One of the standard jobs in software development these days is UX Design. User Experience covers “any aspect of a user’s experience with a given system … addressing all aspects of a product or service as perceived by users.” Products like Google Glass make it clear that we should formalize the development process further to include what we can call “Experience of User”, or XU Design. The XU Designer’s job will be to assess and tweak how third parties experience the users of your product or service. Is the XU experience intrusive? Is it annoying? Do our product’s XU Metrics all point in the direction of “Christ, what an asshole?” As the XU specialty develops we can trace its history back to phenomena like people loudly using cellphones in public, or people talking to you while wearing headphones, and the various ways norms and tolerances developed for these practices, or failed to develop. Right now, though, it looks like Google Glass is shaping up to be the leading XU Design disaster of our time.

Now if you’ll excuse me, I need to trademark the term XU Design and start a consulting company.

You can see this point made in somewhat more detail here.

The dataset we collected has—as best we can tell—635 respondents. More precisely it has 635 unique anonymized IP addresses, so probably slightly fewer actual people, if we assume some people voted at work, then maybe again via their phone or from home. Our 635 respondents made 46,317 pairwise comparisons of departments. Now, in any reputational survey of this sort there is a temptation to enhance the score of one’s own institution, perhaps directly by voting for them whenever you can (if you are allowed) or more indirectly by voting down potential peers whenever you can. For this reason some reputational surveys (like the Philosophical Gourmet Report) prohibit respondents from voting for their employer or Ph.D-granting school. The All our Ideas framework has no such safeguards, but it does have a natural buffer when the number of paired comparisons is large. One has the opportunity to vote for one’s own department, but the number of possible pairs is large enough that it’s quite hard to influence the outcome.

It’s not impossible, however. The distribution of votes across our 635 respondents has a very long tail. While 75 percent of respondents registered just over 70 votes before finding something better to do, and 95 percent of respondents were done after about 250 votes, a brave few carried on for much longer.

As you can see, a small number of respondents cast more than 500 votes, and a few lonely souls cast more than a thousand. I found myself wondering whether these few extreme cases materially affected the final rank order. And in the course of answering that question I found what might fairly be described as somewhat suspicious voting patterns in several—but notably not all—of our Supervoters. In particular, the respondent with the very largest number of votes (1425 in total) had two favorite departments. He or she voted on them multiple times in separate contests—more than thirty times apiece—and both departments won every time. (By chance, this voter was never presented with the two in a head-to-head contest.)

Now, it’s possible for this to happen quite straightforwardly: the departments that emerge at the top of the overall ranking are by definition the ones that win all or nearly all of their head-to-head contests. And given the range of disagreement about which departments should win, as reflected in the error bars around the point estimates, there are quite a few such departments. However, in the case of our Supervoter, the favored departments were somewhat further down in the final rankings, making their 100 percent winning streak seem a little odd, particularly given that no other voters shared their view in several high-disparity cases.

To preserve the confidentiality of the voting process—and bearing in mind that I do not have any identifying information whatsoever about the Supervoter in question—I will refer to the improbably favored departments by the pseudonyms “Hoosier University at Flowerington” and “The Fighting University of Our Lady”, and name the associated Supervoter phenomenon “The Fabio Effect”. Two other Supervoters also displayed somewhat suspicious voting patterns, showing a uniform but perhaps difficult-to-justify preference for Cornelius College and Twin-Citiversity, respectively, over all-comers.

The original rankings still stand, all the same, if only because the original error-ranges easily cover the shuffling around that happens once you re-run the model with the offending Supervoters removed. But I thought it would be worth showing what the rank-order looks like when the Fabio Effect is accounted for.

The model doesn’t take account of any rater effects, but given the general state of the U.S. News ranking methodology I am not really bothered. As you can see, the gradation looks pretty smooth. The first real “hinge” in the rankings (in the sense of a pretty clean separation between a department and the one above it) comes between Toronto and Emory. You could make a case, if you squint a bit, that UT Austin and Duke are at a similar hinge-point with respect to the departments ranked above and below them. Indiana’s high ranking is due solely to Fabio mobilizing a team of role-playing enthusiasts to relentlessly vote in the survey. (This is speculation on my part.)

You can do other things with this data, too. Here are the results of a cluster analysis of the votes, which brings out some interesting similarities.

Finally, Baptiste Coulmont carried out his own pairwise-comparison survey of French sociology departments—at least until he was overwhelmed by the pressure exerted by people who thought the very idea of such a ranking was morally offensive—and presented a nice analysis of it on his blog. Inspired by that, here’s are the results of a principal components analysis of the voting patterns.

The x-axis is more or less the overall ranking itself. The y-axis is harder to interpret. The colors in the plot show how long people typically to vote for that department in comparisons, with Slow and Fast categories representing roughly the bottom and top quintles of the distribution of times. High-ranking but “Slow” departments are interesting here, as it suggests they might be difficult to place with respect to their peers.

(With thanks to Martin O’Neill for the original administrative email.)

Surveys were conducted in fall 2012 by Ipsos Public Affairs … Questionnaires were sent to department heads and directors of graduate studies (or, alternatively, a senior faculty member who teaches graduate students) at schools that had granted a total of five or more doctorates in each discipline during the five-year period from 2005 through 2009, as indicated by the 2010 “Survey of Earned Doctorates.” … The surveys asked about Ph.D. programs in criminology (response rate: 90 percent), economics (25 percent), English (21 percent), history (19 percent), political science (30 percent), psychology (16 percent), and sociology (31 percent). … The number of schools surveyed in fall 2012 were: economics—132, English—156, history—151, political science—119, psychology—246, and sociology—117. In fall 2008, 36 schools were surveyed for criminology.

So, following Neal, this tells us the Sociology rankings are based on a survey of 117 Heads and Directors with a response rate of 31 percent, which is thirty six people in total. For Economics you have 33 people, for History 29 people, for Political Science 36 people, for Psychology 40 people, and for English 33 people. The methods page also notes that they calculate the scores using a trimmed mean, so they throw out two observations each time (the highest and the lowest). The upshot is that the average score of a department is likely to have rather wide confidence intervals.

Update: These numbers are too low. Read on.

I guess it’s possible that U.S. News might mean that the effective N of, e.g., the Sociology survey is 117, and that’s the result of a larger initial survey which yielded a 31 percent response rate. On that interpretation they initially contacted 378 departments (or thereabouts). That would be a non-standard way of describing what you did. Normally, if you give a raw number for the sample size and tell us the response rate, the raw number is the N you began with, not the N you ended up with. More importantly, there aren’t 378 Ph.D granting Sociology departments in the U.S.—a quick check of the Survey of Earned Doctorates suggests that there were 167 in 2010. That suggests that 117 is about right for the number who had awarded five or more in the past five years, and that this is the total initial N. Same goes for Economics, which has 179 Ph.D programs in the 2010 SED.

Then again, the wording in the methods can also be read as saying every department received two surveys (“Questionnaires were sent to department heads and directors of graduate studies … at schools that had granted a total of five or more doctorates … during the five-year period from 2005 through 2009”). Looking more closely at the available SED data for 2006 to 2010 (one year off the USNWR dates, unfortunately), I found that 115 Sociology Departments met the stated criteria of having awarded five our more doctorates in the previous five years. If both the Dept Head and DGS in all those departments got a survey, this makes for an initial maximum N of 230. But that is still quite far from the 378 or so needed, if 117 is supposed to mean the 31 percent who responded rather than the total number initially surveyed.

So it seems like the most plausible interpretation is that for Sociology the number of schools surveyed is in fact 117, that every school received two copies of the questionnaire (one to the Head, one to the DGS or equivalent), but that the 31 percent response rate means “percent of schools from which at least one response was received”, and so the total N surveys for Sociology is somewhere between 36 and 72 people, with a similar range of between 30 and 80 for the other departments.

The best thing to do is to implement your own system, and do it better, if only to introduce confusion by way of additional measures. Omar Lizardo and Jessica Collett have already pointed out that U.S. News decided to cook the rankings by averaging the results from this year’s survey with the previous two rounds. They provide an estimate of what the de-averaged results probably looked like. Back in 20011, Steve Vaisey and I ran a poll using Matt Salganik’s excellent All Our Ideas website, which creates rankings from multiple pairwise comparisons. It’s easy to run and generates rankings with high face validity in a way that’s quicker, more fun, and much, much cheaper than the alternatives. So, we’re doing it again this year. Here is OrgTheory/AAI 2013 Sociology Department Ranking Survey. Go and vote! Chicago people will be happy to hear can vote as often as you like. So, participate in your own quantitative domination and get voting. For the duration of the survey, you can even do it right here with this handy widget:

You should think carefully about whether to have kids. It’s a distinctively modern decision. Until comparatively recently, producing an heir, supplying household labor, insuring against destitution, or being fruitful and multiplying was what having a child was about. Nowadays the decision to bear a child is freighted with a more personal significance—assuming you are physically able to do so, and lucky enough to be well-off and well-situated. Children are an enormous responsibility, we are told, and you should be sure you really want to have one before you go ahead and do it. In particular, you’re supposed to reflect carefully on what it would be like. You weigh the options and make a decision.

Crucially, this involves assessments of your future experiences. You imagine your life with and without kids, and think about what it would be like or feel like to have that experience. In the language of philosophers, you must think about the phenomenology of the experience. When it comes to children, people argue endlessly about what you ought to do. Some claim motherhood is a supremely fulfilling vocation. Some wearily raise their hands (after wiping off spit-up milk) and beg to differ. Others see liberation in the decision to avoid parenthood. They complain about the presumptions of a culture that equates child-rearing with happiness or self-realization, or that looks with pity or suspicion on the indecently happy and child-free. Insofar as there is any detente in the Mommy Wars, though, it’s around the idea that you should personally reflect with great care on these issues and decide for yourself whether this … this—what? Grand adventure? Prison sentence?—this experience is for you.

That sounds like a reasonable compromise, until you realize no-one knows what it’s like to have a child, until they have one. It’s a phenomenologically transformative experience. Fear not, veterans of the Mommy Wars. We are not saying it’s wonderful, or that people who don’t experience it have somehow failed at life. We just mean that people are very different afterwards, in ways they cannot anticipate. The evidence for this is everywhere. New parents laugh ruefully at their detailed pre-kid plans to fit “the baby” into their existing lives. “You ruined everything/In the nicest way”, as songwriter Jonathan Coulton says. Those who choose to remain child-free, meanwhile, bridle at the insulting suggestion that they are missing out on something. Yet they see their friends get body-snatched one at a time, cocooned in minivans and unable to stay out past eight, lost to civilized life, unrecognizable. Something happened to them. Even the parent who reacts to their new situation with numb disbelief, or shock and depression, has a transformative experience. These reactions have their own cruel character because they break so sharply with the official story.

Crying baby courtesy of photosavvy.

Stripped of judgmental overtones, the transformative character of becoming a parent is not a controversial idea. The trouble is, transformative experiences throw a wrench in decision-making based on future experiences. In theory, a rational choice is a series of steps: first determine the possible outcomes, and the costs and benefits associated with each one; then assign a probability to each outcome to calculate its value; finally, choose the option that gives the highest expected value. Real decisions are rarely so clean cut, because we are imperfect calculators and it is probably impossible to figure expected values with precision anyway. Yet this is the decision-making standard we aspire to. For it to work, you must at least be able to assess the costs and benefits of the most important outcomes.

But in this case, the most important outcomes include things like “what the experience will be like for me” or “what it will be like to be a parent”. If becoming a parent is a transformative experience, you can’t know in advance what it will be like for you. You can’t assess the costs and benefits of these outcomes, since you can’t know their values—and so if you choose based on what you think it will be like for you, you can’t even approximate a rational decision-making procedure. Our ordinary understanding of the choice to have a family or remain childless—all that careful weighing of options based on what it’s going to be like for you in the future—is based on a fantasy. You don’t know what it is going to be like. So you can’t rationally make the choice by weighing options involving the experience of parenthood.

You probably have some objections. You might say, “What if I decide to have a child solely because I want to pass along some DNA?” Or, “What if I decide to remain child-free solely because there are too many people on this earth already?” That’s fine. If you’re really not basing your decision at all on what being a parent is going to be like for you then you can make a rational decision. But relying only on criteria like that is not the usual way to decide to have kids.

You might say, can’t a rational decisionmaker adopt a different decision rule, one specially designed to deal with difficult choices? She can—but at a price. For example, consider a play-it-safe rule that says, “Simply choose the option whose worst case scenario is the best one relative to every other option’s worst case scenario.” This rule could help you choose between options without incorporating any special knowledge about what it would be like for you to have a child. Sounds reasonable, but it leads to strange results when considered from any particular individual’s point of view. Take Suzy, for instance, who believes she’d love to have a baby. If the best worst-case scenario involves not having a child (if this is better than having a child and bitterly regretting it, say), then if Suzy follows the play-it-safe rule, she should stay child-free, regardless of her own feelings—as should everyone else following the rule.

What about testimony from people like yourself? Can’t you look at them and rationally expect to have a similar experience if you make a similar choice? No. Without just the sort of self-knowledge you’d get from your own experience of having a child, you can’t know how the experience will affect you, and so you can’t know whether you’re more like the parent or the child-free person. As the saying goes, you get experience just after you need it. Even worse, parental testimony is unreliable. A parent may claim she is happier now than she would have been if she had not had her baby, but that may be because she cannot truly imagine life without it. Once a person has had a child, it becomes psychologically very difficult for her to assess what it would be like if she’d never had it. So even after having the child, she probably can’t weigh the different outcomes.

What about making a simple bet? Shouldn’t you just play the odds and choose to have a family, setting aside what you personally think it will be like? Isn’t it just obvious that having a child will make you happier? The standard account of choosing to be a parent certainly reinforces this view, with its endless talk of deep fulfillment. But the evidence suggests that’s nonsense. The highs may be higher for parents, but the lows are lower. Measures of overall personal happiness suggest that parents with children at home are less happy than those without children. Moreover, individuals who have never had children report similar levels of life satisfaction as individuals with grown children who have left home. If you merely want to play the odds, one shouldn’t have a child. But does that mean you shouldn’t have a child? No! You might be one of the people who find the experience of parenthood fulfilling.

We are not arguing that it is right or wrong to have a child. Nor are we saying people shouldn’t be happy with their choice. You can be happy with a child or blissfully child-free. But if you are happy, you shouldn’t congratulate yourself on your wise decision—you should be thankful for your good luck. Choosing to have a child involves a leap of faith, not a carefully calibrated rational choice. When surprising results surface about the dissatisfaction many parents experience, telling yourself that you knew it wouldn’t be that way for you is simply a rationalization. The same is true if you tell yourself you know you’re happier not being a parent. The standard story of parenthood says it’s a deeply fulfilling event that is like nothing else you’ve ever experienced, and that you should carefully weigh what it will be like before choosing to do it. But in reality you can’t have it both ways.

Rather than try to cover all the many great features of every package the kit installs—all the products of a terrific amount of work by other people, and some of which, like AucTeX and ESS, are very large and sophisticated—I am just going to mention some of the useful things you might immediately use once you’ve installed the starter kit.

1. Ido-mode and Smex mean you don’t have to Memorize a Huge Battery of Shortcuts Like any powerful text editor, time-saving keyboard shortcuts are central to using Emacs efficiently. While it’s easy to grasp the general logic of Emacs key-chords, getting specific ones into your muscle memory takes longer, especially for features which you might not use all that often. In the interim, the SSSK sets up ido-mode and Smex, and together they allow you to search buffers, file names, and functions quickly and flexibly. They also remember what you’ve searched for before, so they are more likely to prompt you with the right option the next time. So, for example, if you want to browse the kill-ring (Emacs’s multi-item clipboard) but you don’t remember the keyboard shortcut, starting to type M-x browse-kill-ring will complete the command very quickly. If you can’t remember the command to insert multiple cursors in a region (see below), starting to type M-x mc/mark will bring up all the commands that match, and you can use C-s to cycle through the options. Thanks to ido-mode, search functionality for finding files, open buffers, functions, and other commands is very smart and flexible. If you partly remember the name you’re looking for, or type a shortened or partial version of it, it will likely be found, and it gets better the more you use it.

2. Use M-x without the Meta key Most Emacs commands are accesed either via C-x or M-x. As is standard, Meta is mapped to the Mac keyboard’s “Option” key. But you can also invoke M-x by doing C-x m, which means most things can get done using either C-x or C-x m. It’s a little more convenient.

3. Search inside multiple files with rgrep or ack You can search for text inside files in your current directory (and recursively all directories underneath it) in one of two ways. C-x C-r invokes rgrep, which will ask you for a search term and a file type. If you have Ack installed, you can do C-x C-a to search all files in your working directory for some term. You can also do M-x ack-same to search only in files of the same type as the current buffer.

4. Autocomplete text as you write Autocomplete mode is useful for completing functions in the scripting or programming languages it knows about (including R), and it can also be useful in long text documents. It’s turned off by default for most text modes (except org-mode) in the SSSK, but you can enable it in a buffer with M-x auto-complete-mode.

5. Quickly Resize, Rotate, and Cycle through Windows Early on when using Emacs you learn to split a frame into two or more windows with C-x 2 or C-x 3. As you work (e.g., with R code in the left-side buffer and an R session running in the right-side buffer), you often want to move the cursor to a different window, move buffers from one window to another, or cycle back and forth between different window configurations you’ve used. The starter kit lets you move the cursor between windows using Shift and the arrow keys. C-c m “rotates” windows, e.g. by moving the left window to the right side and vice versa. C-c-<up> and C-c-<down> cycle through the window configurations you’ve created in the past. Shift-C-<left>, Shift-C-<right>, Shift-C-<up>, and Shift-C-<down> resize split windows a little bit at a time.

6. Make Shift-Enter do a lot in ESS When working with R and ESS, if you open an R file in the main buffer then htting Shift-Enter vertically splits the window and starts R in the right-side buffer. If R is running and a region is highlighted, shift-enter sends the region over to R to be evaluated. If R is running and no region is highlighted, Shift-Enter sends the current line over to R. Repeatedly hitting Shift-Enter in an R file steps through each line (sending it to R), skipping commented lines. The cursor is also moved down to the bottom of the R buffer after each evaluation.

7. Turn on Highline-mode when you need it Some people find it easier to edit code when they can see exactly where the cursor is on the screen. M-x highline-mode adds a highlight to the current line, so you know where you are.

8. LaTeX Symbols display nicely in the buffer The kit includes latex-pretty-symbols. When editing a text file, you can type TeX commands like \sigma and \gamma and they will display properly (as unicode symbols) right in the buffer. The LaTeX command is still there underneath.

9. Multiple Cursors One of the flagship features of Sublime Text 2, editing with multiple cursors can be a really useful dynamic alternative to search-and-replace when you want to make the same changes to a bunch of lines in a file. Thanks to Magnar Sveen, multiple cursor support is now available in Emacs and the SSSK installs the package. The easiest way to see it in action is to select a region and do M-x mc/edit-beginning-of-lines or select a region and try M-x mc/mark-all-in-region to search for a bit of text to replace on each line.

10. Minimal Mode When working full screen, I like to make scroll bars and the dividing lines between windows more unobtrusive. M-x minimal-mode (or C-c s for short) toggles them on and off.

There’s a lot more in the SSSK, from handy functions to a few themes (try M-x load-theme solarized-dark or zenburn) and you can browse the notes in the kit’s org files to learn more about what it sets up for you, and how to customize it to your own needs.

The main problem I encountered while testing the new version is a certain flakiness on the part of Emacs when it contacts the ELPA and Marmalade servers to download the various packages. I’ve encountered three different errors: a long period of hanging without response; a generic “can’t parse HTTP buffer” message; and a specific insistence that “The package ‘auctex’ is not available for installation”. The latter one in particular is strange, because it’s definitely there in the ELPA package repository.

Dealing with these errors is annoyingly close to magical handwaving. The first option is to simply quit and relaunch Emacs and have it try again. This often works (though more than one try is sometimes required). The second, specifically for the auctex error, is to do M-x list-packages and then retry via quitting and relaunching Emacs. Sometimes the installation goes smoothly first time, sometimes this kind of messing around is required before Emacs finally decides auctex is in fact available and installs it. If these problems persist after a couple of tries, the best thing to do is simply to install the offending package or packages manually. Do M-x list-packages and in the resulting buffer search or scroll down the list to, e.g., auctex, mark it for installation by pressing i and then install it (or them) by hitting x. With the packages in place, restart Emacs and the starter kit will finish setting itself up.

Unfortunately I don’t think there’s anything else I can do to make these intermittent installation errors go away. If you use the kit, let me know if you run into any other problems with the installation.

Top Ten Ways that Hastings-on-Hudson might be a Village in a Wittgensteinian Sense

1. It is filled with very rich people affecting to be quite poor people.
2. It’s located in a Remote Part of Norway.
3. If a lion could live in this village, we would not be able to find it a decent duplex. Maybe a condo.
4. The HOAs are unbelievably picky about exterior paintwork, door design, and appropriate methods of kite-flying.
5. The configuration of the objects forms the atomic fact. In the atomic fact objects hang one in another, like the members of a chain. However, hanging laundry on chains at any time is absolutely forbidden.
6. Cutting-edge methods of elementary school instruction designed to enhance discipline, focus, and respect.
7. A property is internal if it is unthinkable that its object does not possess it, and is located inside the line demarcated on the relevant county plat map page.
8. Feeding the duckrabbits is forbidden by local ordinance.
9. Whereof we cannot speak, thereof we must be silent after 10pm except on public holidays.
10. Slightly distressing sense of family resemblance amongst everyone you meet.

Like many thousands of people, I had some contact with him via email, blogs, and twitter over the years. I gave him some data he asked for once. He contributed guest posts to Crooked Timber. He had a terrific grasp of sociology, by the way, and a far better facility with it than many people who think they have a “sociological imagination”. He pushed something my way once that I stupidly failed to take advantage of. I made him laugh a few times.

Unlike some people, I was not his friend, or confidant, or co-worker, or family. To them, I can only say I am sorry. He was so obviously a remarkable man, even at a distance. I wish the U.S. government had not hounded him on such an empty pretext. I wish he had found some other way to cope with his illness.

Bímid buan ar buairt gach ló
Ag caoi go cruaidh ‘s ag tuar na ndeor
Mar scaoileadh uainn an buachaill beo

I am sorry he is gone.

Those numbers come from the first chapter of Becky Pettit’s new book, Invisible Men: Mass Incarceration and the Myth of Black Progress. You can read the first chapter for free, but I recommend you buy the book. Pettit’s argument is that mass incarceration is such a large and intensive phenomenon that it distorts our understanding of many other social processes.

Pettit and others have been arguing for a long time that incarceration is by now a modal event in the life-course for young black men. Black men are more likely to go prison than complete college or serve in the military, and black, male, high-school dropouts are more likely to spend a year in prison than to get married. These social-structural changes have consequences for measuring and counting those involved. The incarcerated population is hard to count properly, and in many commonly-used data sources, like the Current Population Survey, its size and composition is poorly estimated or simply excluded. This has knock-on effects for our understanding of trends—especially changes in racial gaps—in things like rates of educational attainment, the employment-to-population ratio, group differences in average earnings, and voter turnout. In many of these cases, Pettit shows, what look like improvements over time are partly or mostly explained by miscounting due to the incarcerated population. This invisibility can happen both while people are in prison (as many surveys don’t adequately count the prison population) but also on release, where the marginal status of many ex-convicts makes them hard to reach in surveys built around sampling individuals stably attached to single households.

So, for example, Pettit shows that

Reliance on data from the Current Population Survey might lead one to believe that the high school dropout rate has fallen precipitously and that racial inequality has narrowed during the period of penal expansion … [CPS data] imply that the black-white gap in high-school completion, through either formal schooling or a GED, narrowed from 13.6 to 6.3 percentage poitns between 1980 amd 2008. Including inmates, we find little improvement in the black-white gap in high school for the last twenty years … including inmates suggests that the racial gap in high school completion among men has hovered close to its current level of 11 percentage points for most of the past twenty years” (55, 60).

Much the same seems to be true of group-level estimates of employment rates, wages, and voter turnout. Pettit argues in passing that the strong turn in the social science literature towards estimating causal effects (between education and wages, say) has led to a lack of attention to the quality of the purely descriptive numbers, and to a tendency to ignore the “acute sample bias associated with the exclusion of socially marginal groups from sample surveys” (107). This neglect feeds forward into that research, however, even if the research is not about inmates or criminal records at all, as it tends to bias estimates of the effects of, say, education on earnings. Moreover, Pettit argues,

… incarceration is so common in some sociodemographic groups that there are few comparable individuals in the population who have not experienced incarceration. The ubiquity of incarceration among low-skill black men undercuts research designs that require comparison groups to isolate the effects of incarceration from other factors like race, low education, or living in disadvantaged neighborhoods. A significant body of causally oriented research uses quasi-experimental designs … that compare the outomes of inmates or former inmates with similarly-situated indviduals who have not been to prison or jail. There is no valid comparison group for many of America’s inmates exactly because incarceration now inheres in whole sociodemographic groups ..

A broader theme of the book is that inmates and ex-inmates are at once constantly under surveillance and effectively unseen. Crime coverage is ubiquitous even as crime rates have been falling for years. The incarcerated and formerly-incarcerated population is enormous, but its real presence is invisible in many standard sources of data about the American population. The downstream consequences for our basic picture of what America is like and how it has been changing are underappreciated. In an era where “Big Data” is already an overused buzzword, Pettit’s book is a sobering reminder of the consequences of having a numerically large, socially consequential, but often statistically invisible population.

Yesterday I got an email from an editorial assistant at the Times:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

Hi Professor Healy,

We are publishing a column today that may 
reference the data you use here in your post here:

http://www.kieranhealy.org/blog/archives/2012/12/18/assault-death-rates-in-america-some-follow-up/

You mention that the OECD stats are gated — 
any chance you could share them with us, 
for fact checking purposes?

Thanks
Natalie

Natalie Kitroeff
Editorial Assistant
The New York Times

How would you interpret a request like that? I figured someone at the Times was going to refer to that post in passing, and the paper wanted some reassurance that I hadn’t just invented the numbers. I made the code for those charts publicly available, but the OECD data itself is buried in a gated database, which means I can’t publicly share it. So, I replied and attached a spreadsheet with the data, together with a description and a link to the OECD. I asked who was going to make use of it. She replied saying “Charles Blow, an opinion columnist here, may well include it in a chart he is compiling for tonight. Thanks very much!”

Today the New York Times has a column by Charles Blow presenting a table of various comparative, cross-national measures of violence. Here’s a clip:

The table’s “Assault deaths per 100,000 population” column uses the OECD data I sent them. It’s credited to the OECD, as it should be. But I’m slightly irritated. If you email me asking for data you can’t be bothered or don’t know how to dig up yourself, then say so. If that’s what you want, please don’t say you’re emailing me “for fact-checking purposes”. You’re not fact-checking something I did: you just want the data for yourself. In that case, the polite thing to do is acknowledge where you learned of its existence.

────────

Update: I got a follow-up email from Natalie Kitroeff at the Times. Here it is:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Hi Professor Healy, 

Thanks so much for your assistance with this yesterday. 
I just wanted to send a quick note to clear up a 
misunderstanding. After receiving the OECD data on assault 
deaths, we actually received the same dataset directly 
from economists at the OECD. We would have had absolutely 
no problem citing your blog as the source for the data, 
but our policy is to cite the source of the data directly 
when they provide it to us directly. 

Again, I really appreciate your help with this, 
and apologize for the misunderstanding. 

Best,
Natalie

I appreciate her following up. I have three things to say in reply. First, I don’t think this is particularly responsive to the problem of using the language of “fact-checking” to get hold of data. Second, Charles Blow opens his column by saying “Sometimes I think the best argument is raw data. This is one of those times.” It’s easier for it to be one of those times when someone has done the tedious work of ferreting out a specific time-series, drawn a picture of it that shows you why it’s so striking, and then sent you both the actual data and a reference for the hard-working economists at the OECD to follow up on. Third, as I said on Twitter earlier, this isn’t a big deal—in fact, it’s a little embarrassing to post about, because it feels like carping. But I do think that when a topic is as politicized as gun violence in the US, and when people are only too happy to discount data, or dismiss figures, or claim that research is biased, it’s worth being courteous about sources. It’s not hard to do. Consider how a hard-working economist at the New York Times dealt with it in similar circumstances back in July.

So, that’s that.

The original posts clearly explain what the data show and what the sources are. Nevertheless, people reshare pictures quickly, the specific topic of gun violence in the United States is highly politicized, and many people are heavily invested in it one way or the other. So when you present data that people glance at, under circumstances like a school shooting, their interpretation of it—and their desire to dismiss or pigeonhole it—is often colored by their own view of things. The result is that I get email, or people write blog posts, asking questions about the data’s source, its meaning, and my own motives or competence. The tone ranges from the polite and interested to the abusive or dismissive. People assert things about the provenance of the data, or its alleged selectivity. Sometimes they refer to other data or media reports that contradict the trends shown.

If you believe I am actively dishonest or desperately stupid there’s probably not much I can do to convince you otherwise. And, frankly, I am not inclined to try. More than a decade of blogging has taught me that there’s little point engaging with people who consistently assume you’re a liar or an idiot while simultaneously demanding that you carry out research tasks for them. I am not interested in having a pissing match with angry trolls. What I will do here is go through some questions that have come up in the past few days in connection with the figures I posted. Some of them were asked in good faith, others not. I’ll answer them as straightforwardly as I can.

Here’s the most widely circulated picture:

Click for a larger PNG or PDF.

1. Where do these data come from?
As discussed in the original posts, the data for the figure above are from the OECD Health Series. Data for the follow-up post comes from the OECD and the CDC WONDER mortality database. Access to the OECD statistics is gated. Try your local University library. You can query CDC WONDER yourself.

2. Why Assault Death Rates and not just Gun-Use Deaths?
The problem across the board here is getting consistent, reliable, cross-national data. The goal of the original post was to set the U.S. in some kind of longitudinal context with broadly comparable countries—in brief, the advanced industrial democracies—and the OECD is the best source for time-series data of this kind. But there is no good long-run cross-national data on deaths due to gun-use alone. Death as a result of violent assault is the best available series. There are some single-year cross-country datasets, and there are some smaller country-to-country comparisons, too. As a reminder, my original posts did not make any claims at all about guns, gun violence, or gun laws. I was interested in comparing rates of violent death across developed democracies, and that’s exactly what I did.

If you want to know why I was interested in doing this to begin with, it’s not because I have some axe to grind about gun control. It’s because one of the things I work on is cross-national variation in rates of organ donation. National differences in rates of certain kinds of sudden, violent, or accidental death (e.g. due to traffic accidents, gunshot wounds, or stroke) is an important predictor of variation in the availability of organs for donation.

3. Why these specific countries? Isn’t this just cherry-picking?
The purpose of the original posts was to put the U.S. rate of violent death in context. Which context is relevant? My view was that the U.S. was most usefully compared to other advanced industrial democracies. Of course the United States is not the most violent country in the world. But when it comes to questions of living standards, public safety, or social policy, Americans do not typically rush to compare themselves to Honduras, Kyrgystan, or South Africa. So the original comparison set was the following group: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Korea, Luxembourg, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom. Again, the original post showed the trends for these countries separately as well as in the combined figure. Others are welcome to make other comparisons.

4. But what about the OECD Member Countries you excluded?
The original post (and the note attached to the figures) made it clear I excluded two OECD countries with higher assault death rates than the US: Estonia and Mexico. I also excluded some other OECD countries whose rates are lower than the US, again on the grounds of useful comparison. These were Chile, the Czech Republic, Israel, Poland, Slovakia, and Slovenia. Here is what the figure looks like with all of these countries included:

Click for a larger PNG or PDF. The individual time series broken out by country are also available as a large PNG or PDF file.

As you can see, the US comes in third in overall in the OECD countries for which this data is available. In my view, Estonia and Mexico are not good comparison cases for the US. You may disagree. The countries whose average rates are below the US but come closest to it in some years are Chile and Israel. Consult the individual country panels to get a sense of when and and how close these countries are to the U.S. rates.

5. What about Turkey?
Turkey is in the OECD but there is no assault death rate information available for it in the OECD database.

6. I heard most gun deaths are suicides. Doesn’t that explain the U.S. case here?
No. It is true that most gun-related deaths in the U.S. are suicides. But the cause-of-death criteria used by the OECD to construct the assault series excludes suicides (or, to give it its official name, “intentional self harm”). So the gap here is not due to gun-related suicides.

7. You say America is violent, but it looks like the rate is dropping!
As I said in the original post, “the most striking features of the data are (1) how much more violent the U.S. is than other OECD countries (except possibly Estonia and Mexico, not shown here), and (2) the degree of change—and recently, decline—there has been in the U.S. time series considered by itself.”

8. What about rates of non-fatal assault? Can’t a country be violent (e.g. with lots of stabbings or fights) but have a low death rate from assault? I read in the Daily Mail that Britain was one of the most violent countries in the world.
You don’t have to think about the problem for long before realizing that getting good cross-national, comparable data on crime of any kind is difficult for many reasons, and that these problems are worse when the crime doesn’t lead to someone getting killed. Before you get down to some underlying rate of events, you have to consider, amongst other things, (1) Whether the act you have in mind is actually illegal in the country you’re interested in, (2) If it’s illegal, whether it is commonly considered a crime worth reporting, (3) If it’s seen as worth reporting, what’s the rate at which it is actually reported, and (4) If it’s reported, what is the the ability or inclination of the police to reliably record this data. The reason death rates are so often used in comparative settings is that there is a body, and it is harder—not impossible, but harder—for people to disagree about the existence of a dead body as opposed to, say, a fist-fight or stab wound. It is also harder—not impossible, but harder—to keep a fatal assault from being discovered, reported, and recorded as such. This is especially the case in developed democracies.

There are bona fide efforts to collect cross-national data on non-fatal violent crime and assault. The United Nations Office on Drugs and Crime tries to do this. You can look at their brief introduction to the difficulties of compiling and comparing international crime statistics and inspect their data for yourself. The time-series are patchy, and hard to interpret. The organization itself cautions that cross-national comparison and even within-country comparisons from year to year may be invalid due to variability in reporting rates, errors, and data definition.

To get a sense of the issues here take a look at the spreadsheet on non-fatal assault rates. If you took the data at face value, you would believe that the rate of non-fatal assault in Scotland is thirty times higher than it is in Uganda, that between 2006 and 2007 Poland’s assault rate went from 76 per 100k to 1.4 per 100k, that Sweden has 3.3 times the rate of violent assault as the United States, and that El Salvador—a country with the world’s second-highest murder rate, according to other figures from UNDOC—nevertheless had approximately the same rate of violent assault in 2010 as Norway.

Now, perhaps all of those statements are true. It’s possible to make up a story in each case about how these could all be real differences, rather than artifacts due to variation in state capacity to police, bureaucratic reliability in reporting, and local differences in the definition of reportable assault. But I would be extremely cautious about using these numbers to rebut the idea that, for an economically developed democracy, America is unusually violent.

8. But isn’t there a lot of variation inside the United States?
Yes, of course. That’s why I wrote a post called Assault Deaths Within the United States as a follow-up, looking at variation by State, Census Region, and race of victim. To keep things comparable with the cross-national data I used the Assault Series from the CDC database. However, if you want to look specifically at gun violence that data is also available from the CDC, and in somewhat different form—allowing for other sorts of comparisons—from the FBI’s crime statistics.

9. You don’t even do research on violence.
That’s right, I don’t. For that reason, despite various requests, I don’t give interviews or quotes to the media on this topic. Sometimes I informally write about and analyze data on my blog. This was one of those times. I stand by the data analysis, presentation, and discussion in the original posts.

10. Can you make some new graphs for me?
If you want to see exactly what I did with the data, the R code is on github. If you want to do your own analysis, go ahead. I remind you that I am not your RA.

Savita Halappanavar (31), a dentist, presented with back pain at the hospital on October 21st, was found to be miscarrying, and died of septicaemia a week later. Her husband, Praveen Halappanavar (34), an engineer at Boston Scientific in Galway, says she asked several times over a three-day period that the pregnancy be terminated. He says that, having been told she was miscarrying, and after one day in severe pain, Ms Halappanavar asked for a medical termination. This was refused, he says, because the foetal heartbeat was still present and they were told, “this is a Catholic country”. She spent a further 2½ days “in agony” until the foetal heartbeat stopped. The dead foetus was removed and Savita was taken to the high dependency unit and then the intensive care unit, where she died of septicaemia on the 28th.

A statement from Galway Pro-Choice notes, in part:

Under the X Case ruling, women in Ireland are legally entitled to an abortion when it is necessary to save their life. However, legislation has never been passed to reflect this. It is the failure of successive governments to do so that led to Savita’s death. … Rachel Donnelly, Galway Pro-Choice spokesperson stated: “This was an obstetric emergency which should have been dealt with in a routine manner. Yet Irish doctors are restrained from making obvious medical decisions by a fear of potentially severe consequences. As the European Court of Human Rights ruled, as long as the 1861 Act remains in place, alongside a complete political unwillingness to touch the issue, pregnant women will continue to be unsafe in this country.”

In 1992 I was in my second year of college, at UCC. Beginning early in that year a string of social and political crises and scandals broke that, in retrospect, marked the beginning of the end of the public power of the Catholic Church in Ireland, especially with respect to sexuality. I’ve written before about how the events of that year are a kind of bookend to the autumn of 1979, when the Pope came to Ireland and the country was filled with a revivalist fervor focused on the country’s youth. Chief amongst the social crises of 1992 was the X-case. A fourteen year-old girl, pregnant as the result of rape by a neighbor, sought to leave the country to have an abortion in England, as thousands of Irish women did and still do. Her parents asked the police whether it would be possible to collect any DNA evidence during this process, which brought the matter to the Attorney General’s attention. He sought, and was granted, a court injunction preventing her from traveling for the abortion under Article 40.3.3 of the Irish constitution, which had been passed in 1983 as the “Pro-Life amendment” and which prohibited abortion in Ireland. The resulting constitutional crisis saw the Supreme Court issue a hasty ruling permitting the girl to travel, but under a strained interpretation of the law. So in November of 1992 the Government introduced three new amendments meant to clarify things. The result was that things reverted much to the status quo ante, which essentially allowed Ireland to export its abortion problem to the United Kingdom while leaving it unclear what doctors facing a medical emergency during pregnancy were in a position to do.

And so things have stood, more or less. When I followed the X-case in 1992, and when I wrote an MA Thesis about it in 1994, I honestly thought that some minimal legal provision for abortion in Ireland would be fairly soon in coming. But that was twenty years ago.

One piece that always gives me a headache, though, is getting Minion Pro set up for use with LaTeX. Minon is a terrific typeface and I use it for my papers. It’s expensive to buy but often comes bundled with various Adobe products, notably Acrobat reader. If you have the font installed on your Mac somewhere, then there’s a package of stuff available to get it to work with pdflatex. But it’s a pain to install. Note at this point that, if you like, you can simply use xelatex instead to use all your installed fonts with latex. But because I have a debilitating obsessiveness in this regard, I know that xelatex doesn’t let you use certain microtype features available to pdflatex. So I need to set up Minion Pro to work with pdflatex. Here is a shell script I came across that will do that for you automatically. Don’t run it unless you know what you’re doing, though. Seeing as I have to go through this any time I set up a new Mac, and always forget the steps, this is pretty useful.

If you are a sensible person, of course, you will not even find yourself facing this ridiculous situation.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101

#!/bin/sh
## Information
## http://carlo-hamalainen.net/blog/2007/12/11/installing-minion-pro-fonts/
## http://www.ctan.org/tex-archive/fonts/mnsymbol/

## 0.1: Install LCDF Typetools
## http://www.lcdf.org/type/
## If you use Homebrew (http://mxcl.github.com/homebrew/), then uncomment: 
# brew install lcdf-typetools 

## 0.2: If ~/tmp doesn't exist, create it.
# mkdir ~/tmp

## Destination. System wide:  
# DEST=`kpsexpand '$TEXMFLOCAL'`
## Or single-user only:
DEST=~/Library/texmf

## Downloader:
DOWNLOAD="curl -L -O"

## Directory where minion fonts can be found:
#MINIONSRC=/Applications/Adobe\ Reader.app/Contents/Resources/Resource/Font/
#MINIONSRC=~/tmp/minionsrc
MINIONSRC=~/Library/Fonts

## Everything gets done in a temporary directory
cd ~/tmp

## 1: MnSymbol
## http://www.ctan.org/tex-archive/fonts/mnsymbol/
$DOWNLOAD http://mirror.ctan.org/fonts/mnsymbol.zip

unzip mnsymbol
cd mnsymbol/tex

## Generates MnSymbol.sty
latex MnSymbol.ins

mkdir -p $DEST/tex/latex/MnSymbol/      \
    $DEST/fonts/source/public/MnSymbol/ \
    $DEST/doc/latex/MnSymbol/

cp MnSymbol.sty $DEST/tex/latex/MnSymbol/MnSymbol.sty
cd .. # we were in mnsymbol/tex
cp source/* $DEST/fonts/source/public/MnSymbol/
cp MnSymbol.pdf README $DEST/doc/latex/MnSymbol/

mkdir -p $DEST/fonts/map/dvips/MnSymbol \
    $DEST/fonts/enc/dvips/MnSymbol      \
    $DEST/fonts/type1/public/MnSymbol   \
    $DEST/fonts/tfm/public/MnSymbol
cp enc/MnSymbol.map $DEST/fonts/map/dvips/MnSymbol/
cp enc/*.enc $DEST/fonts/enc/dvips/MnSymbol/
cp pfb/*.pfb $DEST/fonts/type1/public/MnSymbol/
cp tfm/* $DEST/fonts/tfm/public/MnSymbol/

## I believe that this is not strictly needed if DEST is in the home
## tree on OSX, but might be needed otherwise
sudo mktexlsr
updmap --enable MixedMap MnSymbol.map

# $DOWNLOAD http://carlo-hamalainen.net/blog/myfiles/minionpro/mnsymbol-test.tex
# pdflatex mnsymbol-test.tex

## 2: MinionPro
mkdir -p ~/tmp/minionpro
cd ~/tmp/minionpro

$DOWNLOAD http://mirrors.ctan.org/fonts/minionpro/enc-2.000.zip
$DOWNLOAD http://mirrors.ctan.org/fonts/minionpro/metrics-base.zip
$DOWNLOAD http://mirrors.ctan.org/fonts/minionpro/metrics-full.zip
$DOWNLOAD http://mirrors.ctan.org/fonts/minionpro/scripts.zip

## This will make the otf directory, among other things.
unzip scripts.zip

cp $MINIONSRC/Minion*otf otf/

## Generate the pfb files
## This step requires that the LCDF type tools are installed.  Get them here:
##   http://www.lcdf.org/type/
./convert.sh

## Copy the pfb files to where they belong:
mkdir -p $DEST/fonts/type1/adobe/MinionPro
cp pfb/*.pfb $DEST/fonts/type1/adobe/MinionPro

SRC=`pwd`
cd $DEST
unzip $SRC/enc-2.000.zip
unzip $SRC/metrics-base.zip
unzip $SRC/metrics-full.zip
cd $SRC

sudo mktexlsr
updmap --enable MixedMap MinionPro.map

## Test:
# $DOWNLOAD http://carlo-hamalainen.net/blog/myfiles/minionpro/minionpro-test.tex
# pdflatex minionpro-test.tex

The U.S. system of employer-sponsored healthcare provision is iTunes. It’s complicated and overburdened; it wasn’t originally designed to do most of the things it now does; in fact, at the outset its design wasn’t really thought through at all (there wasn’t time); many of those involved backed it as a distant second-best solution—better than nothing, but not nearly good enough. Over the years, new features were shoehorned into the basic structure. New problems and inconsistencies emerged and were partially patched. And, inevitably, groups who did pretty well out of the system emerged and entrenched themselves, too. In situations like this, some reforms are possible around the edges, but it’s clear to most people that real structural reform is needed. The people in charge of it come to face a dilemma familiar to both software developers and policy makers. In the words of a Swiss official quoted in an old study (by Ellen Immergut) of the politics of health care in Europe, “Were it necessary to draft a health insurance bill today, I would never come up with the insane idea of proposing our current system.”

As every developer knows, ripping out an established codebase and replacing it with something better is far easier said than done, especially when it’s huge, and a lot of people are doing quite well from it. Perhaps worst of all—and maybe most similar of all, too—the problem has three jointly unpleasant characteristics: it’s really important for the future; a great deal of political fighting is needed to achieve any real reform; and the details of the required changes are eye-glazingly dull, except perhaps to a small number of odd people and parties with a stake in seeing the wrong thing happen.

It’s best not to push this analogy any further. You can already hear it creaking. Sociologists and political scientists who study institutions and policy discuss these issues under the general rubric of ”path dependence”. Engineers have their own more or less formalized lore (in books like The Mythical Man Month and concepts like “Second System Syndrome”) that touch on some of the same issues. In the end—some common nerd fantasies notwithstanding—politics is not simply a matter of engineering. Indeed, the lesson taught by everyone from Fred Brooks to Michael Lopp is that—some common nerd fantasies notwithstanding—often even engineering isn’t simply a matter of engineering. But many software developers know what it’s like to inherit a huge, messy, live codebase that does something very important in a really terrible way. And that’s why the U.S. employer-sponsored healthcare system is iTunes.

Note: If you want to learn more about the guts of health care reform, I recommend two books by one of my teachers, Paul Starr. The first his is magisterial The Social Transformation of American Medicine, which takes the story up to the late 1970s. It is the definitive account how America reached the point where its health care system was in structural crisis. His recent Remedy and Reaction is shorter (though it still provides a good historical overview, including the failed Clinton-era reform which Starr himself was involved in) and concentrates on where we are now.

I’m teaching Weber next week in my social theory class. This afternoon I uploaded some of the recommended reading to the class website—a longish excerpt from the first volume of Michael Mann’s The Sources of Social Power, a book which made a big impression on me when I read it as an undergraduate. It didn’t turn me in to a Big Structures/Huge Comparisons guy, if for no other reason than the ambition it entailed seemed so gigantic, but of all the products of the so-called “Golden Age” of macro-sociology, Sources—and the first volume in particular—seemed to come closest to fulfilling Weber’s vision for what really big-picture sociology could be. Re-reading the first hundred-odd pages this afternoon I was struck by the directness and accessibility of Mann’s approach, and by how much of his theoretical intuition seemed right, given his aims—in particular his insistence that societies are not totalities or systems, and his determination to avoid the pitfalls that come with thinking they are:

Societies are not unitary. They are not social systems (closed or open); they are not totalities. We can never find a single bounded society in geographical or social space. Because there is no system, no totality, there cannot be “subsystems,” “dimensions,” or “levels” of such a totality. Because there is no whole, social relations cannot be reduced “ultimately, “in the last instance,” to some systemic property of it—like the “mode of material production,” or the “cultural” or “normative system,” or the “form of military organization.” Because there is no bounded totality, it is not helpful to divide social change or conflict into “endogenous” and “exogenous” varieties. Because there is no social system, there is no “evolutionary” process within it. Because humanity is not divided into a series of bounded totalities, “diffusion” of social organization does not occur between them. Because there is no totality, individuals are not constrained in their behavior by “social structure as a whole,” and so it is not helpful to make a disctinction between “social action” and “social structure.” … State, culture, and economy are all important structuring networks, but they almost never coincide. There is no one master concept or basic unit of “society”.

Instead, for Mann, what matters are the overlapping networks of social interaction—ideological, military, economic, and political—that can provide the organizational means of attaining goals.

All of which is to say that, after a bit of idle googling, I was surprised to learn that volumes three and four are scheduled for publication later this year and early next, respectively. Mann published volume I in 1986 and volume II in 1993, and while he has done a lot of other things in the meantime, parts of volume II, in particular, gave the distinct impression that the project had gotten seriously bogged down. I’m very glad to see that he’s pushed the project through. The first two volumes are also set to be reissued, with new Prefaces (and covers). I suppose it is too much to ask that they have proper indexes this time, too.