23 August 2008
In Britain (and many other democracies) Members of Parliament often have "Outside Interests" and draw extra income from paid directorships, consulting gigs, or journalistic work. There has been plenty of controversy about MPs having such outside interests ranging from possible ethical issues to concerns over possible impact on MPs' legislative behavior. One concern is that MPs may be less effective as a representative if they "moonlight" from their Westminster jobs.
In a recent paper about the financial returns to serving in the House of Commons, Andy Eggers and I consider the relationship between outside interests and MPs' vote attendance (the percent of eligible votes personally attended or told) for the 2005-2007 period. We find that for both the Conservative and the Labour party, MPs with at least one (self-)reported outside interest (directorships, consultancies, and work in journalism) attended fewer votes compared to MPs with no outside interests; attendance rates are around 4-6 percentage points lower and the differences are all significant at conventional levels. The results are summarized in the jittergrams below (we excluded MPs that hold office as minister, speaker, whips, and chairman of standing committees who are not allowed to vote).
The exception are directorships for Labour MPs where we cannot reject the null
of no difference. We also find no such difference for MPs when comparing those with and
without regular employment (such as work as a barrister, medical doctor, etc.).
There are obviously several other factors that may contribute to low attendance rates such as absence on constituency business, illness, paternity/maternity leaves, etc., but overall the results do suggest that outside interests distract MPs from their legislative work. This finding is also consistent with earlier work by Muller (1977), who found that sponsored Labour MPs were more active than other MPs on issues close to the interests of their sponsors (e.g. mining or railway issues), but that on the whole they were less active members of Parliament, participating in question time, standing committees, and debates far less than non-sponsored members.
If this whetted your curiosity you can analyze the voting data for yourself at publicwhip.
21 August 2008
For those who have collected research data and made it available to others, its nice when people thank you. But it would be nicer to receive formal scholarly citation credit and web visibility for your hard work. The Dataverse Network project is designed to get you that credit and visibility.
The idea is to give you a free "dataverse" (your view of the universe of data) -- which is a virtual archive where you can store, permanently preserve, and distribute your data (or list data from other dataverses) with everyone or only those you approve. Your dataverse is branded as yours, with the look and feel of your web site and on your web site, but since it is served out by an installation of the Dataverse Network at Harvard you needn't install any software or hardware. Some other features include:
12 August 2008
The BBC reports on new findings published in in Archives of Internal Medicine from a longitudinal study of running and health. In 1984, the authors recruited members of a national club of over-50 runners and a control group of similar non-runners. Based on health outcomes observed in the years since, the authors conclude that "Vigorous exercise (running) at middle and older ages is associated with reduced disability in later life and a notable survival advantage." And -- more good news for runners -- the runners did not disproportionately suffer from knee and ankle injuries.
With studies like this, of course, I approach the paper wondering how much of the observed difference is due to the "treatment" (running) and how much is due to other differences between the treatment and control groups. Here I think that unmeasured confounding is significant; I would guess that selection accounts for maybe half of the difference in outcomes between the runners and non-runners. People who are in a running club at age 55 are unusual for many reasons that are not easily observed and controlled for. For one thing, they've usually run for a long time, meaning that the study really compares lifelong exercise regimens and not just whether the subjects run late in life. More importantly, older runners tend to have unusually fortunate genetic inheritances; anyone who didn't would be unable or unwilling to keep up a running regimen at that age. In my experience dedicated older runners also tend to be people with a fierce determination to conquer challenges and stick to a regimen. These are people who eat well and see their doctor and have friends and have all sorts of other health advantages. It's not surprising that people with such genetic and lifestyle advantages live longer, but attributing it to their running -- ie suggesting that anyone could start running and have similar outcomes -- would be to overlook a lot of these confounding factors.
The authors of the study recognize some of the selection problems they face, and they do what they can to address them, conditional on the study design. I was reminded in reading the paper that public health and medical researchers tend to be considerably more diligent than researchers in other social sciences in choosing words that differentiate association and causation. These distinctions tend to be lost in media coverage of the research but in this case the authors do a careful job of recognizing the obstacles to drawing inferences about the effects of running based on their study.
5 August 2008
In Born on the First of July: An (Un)natural Experiment in Birth Timing, forthcoming at the Journal of Public Economics, Joshua Gans and Andrew Leigh examine "introduction effects" (the extent to which people change their behavior to respond to new policies) in the context of a baby bonus that was initiated in Australia in 2004. In May of that year, the government announced that families of babies born on or after July 1 would receive a $3000 cash bonus. Mothers with due dates around that time made special arrangements (mostly delaying Caesarean and other planned deliveries) to get the prize. The authors estimate that over 1000 births were moved; July 1, 2004, witnessed more births than any other day in the period since 1975 for which the authors have data.
The authors note two implications of the study. First, policies can provoke not only long-run distortions (e.g. increases in babies born) but short-run distortions from gaming of the system. Second, the "baby bump" constituted a large disruption in regular procedures in maternity hospitals and staff; they don't find effects on infant mortality, but they suggest that the event could be useful for studying the effects of under-staffing in hospitals.
My first thought in reading the paper was that it was a cautionary tale for regression discontinuity design. The setup of the study has the flavor of David Card et al's paper "Does Medicare Save Lives?," discussed on this blog by the intrepid John Graves, in which the authors examine the outcomes of patients who need medical procedures right around the time when they become eligible to receive Medicare benefits; they find that patients who were barely old enough to receive the benefits do considerably better than the ones who were too young. I figured this study was probably a failed attempt to do something similar, ie to study the effect of extra income on child mortality or other outcomes by comparing kids born just before and after the benefit was handed out. This sort of thing doesn't work when the subjects are able to sort around the threshold. In this case, the parents who gave birth just after the cutoff may have been more desperate for money, or had more power with the doctors, or perhaps were generally more in tune with political events, such that differences in outcomes between recipients and non-recipients of the bonus could be due to these factors and not the bonus itself. In Card et al's case, they focused on emergency procedures that could not have been delayed; this study shows that for many people childbirth is quite postponable. So in addition to the implications Gans and Leigh draw in their conclusion, I would add that this is another case where an RDD-style approach is complicated because subjects can effectively sort.
I do think you could examine the effect of the bonus on child outcomes if you looked at kids born at least 3 weeks before and after the cutoff date, a point at which the sorting is probably not that big of a deal. And date of birth is probably not itself a strong confounder for whatever you want to study, so there are limited advantages to focusing in on the threshold anyway.
At any rate, it appears my initial impression -- that the paper is the artifact of a failed RDD project -- was wrong: the authors have done other examinations of how events affect birth patterns (the effect of the millennium on conceptions, births and deaths, and the ability of parents to move births from inauspicious days like Feb 29 and April 1).