to step forward.

comments Add Comment

Every Monday, the National Bureau of Economic Research, a nonprofit organization made up of some of North America’s most respected economists, releases its latest batch of working papers. The papers aren’t peer-reviewed, so their conclusions are preliminary (and occasionally flat-out wrong). But they offer an early peek into some of the research that will shape economic thinking in the years ahead. Here are a few of the past two weeks’ most interesting papers:


Title: “Do Large Modern Retailers Pay Premium Wages?”

Authors: Brianna Cardiff-Hicks, Francine Lafontaine, Kathryn Shaw

What they found: Big retailers offer better pay than small companies, especially for more educated workers.

Why it matters: In recent decades, U.S. manufacturing employment has declined steadily, while retail employment has grown. Politicians and policymakers have fretted about the replacement of “good” factory jobs with low-paying retail and restaurant jobs. But in this paper, the authors note that while manufacturing jobs do offer better wages on average than retail jobs, not all retail jobs are low paying. In particular, they highlight the rise of big, modern retail companies, such as Wal-Mart and Starbucks, which pay better, on average, than small retailers. For workers with only a high school diploma, companies with 1,000 or more employees pay 15 percent better than small companies with less than 10 workers. For workers with a college degree, pay is 25 percent higher at big companies. Big companies also employ more college graduates and offer a bigger wage premium for managers (presumably because they’re managing more people). But though the retail sector employs a lot of women, it still has a significant gender-wage gap: Even looking only at cashiers (to avoid the problem that men and women often perform different jobs), and controlling for education and other factors, women earn, on average, about 20 percent less than men.

Key quote: “In ongoing policy discussions, it is suggested that resources should go into bringing outsourced manufacturing jobs back to the U.S., or to improving the training that workers need for today’s manufacturing jobs. … There is an alternative to policies aimed at building up manufacturing. That alternative is to prepare workers to be managers in modern retail firms. A manager in the retail sector makes more per hour than an operative in manufacturing, and the need for managers in retail is greater than in manufacturing, as indicated by the higher proportions of managers in all size firms.”

Data they used: Current Population Survey, National Longitudinal Survey of Youth


Title: “Should Student Employment Be Subsidized? Conditional Counterfactuals and the Outcomes of Work-Study Participation”

Authors: Judith Scott-Clayton, Veronica Minaya

What they found: The Federal Work-Study Program improves grades and graduation rates among students who would have worked while in school anyway, and improves future employment prospects for students who would not have.

Why it matters: The Federal Work-Study Program is a major piece of the government’s student aid efforts, providing subsidized employment for 700,000 students every year, including 1 in 10 full-time freshmen. But there has been relatively little research into the program’s impact. In this paper, the authors find that six years after entering college, work-study participants are 3.2 percentage points more likely to have a bachelor’s degree and 2.4 percentage points more likely to be employed than similar students who didn’t participate in the program. But those figures are a bit misleading. About half of work-study participants probably would have worked anyway and are actually able to work less because of the program; the other half work more hours because of the program. The program affects the two groups differently. Those who would have worked anyway see their GPA rise and graduate at a higher rate, but they don’t see any gains in post-graduation employment. Those who wouldn’t have worked without the program don’t experience the same academic gains and may do worse academically, but they have better work prospects when they graduate. Lower-income students experienced especially large academic gains. Somewhat surprisingly, work-study participants also had significantly higher debt loads.

Key quote: “Groups that we would expect to have a higher likelihood of employment in the absence of the subsidy experience smaller employment impacts, and more positive subsequent impacts. These positive impacts appear driven by reductions in weekly hours and improvements in job amenities (such as on-campus location and relationship to major) for students who would have been working anyway had they not received [Federal Work-Study]. An implication is that effectiveness of Federal Work-Study funds might be increased by modifying the allocation formula — which currently provides disproportionate support to students at elite private institutions — to better target lower-income and lower-scoring students.”

Data they used: Beginning Postsecondary Students Longitudinal Study


Title: “Expanding the School Breakfast Program: Impacts on Children’s Consumption, Nutrition and Health”

Authors: Diane Whitmore Schanzenbach, Mary Zaki

What they found: Expanding access to school breakfast programs has no statistically significant impact on students’ health, attendance or test scores.

Why it matters: School lunch programs have been highly successful at addressing childhood hunger, but the related breakfast program, which launched in the 1960s, has historically reached fewer students, despite evidence that breakfast is important to childhood development and learning. In recent years, there have been efforts to expand use of the breakfast program by offering it to all students, regardless of income, and by serving breakfast in classrooms at the start of the school day, rather than requiring students to arrive early. In this paper, the authors find the in-classroom programs (but not expanded cafeteria-based programs) make children more likely to eat breakfast. But the programs don’t improve health or raise reading or math scores, and provide at most a minimal boost to attendance rates. The authors did find a small but statistically significant improvement in health for certain sub-groups, such as students in high-poverty, urban schools.

Key quote: “Our results do not indicate that the school breakfast program is not effective. There is already a reasonably high program participation rate among the control group, and a higher breakfast consumption rate among the control group, indicating that some children who do not participate in the school program eat breakfast at home. In other words, our results do not shed light on what would happen if the school breakfast program were reduced or eliminated, nor do they suggest that reducing or eliminating the school breakfast program is warranted. The results speak only to attempts to further expand the program, through universal access or [breakfast-in-classroom] programs.”

Data they used: “Data from a randomized experiment implemented in 153 schools across six school districts.”


Title: “The Micro and Macro of Disappearing Routine Jobs: A Flows Approach”

Authors: Guido Matias Cortes, Nir Jaimovich, Christopher J. Nekarda, Henry E. Siu

What they found: Fewer Americans are taking jobs that are easily automated, and more people who have such jobs are leaving the labor force entirely.

Why it matters: One of the hottest topics in economics in recent years has been “job polarization,” the idea, popularized by Massachusetts Institute of Technology economist David Autor, that technology, globalization or other forces have wiped out many middle-skill (and thus middle-wage) jobs, leaving only jobs at the top or bottom of the top or bottom of the skills spectrum. (The thesis, though very influential, hasn’t been universally accepted.) In this paper, the authors look at jobs that involve repetitive tasks that can be performed by computers or robots. Unsurprisingly, they find that fewer people are finding such jobs, and more people are leaving them. But the two trends are being driven by different groups. The decline in people entering routine jobs has been driven by women, prime-age workers (those between 25 and 54) and the more highly educated. The increase in people leaving routine jobs, on the other hand, has been driven by men and by younger workers. Moreover, the rise in people leaving such jobs has been driven by people leaving the labor force (such as to retire or to go back to school) rather than people losing jobs and becoming unemployed.

Key quote: “Our findings suggest that changes in the occupational choices of young workers play a prominent role in accounting for the decline of routine employment. A further empirical analysis of these changes and their implications for entry wages and future career progression would be an interesting direction for future research.”

Data they used: Longitudinally linked Current Population Survey data.

comments Add Comment

The Hill, an inside-the-beltway political newspaper, is out with its annual listing of the alleged 50 hottest people working in D.C. It’s a Washington tradition in which members of an industry predicated on avoiding the press talk about how best to avoid gluten.

You can see the list on The Hill’s website. But in the interest of time, we’ve compiled the major stats you need to know from the piece, Harper’s Index-style.

Number of Democrats: 24

Number of Republicans: 18

Number of Washington politicos attempting to convince us that they — no, really — don’t have an affiliation: 8

Number who seriously believe that: 0

Number of unaffiliated people who host a show for Fox News: 1

Median age: 29

Number of people 25 or younger: 13

Number who are single: 20

Number who are married, engaged or in a relationship: 28

Notoriously coy senators who said they were merely “unmarried”: 1

People who wore a pin on their lapel: 6

Number of those pins that were flag pins: 2

People who, by D.C. standards, are probably not patriots: 48

Profiles that mention the subject’s dog: 4

Profiles that mention the subject’s cat: 0

Profiles that mention “gluten”: 2

Profiles that mention the person’s gym: 18

People who claim to do yoga: 5

People who claim to lift: 3

People who “run”: 22

People who merely “jog”: 1

comments Add Comment

Joe Posnanski had a fascinating piece at NBC Sports last week regarding comments made by Tiger Woods after Rory McIlroy’s British Open victory (the third major title for the 25-year-old). When asked for a reaction to McIlroy’s win — and his championship-winning form of recent years — Woods said:

Well, as you can see, the way he plays is pretty aggressively. When he gets it going, he gets it going. When it gets going bad, it gets going real bad. It’s one or the other. If you look at his results, he’s kind of that way. Very similar to what Phil [Mickelson] does. He has his hot weeks, and he has his weeks where he’s off. And that’s just the nature of how he plays the game – it’s no right way or wrong way.

Posnanski’s (likely correct) reading of the remarks is that Woods attributes a high-variance playing strategy to McIlroy, implying the young champion is willing to accept bad rounds in exchange for dazzling ones. (It’s hard not to also read between the lines of Woods’s comments; he seems to be contrasting McIlroy’s bargain with variance against his own brand of consistent brilliance when he was at his peak.)

It seems obvious that some golfers are inconsistent and some are steady. (Padraig Harrington made the same comparison between the supposed streakiness of McIlroy and Mickelson last summer.) But as we’ve seen in other sports, such as basketball, the human mind is wired to find patterns and attribute significance to sequences that often turn out to be totally random. So, is Woods’s perception of McIlroy off-base?

At first glance, Woods seems right. If we look at the standard deviation of round-by-round major-tournament scores (relative to the field average) for players who have won multiple majors since 1958 (looking only at the years between their first and last major), McIlroy tops the list as the least consistent:


There’s also a rhyme and reason to the list based on Woods’s reasoning. In addition to McIlroy ranking first (and, coincidentally, Harrington ranking second), John Daly — known primarily as a volatileundisciplined long bomber — sits third. And, limiting the data to multi-major winners since 1980, a regression between the most common PGA Tour skill statistics (driving distance and accuracy, greens in regulation percentage, putts per round and sand save percentage, all relative to the tour average) reveals a statistically significant relationship between increased driving power and round-to-round inconsistency in majors, which jibes with Woods’s argument.

But if Woods is on to something, then we would expect to find some consistency to a player’s, well, consistency. A pattern of wild round-to-round scoring swings should persist across a player’s whole career. But if we split players’ careers into random groups (I used even and odd years), the correlation between their round-by-round scoring standard deviation in one group of years and the other is just 0.15. That’s low, meaning even though a player like McIlroy has appeared quite streaky in majors so far, we should only expect him to be slightly less consistent than average going forward.

The rest of the supposed streakiness Woods saw in McIlroy is probably just the product of randomness and not intrinsic to his game.

comments Add Comment

More >

Powered by VIP