About That POS Rally Table

Not all foreign policy episodes produce a “rally round the flag effect,” and most of those effects are far smaller than one would believe from a Public Opinion Strategies table that everyone was linking to yesterday.

POS posted the table, which lists 13 such bumps in presidential approval ratings, under the comment: “Not including the post-9/11 response, the average presidential approval rating bumps up 13% for an average of 22 weeks.” However, the table is highly selective in choosing these events. In fact, POS has basically selected the largest such rallies in Gallup history, and omits those events which produced smaller rallies – as well as those that produced no rally at all.

The key study of these events is an old article by political scientist Richard Brody, who listed 65 potential rally events from the announcement of the Truman Doctrine in March 1947 through the Iran/Contra Affair in November 1986. The average approval gain? Under 3 points.

There doesn’t appear to be any particular reason for the inclusion of some events and the exclusion of others in the POS table. For example, POS includes the beginning of the Korea War, which produced a 9 point surge for Harry Truman. A good example of a rally! But they omit the Berlin blockade of April 1948, which produced only a 3 point increase. The Ford-era Mayaguez incident (+11) is in, but not the similar Johnson-era Pueblo incident (-7). The Cuban Missile Crisis (+12), but not the Bay of Pigs (+5). By the way, the Pueblo incident was hardly the only negative one – just during the Reagan presidency there are three separate Libya incidents (in August 1981 and February and March 1986) which were all associated with small declines in the president’s approval rating.

Now, Brody’s article predates the major rallies under George H.W. Bush (Gulf War) and George W. Bush (9/11, Iraq War, capturing Saddam)…but presumably it also misses some smaller events under those presidents and Bill Clinton.

The POS table also gives the duration of the rally effect, and I agree with those who say that the duration of the effect is probably as important or more than the height of the spike. I’d be very cautious about their data on these, too, however. While Brody doesn’t give data for the duration of his larger group of cases, presumably including all of the small bounces would cut the average duration dramatically. Even without that, however, I’d urge people not to take the POS table at face value. Gallup only recently began their daily tracking of the president’s approval ratings, and in many of the earlier episodes POS includes, polling was highly irregular. For example, in 1962 Gallup was polling only once a month or so, on no (apparent) regular schedule; it’s very possible that the 40 weeks it took for JFK’s approval to fall back to pre-Cuba levels would have been much shorter if we had daily tracking back then. That, of course, is not POS’s fault, but it’s worth keeping in mind. It’s also worth remembering that after a few days, or at best a few weeks, most rally events are probably not responsible for elevated presidential approval numbers, and it’s very possible that some other intervening event is going on (indeed, when dealing with only a handful of episodes, all sorts of distortions are possible).

All that said, my guess was that the killing of bin Laden would produce a very healthy bounce. We’ll see. The immediate results are well below my 15 point guess (Nate Silver has one roundup, Mark Blumenthal has another). As I’ve written before, Brody found that the key variable is the reaction of opposition party politicians; if they praise the president, the rally happens, and if criticize him, no rally. Either way – the 13 point POS estimate is far, far too high for an average rally event.