Strategic Assessment in War

Scott Sigmund Gartner

Yale University Press, 1997; Pp. vii,248

As Clausewitz remarked long ago, information and its interpretation are among the most important elements of warfare. War moves quickly and generates a deluge of information. One of the most difficult tasks of military commanders and civilian decision makers is to use this information to determine whether their strategies achieve the desired goals. (Strategy is here defined simply as the way an organization operates its military forces to achieve specific goals; p.18)

In this short book, Gartner examines how decision makers use battlefield indicators to assess the performance of their strategies and, should they consider it necessary, switch to alternative ones. He argues that organizations form beliefs about the likelihood of success, as defined through the organizational lens, and then use quantitative indicators to measure their progress. These indicators will be closely watched during the war, and will be continuously evaluated. Sudden and dramatic changes send signals that decision makers can distinguish from noise, and can inform about the performance of their organization.

The first part of the argument is a well-known theory from the organizational literature (e.g. Simon, March) which posits that organizations do not have the resources and time to continuously search for best strategies, as usually required by strict rationality, and that they essentially settle for those that are only "good enough" given the information and the goals (bounded rationality). This literature usually defined "good enough" in terms of absolute values of some indicators considered most important by decision makers. This approach, dubbed the "absolute value" approach by the author, has been used by historians to explain wartime changes of strategy.

Gartner goes a step further, and this is the contribution of the book, by arguing that this approach has a glaring hole in it: There are many instances that indicators reach maximum (or minimum) absolute values without corresponding reaction from decision makers. Also, the absolute value approach has the small problem that it overpredicts changes because it is not until after the war that highest figures become available. During the war, only past figures can be used for comparison, so any increase is a maximum. Usually, the organizational theorists "solve" the problem by assuming a level of satisfaction (aspiration level) that must be reached, but it has never been clear to me how this is determined.

Gartner, on the other hand, argues that decision makers do not change strategy unless the situation gets worse and does that at an accelerating rate. In other words, they do not use the absolute values of the indicators, but their first and second derivatives, to evaluate performance. Thus, even record levels of bad outcomes will not necessarily lead to changes unless they also indicate a situation that is deteriorating and doing so at a faster pace.

The other remarkable finding in the book concerns the choice of indicators. Gartner argues that organizations interpret their missions in very specific ways that translate into choices of what information to use and how to use it. It is entirely possible that two organizations can look at the same data and reach two diametrically opposed conclusions about the progress of war. Having chosen an indicator, organizations are more likely to stick with it than change, and so institutional conflicts can develop. Finally, Gartner also notes how an organization can become deadlocked when its indicators send it mixed signals. When one indicator shows improvement at an increasing rate and the other shows accelerating deterioration, the organization will become unable to assess its performance, and will not be able to make coherent decisions.

Gartner tests his theory against four case studies: British submarine campaigns in World Wars I and II, U.S. ground strategy in Vietnam, and U.S. decision making during the Iranian hostage crisis. I found the World War I and Vietnam cases most intriguing (which is not to say that the others were not), although one may disagree with some of the interpretation.

Throughout World War I, the Royal Navy used the number of sunk German submarines as its dominant indicator. The civilian government, on the other hand, used tonnage of shipping lost to enemy submarines. The admirals believed that patrols, a seek and destroy strategy, was most effective for sinking enemy subs, and they saw their mission as ensuring that they could do that at a faster rate than the Germans could build them This was a strategy that they believed would in turn ensure the safety of the merchant fleet in the long run. The standard story is that the admirals were "wedded" to their strategy and that even after it began to fail, they were unable to see what everyone else did: That switching to convoys would be a more effective way to deal with unrestricted submarine warfare employed by the Germans. Only after Lloyd George ordered them to implement convoys, did they do so. Remarkably, they also deemed the new strategy an immediate success (p.68).

Gartner notes that the standard story raises more questions than it answers. For example, shipping losses reached record levels at the end of 1914, in 1915, and in 1916. Yet the civilians did not react until 1917. Why did the data, once they convinced the civilians failed to persuade the admiralty? Why did the admiralty judge convoys a success? Gartner argues that it was not until 1917 (when unrestricted submarine warfare was resumed) did the merchant shipping experience losses that were increasing at an accelerating rate. (There was one previous instance in 1915 but then Germany abandoned the strategy under U.S. pressure before the British adjusted strategy; this is outside the model and could not have been predicted by it.) At the same time, however, the number of enemy subs sunk was also increasing at an accelerating rate, and so the admirals saw their strategy as working. Thus, the civilians and the military reached opposite conclusions from the same data. However, in April 1917, the number of subs destroyed fell dramatically, and thus the two indicators were finally moving in the same direction, and the admirals changed to convoys. When it turned out that convoys were even more effective in sinking German submarines, the navy judged the new strategy a success.

The Vietnam case is even more interesting. There were three principal actors involved: the U.S. army, which used number of enemy soldiers killed as its indicator, U.S. Congress, which used the number of Americans killed, and the Johnson administration, which used both because it wanted to win the war but was sensitive to domestic political costs. During the Tet Offensive, the number of Vietcong soldiers killed dramatically increased at unprecedented rates, prompting the army to evaluate Tet as a complete success. At the same time, the number of Americans killed also jumped at record rates, leading Congress to conclude that the war was failing. The administration, having received mixed signals, became deadlocked and unable to reach consensus as it had been before Tet. In addition, Gartner gives an example of the U.S. Marines, which used a "combined action platoons" (CAP) strategy that emphasized local stability and correspondingly used an indicator (e.g. rice production) decidedly different from the army's and its search-and-destroy strategy. Since CAP resulted in little Vietcong casualties, it was deemed a failure by the army at the same time when the Marines' indicators showed success. Eventually, the Army pressured the Marines into switching strategies.

This excellent book does have some shortcomings. First, and most importantly, although the author cites decision makers mentioning quantitative data and their movement, there is little historical evidence to prove that they actually use the rate of change and its speed as the author claims. With few exceptions, the statements are so vague that virtually anything can be inferred from them. Ideally, I would have liked to see evidence (memos, reports, memoirs) that shows decision makers actually using the derivatives. Second, the model is a bit silent as to the source of change driving the indicators. For example, the claim is that the Carter administration resolved on the rescue attempt once the president's rating declined at an increasing rate. But the source of the decline was not the hostage crisis (even during the decline the public retained the favorable opinion of how he handled that). Rather, it was the Soviet invasion of Afghanistan that resulted in this drop. Why did the Carter administration believe that an action where its current policy was seen as fine would improve the situation in an area where it was failing? Finally, some of the interpretation appears a bit stretched. For example, the deadlock of the Johnson administration is seen as resulting from movement in opposite directions of the Army and Congress indicators (pp. 140-1). But one could explain it simply by pointing out that, unlike Congress, the administration wanted to win the war badly, but when the costs climbed, it became conflicted because it could not see itself sustaining the war at the rate it was incurring losses. Thus, it was not a matter of inability to "determine whether the U.S. situation in Vietnam was good or bad," but whether victory can be had given the costs.

There is also a lot of room for future work. The model is silent as to the choice of indicators. Why do organizations pick the ones they do? Although we can explain some choices ex post (e.g. the Army, which was prepared to fight a huge European war against the Warsaw Pact emphasized traditional measures; while the Marines were prepared for small-scale wars), is there a way to do so ex ante? The model is also silent as to whether a change in policy will, in fact, occur. It only predicts that decision makers will see the need for change. This is a huge gap in the theory because we effectively do not know whether the movement and rate of change of indicators will actually result in a strategic shift. The author picked examples where they do, but this may not be necessarily so.

Four appendices with data for the relevant indicators in World Wars I and II, the Vietnam War, and President Carter's approval ratings.

February 23, 2002

@book{gartner-97,
  title     = {Strategic Assessment in War},
  author    = {Scott Sigmund Gartner},
  year      = {1997},
  publisher = {Yale University Press},
  address   = {New Haven},
  isbn      = {0-300-06034-3},
  note      = {Index, data appendices; Pp. vii,248}
}