Book ideas #10: The Billion Dollar Spy
Ergodicity, value capture, tragedy of the commons, and other lectures from espionage
A couple of weeks ago, I came across Michael Vickers’ excellent memoir, much of which is about his years in the CIA. This then made me put my planned reads on hold and prioritize CIA-related books on my to-read list. One of the reading highlights was David E. Hoffman’s The Billion Dollar Spy. The book explains how Soviet engineer Adolf Tolkachev handed over military secrets to the CIA from 1978 until his capture in 1985. In this post, I highlight 5 things I think are interesting even for those who don’t care about espionage per se.
Choose your error
Should you go to that meeting with your spy? This is a perennial question faced by CIA case officers. If the officer goes, and everything is fine, he gets valuable information from his asset (i.e., the spy). If he goes, and things are not fine (i.e., they are caught while meeting), then his asset is compromised and likely to be jailed or executed. If he doesn’t go, there is no danger of apprehension, but nor is there information gain. So, the task of the CIA officer is to balance the two types of errors: the error of going when he shouldn’t and the error of not going when he should.
The two errors aren’t equally costly. Going when you shouldn’t go is typically a much more costly error, warranting little tolerance for failure. The CIA officer, naturally, prefers to avoid capture entirely. But if it’s inevitable, he has a strong preference for being apprehended before—rather than during—the meeting. This way, his spy is not compromised.
This preference can be so strong, in fact, that the CIA officer might take extreme measures to check if he is being followed. For example, one of Tolkachev’s case officers, David Rolph, did all of the following before one of his meetings with Tolkachev: he stopped at a theatre, walked to an antique store, walked into a random apartment building, and wandered around in a small, empty park. These activities might seem banal to us, but they would send off red flags at the KGB either due to the nature of the activity (e.g., surveillance is not possible inside a random building), its timing (e.g., being in the park late at night), or simply because Rolph wouldn’t typically frequent these places, which would make any visit to these places suspicious. He was, in effect, saying: “I’m here, KGB, doing suspicious things, catch me now if you are following me. If you are not catching me, it’s because I’m not followed, so I can safely proceed to my meeting with Tolkachev.”
Value capture as a spy
Tolkachev provided value to the U.S. that was worth potentially billions of dollars (hint: the title of the book). Comparatively, he got very little in return. I don’t mean he got an unfair deal necessarily, just that, in absolute terms, he captured a small share of the value he created for the U.S. Which raises the following question: How much of the value created by a spy is captured by him, and how does this number relate to what can be captured by people in somewhat similar domains?
Not worrying about the exact numbers, just focusing on the order of magnitude, let’s say Tolkachev saved somewhere between 1 and 10 billion dollars for the U.S. in research spending (at one point, long before his capture, the U.S. Air Force estimated his contributions in the neighborhood of 2 billion dollars). It’s also mentioned that his account balance stood at 2 million dollars at some point. So, let’s say he captured 0.1% of the value he created.
Contrast this with the case of stock trader Ivan Boesky. He got insider tips worth around 50M dollars, promising a 5% cut to his contact, Dennis Levine, on new stock purchases (and 1% if Boesky already held the stock). Levine’s share was at least a magnitude larger than Tolkachev’s. Why?
Surely, there are many differences between the two cases, and we shouldn’t draw large conclusions from two hand-picked examples. That said, three factors come to mind that could contribute to the observed gap.
First, unlike insider trading where the value of information is transparent (the information eventually becomes public, stock prices get updated, and the trader’s profit can be calculated), there is asymmetric information in espionage. Tolkachev didn’t know the exact (or even the approximate) amount he saved in R&D costs for the U.S. He was aware of his value and frequently alluded to it when negotiating, but it’s not inconceivable that he underestimated his value by an order of magnitude.
Second, he didn’t have a lot of bargaining power because the market for his military secrets was very small. He could go to the U.S., maybe the U.K., or some other Western country. In contrast, in insider trading, the potential pool of people who want to benefit from private information is much larger. So, even if the number of sellers in both markets is similar, in espionage you only have a few buyers, while in insider trading you have many.
Third, it’s not just economic market power that was asymmetric. Tolkachev’s personal safety depended on the very actor to whom he handed over information, further weakening his bargaining power.
All these accounts assume that Tolkachev wanted to maximize his monetary return in the first place and that economic forces prevented him from doing so. Most likely, however, his motivation was different, which makes for a rather anti-climactic conclusion: if he didn’t care about money, it was unlikely he would get a lot of money. But if not money, then what was his motivation?
The CIA frequently classifies assets like spies based on their motivations with the acronym MICE (money, ideology, coercion, ego). Many of the CIA’s Soviet assets were motivated by ideology; they disliked the system or wanted revenge for the harm they and their families had suffered. Like most of their contemporaries, the parents of Tolkachev’s wife were persecuted during Stalin’s purges. It’s impossible to know one’s motivations for sure, but it seems like money was not a major factor for him; he thought of it more as a (costly) signal that the U.S. valued his services.
Tragedy of the commons
Initially, Tolkachev handed over secrets about radar systems, his area of expertise. Over time, however, he expanded the scope of his operations, first copying documents that, though not directly in his area, he could plausibly need for his work. Then he also started copying documents that were unrelated to his work.
As news about a prolific Soviet spy (i.e., Tolkachev, codename: CKSPHERE) with access to military secrets spread in the relevant U.S. military circles, requests for additional secrets started pouring in. On one occasion, for instance, he was sent a detailed list of requests, including stealing details about supersonic strategic bombers, a topic clearly outside his line of work.
Access to Tolkachev’s resources is an example of what economists call the tragedy of the commons, where individuals’ unrestricted use of a shared resource leads to its premature depletion. Here, this would play out as follows. Every scientist or military department only cares about receiving information relevant to their field. But if the spy tries to steal all this information, he increases (his already high) chances of getting caught. In other words, the spy has limited capacity, and access to his resources must, therefore, be managed in a sensible way.
In tragedy of the commons cases (e.g., overfishing in unregulated waters), one solution would be to have a quota system to manage access to the scarce resource. Here, the quota allocation role is played by the CIA case officer. The goal is to manage the scarce public good—access to Tolkachev—in a way that guarantees that this public good would not be depleted quickly (i.e., avoid capture by the KGB). Maybe a spy’s excess capacity could be allocated using a quota system too, but with risk-adjusted limits for each information-requesting department.
Ergodicity and survivorship bias
Every operation (e.g., a meeting with the CIA officer, a dead drop—leaving objects at a previously agreed location) comes with a given risk. For simplicity, I’ll assume that this risk is the same for every operation—say, a 1-in-100 chance of capture. So, you would lose one spy, on average, in every 100 operations. Theoretically, the 100 operations could come about by 1 spy doing 100 operations or 100 spies doing 1 operation each. This property is called ergodicity: the expected outcome is the same regardless of whether it is the result of a single agent doing something many times or many agents doing something once. Now, being a spy is clearly non-ergodic: if we only have one spy and he is captured in Operation No. 37, there won’t be any more operations, so it doesn’t matter that his expected capture rate is 1 in 100.
Ergodicity aside, being a spy seems to come with an elevated lifetime risk of capture and death. Famous CIA assets like Tolkachev, Penkovsky, and Polyakov operated for a different number of years (e.g., 20+ years in the case of Polyakov), but their fate was ultimately the same: they were captured and executed.
There are other lines of work that come with a high lifetime risk of death, such as being a bomber plane pilot or working on a submarine in WWII, or to mention contemporary examples, working in commercial fishing, logging, and mining. These professionals, however, face the risk of injury and death only while they work. Unlike spies, they don’t need to worry about betrayals and leaks for the rest of their lives.
We can estimate mortality risk in, say, commercial fishing because we observe the outcome of every fisherman. But we only observe the outcomes of spies who were captured, and not those of successful spies who were not captured (except for some who wrote memoirs or became known for other reasons). Therefore, an estimate based on the known spy cases almost certainly overestimates the real risk, but we don’t know by how much. Espionage could even be a low-risk profession if most spies escape detection during their careers. This seems highly unlikely, but we cannot be sure unless we can observe the full spectrum of outcomes—which we currently can’t.
Spy requests as a window into society
Spies are extraordinary in some ways yet ordinary in other ways. While they tend to be remembered for the former, learning about the latter is very valuable too. Spies living in the USSR, just like ordinary citizens, experienced product scarcity. But unlike ordinary citizens, they had a third option beyond living without the desired product or purchasing it on the black market: they could ask their case officer to procure it for them.
This is what Tolkachev did too. For instance, here is a list of some items that he requested from the CIA: rock music albums (e.g., Led Zeppelin and Pink Floyd) for his son, a Sony Walkman with earphones, pencils of various hardness, Polish blades for his safety razors, rosehip and buckthorn oil to treat his chronic antacid gastritis, a French medicine for his gum disease, bottles of India ink for his son’s drafting equipment, and prescription eyeglasses for him and his wife.
Recall that Tolkachev was a top engineer, living in relatively lavish conditions by Soviet standards. It’s a sad statement about Soviet society that even he, a relatively well-off citizen, felt he needed the CIA’s help to get something as basic as prescription eyeglasses.