The Economics of National Security: Air Strikes in Vietnam
Proving causality is really hard
Did drone strikes during the Global War on Terror increase or decrease terrorist activity? Did overwhelming firepower during the Vietnam War increase or decrease Viet Cong (VC) insurgency events?
Military history and strategy often seem divorced from rigorous, quantified, historical economic studies, but we need economics to help us assess things like when the application of force is and is not successful. Economist Melissa Dell understands this. She is an economic historian who is as likely to study U.S. industrialization in the nineteenth century as she is the impacts of foreign military intervention on insurgency. These topics matter because, as Patrick Collison and Tyler Cowen put it, “humanity needs to get better at knowing how to get better.”
While it feels intuitive today that overwhelming firepower was not an effective strategy in Vietnam, it is actually difficult to demonstrate evidence of causality. For example, did VC activity increase because of air strikes, or were the areas targeted for air strikes the places where insurgency was already on the rise?
One of the biggest challenges about being an economic historian is there are no opportunities for randomized control trials, the gold standard for empirical evidence. Dell can’t ask the Department of Defense to randomly assign drone strikes so she can causally study if they increase or decrease terrorist activity. Instead, our economic historians must look for “natural experiments” or historical accidents, where there is plausibly random variation that influenced the allocation of military force at the margin.
Unearthing these natural experiments is a skill in and of itself, and Dell demonstrates mastery. Her 2017 paper Nation Building Through Foreign Intervention: Evidence from Discontinuities in Military Strategies took advantage of one of these natural experiments via a newly discovered algorithmic component of bombing strategy in Vietnam. She provides evidence that air strikes increased the military and political activities of the VC Communist insurgency and weakened local governance.
Project CHECO is a declassified Air Force Study that describes one of the variables that informed how the military selected targets for its weekly pre-planned Vietnam bombing missions. Hamlets (small villages in rural Vietnam) were prioritized for bombing based on their security rating. The rating was calculated using an algorithm combining data from 169 questions on security, political, and economic factors. Hamlets rated 1 were the least secure, and those rated 5 were the most secure. Less secure hamlets were prioritized for bombing.
While the hamlet security rating ranged continuously from 1 to 5, the output was rounded to the nearest whole number before being printed from a mainframe computer and passed to the military planners. This last detail matters because it allowed Dell to exploit a discontinuity. For example, a hamlet that received a score of 3.49 got rounded down to 3, while a hamlet that received a 3.52 got rounded up to 4. Two hamlets that were very similar received very different bombing treatment if they fell on different sides of the rounding threshold. Dell exploits the randomness introduced by the rounding thresholds to establish causality:
Moving from no strikes during the sample period to the sample average increased the probability that there was a village VC guerrilla squad - which consists of local fighters - by 27 percentage points. It also increased the probability that the VC Infrastructure - the VC’s political branch - was active by 25 percentage points and increased the probability of a VC-initiated attack on local security forces, government officials, or civilians by 9 percentage points.
That extensive data existed for Dell’s study is a reflection of the times. Secretary of Defense Robert McNamara and his Rand corporation “Whiz Kids” infamously tried to quantify the Vietnam War using systems analysis. McNamara drew on his previous experience as President of Ford Motor Company to introduce operations research and computing into Department of Defense (DoD) operations. While McNamara’s attempts to use statistics to forge an American victory in Vietnam were unsuccessful, they did produce an unusual trove of wartime data for economists today to mine.
Even though the era of McNamara produced a lot of data, it is not easy to parse today. Dell shows an impressive level of ingenuity to revive data sets many would have declared dead. During the Vietnam War, field data was key-punched into IBM 360 mainframe computers in Saigon and Washington. However, the Hamlet’s continuous scores (i.e. the non-rounded numbers), were never printed or saved from the mainframe’s memory. Dell had to re-calculate those scores using the original conditional probability matrices in uncatalogued documents at Fort McNair and the question responses from tapes held at the U.S. National Archives.
The flawed hamlet rating system with its arbitrary rounding threshold is a good example of a complex, ostensibly rational system that is mostly style over substance. It may seem contradictory that I’m critiquing McNamara’s data-driven war but simultaneously calling for more quantified studies like Dell’s. Let me try and thread the needle: McNamara and the DoD used data as a cudgel for justification of actions in Vietnam (this folly is explained in the superb The Best and the Brightest); Dell uses data to parse meaningful signal out of an inherently complex situation.
Interesting article. The Hamlet Security rating was certainly not perfect, but I think it is the best data the US could have realistically gotten given the times. If you take it as a general guide, the data does accurately describe which side controlled a village. I believe that something like it should have been adopted in the early 1960s.
One thing that needs to be taken into account on the effectiveness of air strikes and artillery strikes on Vietnamese villages is what happened afterwards.
From subjective commentary at the time, if US/ARVN forces captured the village and pacified it with local forces 24/7, they got the support of the villagers in the long-term. This typically occurred after 1968.
If, however, the Allies shelled a village and then let the VC maintain control of the village afterwards, then the villagers blamed the US/ARVN. This was most common before 1968.
In general, I believe that the Vietnamese villagers wanted peace and stability in their village, and they did not care who gave it to them.
Not sure if you can parse the existing data to prove or falsify this theory with existing data, though.
Madeline - I would be interested in your take on Mark Moyar's Triumph Regained: The Vietnam War, 1965-1968. Mark had access to fully declassified records from the US, Russia (then Soviet Union), the PRC, and Vietnam (formerly North Vietnam).