Graphical anomaly in Ziobrowski et al (2004)

Congressional stock trading is back in the news, so I recently had occasion to look back into the issue including my work with Jens Hainmueller about 10 years ago.

It reminded me of an intriguing anomaly in a 2004 paper by Alan Ziobrowski and coauthors (link). This paper was the first to claim that members of Congress enjoyed a systematic investing advantage. It looks at common stock transactions by members of the Senate in the period 1993-1998.

The first and most striking evidence of Congressional trading acumen in Ziobrowski et al (2004) is the figure below (Figure 1). The dark line shows the performance of a portfolio of stocks Senators sold, where the vertical line at zero indicates the day when they sold. In the year before the sale, these stocks beat the market by a whopping 25%; after the sale, they basically matched the market. Conversely, the stocks Senators bought were flat before the purchase but then beat the market by about 25% in the following year. This is stunning trading acumen.

But there is something odd about this figure. It is supposed to show cumulative abnormal returns, meaning the amount by which each portfolio outperforms the market in the year before and after the transaction (i.e. +/- 255 trading days). So it shouldn’t be surprising that the Buys series and the Sells series both start at 0. But it is surprising that they both end precisely at 28.6%. There is no reason to expect that.

On closer inspection, the Sells series looks like it is the same as the Buys series, just flipped and reversed. To see this, I rotated the figure by 180 degrees, colored it orange, and made it semi-transparent; then I overlaid it on the original, which I colored blue for contrast. Below is the result. You can see that the two series line up exactly.

Jens and I noticed this anomaly back when we were working on our papers (around 2010), but we didn’t do anything about it. Ziobrowski had already refused to share his data with us or with other researchers, so we couldn’t check what happened.

Jens and I ultimately published a paper arguing that the evidence of Congressional investing acumen in Ziobrowski et al (2004) and their 2011 follow-up (focusing on the House) is weaker than the authors claim, in part because they emphasize the strongest results among various possible specifications. I still think that’s a convincing critique of these papers.

I think it also would have been appropriate to highlight this anomaly, though at the time I was too timid to get into it. It’s hard to tell whether this was just an error in the presentation of the cumulative abnormal returns or something that affects the calendar-time portfolio analysis that is reported in the subsequent tables. Even if it was just this figure, I think it’s worth noticing because this figure is the most striking piece of evidence in the paper.