- EphBlog - http://ephblog.com -


The Equality of Opportunity Project gathers amazing data on the incomes of college graduates and their families. The New York Times provides an overview and this summary (pdf) of Williams data. Day 3 of 5 on this topic.

The central point about socio-economic diversity that I have been making for more than a decade is that there is no evidence that Williams is more economically diverse now than it was 30 years ago, and probably even 50 or 100 years ago. It is embarrassing how often the Williams administration (names like Payne, Schapiro, Hill, Falk and Dudley come to mind) claim that we are more economically diverse and how quickly naive reporters like David Leonhardt of The New York Times are to believe them. Recall the question that I have suggested for years:

In 1998, the 426th poorest family at Williams had a family income of $63,791. What is the family income of the 426th poorest family at Williams today? How has that number changed over the last two decades?

If the Record were a competent paper, or David Leonhardt were a competent reporter, than this is the question that would be asked. It/he isn’t, and so we have been left with just my rants. But now we have data!


Summary: Williams did not become (meaningfully) more economically diverse between the classes of 2005 and 2013. Eyeballing the chart, it looks like about 19% of the students in the class of 2006 were from families in the bottom 60% of the income distribution. For the class of 2013, it was 20%. Surprised? You shouldn’t be. Recall my analysis from 2008:

We can see that there is no evidence that the socio-economic diversity of Williams has increased in the last decade and some circumstantial evidence that it has stayed the same.

The EOP proves that I was right. There was no good evidence that economic diversity had meaningfully increased at Williams between 1998 and 2008. The EOP data, which goes through the class of 2013, shows the same thing.

More importantly, we know that the same trend has continued up through the class of 2021, as we discussed on Monday. In fact, this sure seems similar to the data we know for the class of 1998.


1) The above chart is drawn from this collection, which shows the trends for various cuts of the income distribution. There is no perfect single measure of income inequality. Other charts, like that for the percentage of students from the top 20%, might put Williams in a better light. But even these charts, to the extent that they show changes in the direction of more economic diversity, show incredibly small changes, perhaps even within the appropriate confidence intervals.

2) We are being fast and loose with many of the relevant details. The numbers we studied in 2008 were based on all the students at Williams over the years between 1998 and 2008. In other words, each number was provided for all 2,000 students on campus in a given academic year. The EOP data is, I think, based on birth year, which provides, at best, an imperfect mapping to graduation class.

3) We should try to get our hands on the underlying data for Williams and some other peer schools. Any volunteers? Any readers with connections to Chetty et al?

4) Any predictions as to whether or not US News will use this data in its next set of rankings? Should it?

Comments Disabled (Open | Close)

Comments Disabled To "EOP III"

#1 Comment By Williams Alum On January 25, 2017 @ 8:46 am


Can a statistician confirm that confidence intervals don’t make sense when talking about census data?

I.e. suppose class of 2020 had median income of $70k. Class of 2021 had median income of $71k. You can’t say this is within the confidence interval and didn’t statistically change, if those numbers are a census of all families. The number definitively did change.

So I think that part in your piece makes no sense, presuming that the data set is census based and not survey based (which I am not sure because I am too tired to click through to the data set).


#2 Comment By David Dudley Field ’25 On January 25, 2017 @ 9:27 am

The data is survey-based so your point is well-taken. But, as with all surveys, there are not-matched records, thereby creating some uncertainty. There is also uncertainty arising from the lack of international students (this analysis uses US tax records only) and changing numbers of international students. Recall that, over time, we have somewhat more international students and that they are richer than they used to be.

All that said, the key is not so much the confidence intervals, whatever they are, but the raw magnitude of the change. If, over almost 25 years, the portion of Williams students from families in the bottom 60% has fluctuated within a band of 17%-20% (which seems to be the case), then, for me, this means that there has been no meaningful increase in economic diversity.