John Mueller's War, Presidents and Public Opinion

Warning: This post is pretty long, even for me. It also talks about polls and regressions and logarithms, oh my! What I'm arguing here is that one of the academic studies that is used to justify a very heavy reliance on air power in American wars doesn't hold up very well to historical scrutiny.

The concept that the American public is particularly "casualty averse" in wars has a strong hold on military planners and politicians. And at a basic level, that's a good thing. Generals and Presidents and members of Congress should be very hesitant to send American soldiers to war, so that they would only do it when there's a very good reason. Most people do have the good sense to know that war is a very bad thing.

But our Dr. Strangeloves have turned that healthy instinct on its head. In PentagonThink, Americans will only object to war if there are a lot of American casualties. The way you minimize American casualties is to rely heavily on long-range firepower and, especially, air power. Even though doing that may wind up blowing civilians away by the tens of thousands in the enemy country, thereby severely undermining the prospects for a quick resolution of a war like the one in Iraq.

But, the thinking goes, this is "minimizing casualties", i.e., American casualties, the only ones that supposedly count. That then ensures broad public support for the war because people only oppose war when there are American casualties. And since the American military is invincible, the "center of gravity", i.e., the only weakness, of the American military is the loss of public support. So, use heavy airpower and artillery, you minimize American casualties, the American public supports the war and our invincible generals can't avoid winning.

That way, you can have more wars because the public will support them. So you see? The fact that normal people dislike war becomes a good reason to have more wars!

With thinking like this, is it any wonder that our infallible generals lost the Iraq War?

John Mueller's 1973 study of War, Presidents, and Public Opinion has become a frequently-cited ideological justification for this peace-is-war goofball theory. Which is ironic in a way, because Mueller's book is on the surface a careful political scientist's analysis of hard quantitative polling data, a social scientist's wet dream.

The problem with quantitative data, though, is that it can be seductively misleading. One of the two things I remember most from a forecasting class I once took was arguing with the professor about a statistical regression model for Presidential elections that relied heavily on economic statistics for its predictions that proved conclusively that the Vietnam War had nothing whatsoever to do with the outcome of the 1968 Presidential election.

The other thing I most remember from his class was his first rule of forecasting: if a forecast looks wrong, it probably is. In this case, the retroactive forecast was a test of the model to make sure it was using accurate predictive assumptions. But anyone with even the most superficial knowledge of the 1968 election could see the problem with that result. There was also extensive public opinion (marketing) data that measured what issues were important to people, which would have provided quantitative disconfirmation.

The other problem with the result was that it showed clearly that Hubert Humphrey was elected President in 1968. You might think that even an economist would see the problem with that result. But the numbers don't lie, right?

Mueller's results suffer in part from the same problem of letting the false concreteness of polling numbers cloud the results.

Mueller's focus was comparing the public opinion polls on the Korean War and the Vietnam War. His conclusion that is grist for the mill of the war planners who want to minimize public opposition to war by maximizing the death and destruction on the Other Side is the one that says that public opposition to wars is driven by the rise in American casualties, period. He restated that result more recently in a much-cited article in Foreign Affairs, The Iraq Syndrome Nov/Dec 2005:

American troops have been sent into harm's way many times since 1945, but in only three cases - Korea, Vietnam, and Iraq - have they been drawn into sustained ground combat and suffered more than 300 deaths in action. American public opinion became a key factor in all three wars, and in each one there has been a simple association: as casualties mount, support decreases. Broad enthusiasm at the outset invariably erodes.

My basic reaction to this is, hmmm, let's see. As a war goes on month after month and year after year and casualties mount, it becomes less popular. Well, duh!

He then continues:

The only thing remarkable about the current war in Iraq is how precipitously American public support has dropped off. Casualty for casualty, support has declined far more quickly than it did during either the Korean War or the Vietnam War. And if history is any indication, there is little the Bush administration can do to reverse this decline.

Now, support has dropped off for the Iraq War more quickly than for the Korean and Vietnam Wars, even though casualties are not perceived as being so severe. That makes me think that the American public may be applying other standards of judgment than simply counting American casualties. To me, that's also a "Well, duh!" observation.

But Mueller finds that result "remarkable". I see two basic possibilities for this. One is that his findings and the resulting generalization derived from it, that mounting American casualties and only that make people turn against a war, are so strongly established that the Iraq War must have something about it that make it a radical exception. The other possibility is that the simple-minded conclusion that American casualties and only American casualties drive public opinion about war is somehow deficient. I'm betting on the latter.

Here's the nitty-gritty political science version from the 1973 book, with the key table thrown in as a bonus. Don't feel bad if you feel a blurry feeling in your brain as you look at this; it's perfectly natural.

These concerns can be incorporated into the analysis by seeking to relate popular support for the wars to the logarithm of the total number of American casualties that had been suffered at the time of the poll. That is, one assumes that the public is sensitive to relatively small losses at the start of the war but only to rather large ones toward its end. Specifically, one does not expect casualties to affect attitudes in a linear manner with a rise from 100 to 1000 being the same as one from 10,000 to 10,900. Rather, a rise from 100 to 1000 is taken as the same as one from 10,000 to 100,000. Thus the distance between the numbers, 10, 100, 1000, 10,000, 100,000 and 1,000,000 is made equal.

This sort of a transformation is applied for Table 3.4 and the result is a set of equations suggesting strikingly similar drops in popular support for the two wars: in each war, support is projected to have started at much the same level (the intercept figure), and then every time American casualties increased by a factor of 10, support for the war dropped by about 15 percentage points. Results are not so neat when the dependent variable is the percentage opposing the war, but the patterns remain largely similar: opposition to Vietnam is taken to have begun at a somewhat lower level than in Korea and then to have increased at a somewhat faster rate. Part of this difference is due to the generally lower frequencies of no opinion responses during the Vietnam period. (italics in original)

In other words, if you just look at the trends against casualties, the relationship between opposition to the war is not obvious when you compare the Korean and Vietnam Wars. But when you run the logarithm, it shows a pretty similar pattern that as casualties mount so does opposition to the war.

And most readers who don't spend their spare hours pouring through polling statistics and running their own regression analyses will be very tempted to say after reading that, "Oh, the logarithm says that's the way it is. Well, okay then! Now, what's the latest on Anna Nicole Smith's baby?"

I didn't redo Mueller's statistical analysis. But it's fairly obvious without running the numbers that mounting casualties would also correlate to length of the war and therefore other factors perhaps not so easily quantifiable. Which should be a big red flag in itself.

Then there's also the fact which Mueller found so wondrous in his 2005 article that the Iraq War shows a very different result. And the most likely explanation of that to me is that his explanation in the 1973 book was wrong. Or, at best, oversimplified. Because most people don't need a statistician to recognize that the number of casualties in a war are an important factor in how people look at war and almost certainly have been since, say, prehistoric times.

An obvious comparison would be the Second World War, which involved a far greater loss in American lives than either the Korean or Vietnam Wars. And, in fact, Mueller makes a brief comparison in the 1973 book:

To put the poll statistics from Korea and Vietnam in somewhat broader perspective it may be useful to look at poll data from World War II, presumably the most "popular" in American history. A question comparable to the one under consideration here was posed by Gallup in early 1944, "Do you think you, yourself, will feel [in years to comel it was a mistake for us to have entered this war?" Only 14 percent answered affirmatively with 77 percent in the negative, a support rating solidly more favorable than those attained by the two later wars (Cantril and Strunk 1951:978).

Still, the picture is not quite so clear-cut. In 1967 Gallup posed an updated version of a question that had been asked frequently during World War II: "Do you feel you have a clear idea of what the Vietnam War is all about—that is, what we are fighting for?" Not surprisingly, great confusion was found—only 48 percent felt they knew (GOI 25). The comparable question during World War II elicited a more confident judgment — but not as much greater as might be expected. In fact in June 1942, 6 months after Pearl Harbor, only 53 percent of the public felt it had a clear idea of what the war was about. This proportion increased after that but approached 80 percent only in 1945 and at one point, the spring of 1944, dipped below 60 percent (Cantril and Strunk 1951:1077-78; Cantril 1967:48).

Sentiment for withdrawal from the Korean and Vietnam wars may very crudely be compared to an affirmative reply to the following query posed frequently during World War II: "If Hitler offered peace now to all countries on the basis of not going further but of leaving matters as they now are, would you favor or oppose such a peace?" Few supported the proposition in the early years of the war, but in early 1944 — while Hitler still held France — it was endorsed by more than 20 percent of the population and by about 15 percent thereafter (Cantril and Strunk 1951:1077-78). (my emphasis)

Okay, what does this tell us? In the Second World War, we see an increase over time in the number of people who perceived a clear purpose for the war, which I agree is some kind of reasonable approximation for confidence in the war. It fluctuated at times, like in spring of 1944 when the opening of an Allied second front in Europe was beginning to look long-delayed. Also, at around that point in early 1944, after large numbers of American deaths already incurred in the war and the draft was bringing millions of men into the service and keeping them there "for the duration", 20% looked favorably on something like a cease-fire in place with Germany.

To me, that looks like remarkably strong support for the war, despite the mounting casualties. In other words, Mueller couldn't show anything like the same effect he claims to have found in the Korean and Vietnam Wars, which is "a simple association: as casualties mount, support decreases".

But here's what Mueller concluded in 1973:

Thus World War II, although unquestionably much more highly supported by the public than the Korean and Vietnam wars, seems to have been rather less consensual than might be supposed. This may be so in part because the truth about Hitler's death camps only reached the American public in 1945 ... In mid-1943 only half the population thought that the death camp "rumors" were true. At the end of 1944 this portion had risen to 76 percent, but few anticipated that the death toll would be greater than "thousands." Therefore a major reason for supporting the war was largely unappreciated while it was going on. (my emphasis)

To put it bluntly, this doesn't make jack for sense to me. It would be putting it mildly to say that it's a forced interpretation of the data he cites.

And the business about the death camps comes out of the blue. I'm willing to believe that news of the camps (I'm assuming he's using "death camps" to refer to all the concentration camps; only four camps were technically designated "death camps") reinforced a positive public opinion of the American war effort. But, as pleasant as it might be to believe, I doubt if the news of the camps was nearly as decisive in the American public's support for the war as Mueller's paragraph just quoted suggests. It's far more likely that the public conviction that the war was genuinely in America's national interest drove the support for the war effort. His argument about the camps is essentially retroactive: if the public had known more about them, the war would have had even more support.

Even if Mueller's comments about the effect of the camps on public opinion is correct, it certainly seems to me that it contradicts the results he adduces from the Korean and Vietnam Wars that public support is driven by "a simple association: as casualties mount, support decreases." He's arguing that the camps much outweighed such considerations in the Second World War.

The bottom line is, that "simple association" he claims to have found doesn't wash. Mueller's analysis fails to sustain that connection. But it's fairly obvious why it would be appealing to fans of frequent war, or of the kind of perpetual war that Dick Cheney seems to envision, to use Mueller's conclusion as support for the line of reasoning I outlined at the beginning of this post.

In summarizing the history of the polling data, though, Mueller does present some more straightforward findings, some of which may be surprising to those who are going on conventional assumptions about public opinion and the Vietnam War (Mueller's data includes the Korean War): partisan influence (Democratic and Republican) seemed to be more influential on people's opinions of the wars than more ideological measures like "hawk/dove"; "hard hats" (blue-collar) workers were less supportive of the wars than more affluent people; the more educated tended to support the wars more, a measure that also correlates with class factors; younger people were more supportive of the wars than older ones, women less supportive than men, and blacks less supportive than whites.

One of the most interesting findings is that there seemed to be no significant association between direct personal risk, as in people likely to be drafted or with close family members likely to be drafted, and opposition to the wars. But that's another one that I would need to take a closer look at the data before I could fully accept it.

Mueller doesn't go into as much detail about attitudes toward antiwar activists. But he does suggest, fairly vaguely actually, that antiwar protests harmed the antiwar cause more than it helped, which also strikes me as an Alice-in-Wonderland notion:

Thus it seems entirely possible that, because their cause became associated with an extraordinarily unpopular reference group, any gain the opposition to the war in Vietnam may have achieved by forcefully bringing its point of view to public attention was nullified. But, again, it must be observed that the protest may still have been effective in a general political sense if it was successful in altering attitudes among elites and decision-makers.

A rude, snarky blogger might be tempted to say that it seemed "entirely possible" to a lot of people in 1973 that Richard Nixon wasn't a crook, too. And they were wrong about that.

A couple of things make this conclusion especially questionable, and are probably all but impossible to clarify through polling data. One is that the purpose of protest in an unpopular cause, which opposing the Vietnam War certainly was in 1965-67, is to bring problems to the attention of the public and challenge the prevailing opinion. In other words, being a pain in the rear. The fact that people find protesters to be pains in the rear is not the same as saying that the protests are ineffective.

The other things is that this involves the issue that liberal bloggers have recently been referring to as the "dirty hippies", better known as "culture war" issues. Guys with long hair, women on the birth control pill, black people challenging their status as second-class citizens, urban riots, antiwar protesters, dope-smoking hippies: for many conservative whites, these images all blended together into a general threat to the accustomed way of life of boring white people, an association that the Nixon administration actively encouraged. It would surely be a major challenge to sort out those more general "culture war" attitudes to focus specifically on antiwar protests. Much less to correlate that precise set of opinions with the effectiveness of the protest overall.

The increase in antiwar protest in the Vietnam War was also correlated with the length of the war, just as casualties were. But if Mueller was so ready to embrace the "simple association: as casualties mount, support decreases", it's hard to see how he could so easily dismiss a correlation between antiwar protests and changing attitudes.

Tags: iraq syndrome, john mueller, war presidents and public opinion