Joshua PollackBombs and Dollars

Two professors who study nuclear proliferation — Nick Miller at Dartmouth and Vipin Narang at MIT — have a neat little piece in Politico Magazine about which poli-sci theories predicted North Korea’s bomb and which didn’t. Long story short: most didn’t. Etel Solingen’s Nuclear Logics (2007) holds up well. It was, of course, published four years after North Korea’s withdrawal from the Nuclear Non-Proliferation Treaty, and the year after its first nuclear test. Still, case studies aside, her earlier work presented the same essential set of ideas. So, bravo!

There are many competing ideas in the field, but the widest divide is between those who privilege capabilities and those who privilege intentions. Or, if you prefer the lingo of microeconomics, there’s a “supply” school and a “demand” school. If you’re paying attention at all to my occasional scribblings, like this recent feature with Catherine Dill, you’ll know that I’m on the demand/intentions side. That’s not to say that some very basic level of resources and especially managerial capacity isn’t needed. (Over to you, Malfrid Braut-Hegghammer.) But most states of interest have what it takes.

I bring this up because reading the Miller-Narang piece reminded me of a data table I pulled together a few years ago while working on a project that’s still on a back burner. The idea was to compare states’ resource levels when they “went nuclear” and see what the pattern looks like. In particular, where does impoverished North Korea fall in the list?

First, I went through the list of countries that have conducted nuclear explosions, or are believed by scholars to have built devices for that purpose. There are ten of them. In each case, I identified the year of the first such test (or my sense of the consensus of the earliest time when such a first test could have happened). Then I looked up each country’s gross domestic product for that year, indexed for inflation.

All this is based on the view that GDP–or some other, comparable measure of economic output–is the single best proxy for “resources.” We often make the mistake of thinking of countries as “rich” or “poor” as a whole based on their GDP per capita. While that’s a good measure for level of development, it doesn’t tell you what sort of pool of resources the state potentially has at its command. For that question, just plain old GDP is the way to go.

Here’s what I found:

Country & year GDP ($1M 1990 USD)
USA 1945 $1,644,761
India 1974 $500,146
USSR 1949 $465,631
China 1964 $450,312
UK 1952 $357,585
France 1960 $344,609
Pakistan 1998 $139,063
South Africa 1978 $116,077
North Korea 2006 $25,310
Israel 1967 $16,758

Ital. = no confirmed nuclear tests.

Yep, that’s right! North Korea, as poor as it is, isn’t at the bottom of the list.

Israel in 1967 was, as a whole, poorer than North Korea in 2006, as a whole. For perspective, Israel had fewer than 3 million people then; North Korea had about 25 million in 2006.

Also surprising: India and China are near the top of the list. But the same sort of observation applies: they were (and are) hugely populous, much moreso than the European countries just below China in the list.

The “GDP per capita” version of the list (not shown here) is much more intuitive: the bottom three countries are North Korea (2006), India (1974), and China (1964), China being the poorest per capita. But, as I’ve argued above, that’s not relevant to a centralized national project like making the Bomb. Our intuitions about wealth and poverty mislead us on this score.

The two ends of the list of bomb-builders are separated by a couple orders of magnitude’s worth of GDP. A country with less than $17 billion in annual output (in 1990 US dollars, like all the other figures above) managed to pull it off, according to Avner Cohen among others. That’s just somewhat more than half of the output that North Korea is estimated to have had at the time of its first nuclear test.

The bottom line: resource levels–as approximated by GDP–are not the key.

(Caveats: The figures come from Angus Maddison’s historical statistics project. The issue of opaque nuclear proliferation means that Pakistan’s correct year arguably should be sometime in the 1980s. Those with different analytical preferences are free to build their own versions of the table, although I doubt it would end up being drastically different.)


  1. Allen Simmer (History)

    So, who’s next? Or rather, who should we expect to be next given your data?

    • Joshua Pollack (History)

      Two answers:

      1. Nobody. Everyone else (except South Sudan) is in the NPT and shows no obvious signs of leaving. Iran might be pushed out, but for the time being, they have made their choice.

      2. Anybody. Other than, say, Vatican City, it seems to me that there are very few countries that could not join this list if they were sufficiently determined to do so. But see (1) above.

  2. Benjamin Shen (History)

    There was also the time advantage. The level of technology, especially electronics, available to North Korea in 2006 was much more advanced than that available to Israel in 1967.

    • Joshua Pollack (History)

      Yep. Here’s how I think about chronology, in a nutshell. The earlier a country started, the less external pressure it faced, simply because it took awhile for nonproliferation norms to gain traction. Thus the shift to opacity in the 1960s and 1970s, followed by a transition to near-universal adherence. But the later a country started, the more technological (and, typically, economic) advantages it had.

      How much these respective considerations matter is a good question, although I think the norms are the most powerful consideration in play. At least, that’s what I think explains this sequence:

      [NPT talks begin]
      1967 [era of opaque proliferation begins]
      [NPT achieves near-universal adherence]

  3. Bradley Laing (History)

    Question” some of the 1952 UK Scientists were Ex-Manhattan Project guys. Does any of the US money spend on its own bomb research count toward the UK program? Can you add or subtract some amount of money from the UK column, based on that fact?

    • Joshua Pollack (History)

      I’m tempted to call that an analytical rabbit hole, but it’s more like a rabbit warren. The UK and Canada helped the US; the USSR helped China (up to a point), and France helped Israel. No country is an island, although there’s quite a lot of variation in the (known) degree of assistance across the cases. Rather than discounting GDP figures, I’d treat that as a separate question.

    • ajay (History)

      “Does any of the US money spend on its own bomb research count toward the UK program?”

      Or indeed the Soviet program (via Klaus Fuchs, etc.)

    • Joshua Pollack (History)

      It’s safe to say that the first successful bomb program (the Mannhattan Project) faced obstacles that none of the later ones did. They didn’t really know it was feasible!

  4. ajay (History)

    Don’t be too quick to dismiss the importance of per capita gdp. David Edgerton, in “Britain’s War Machine”, makes the point that Britain and Germany had similar GDP in 1939, but Germany had a much bigger population. He argues that this gave Britain, not Germany, the advantage in war economic terms, because Britain had far more manufacturing capacity left over for war production once it had met its population’s basic needs for food, clothing etc.

    • Joshua Pollack (History)

      I see what you mean. On the other hand, China and North Korea experienced famines while they were developing their nuclear programs. And most South Africans lived under Apartheid. So feeding, clothing, and housing people were not always high on the list. Perhaps that should be considered a regime-type question.

      For the record, here’s the GDP-per-capita version of the chart (all figures in 1990 USD):

      USA 1945: $11,708.65
      France 1960: $7,397.56
      UK 1952: $7,090.72
      Israel 1967: $6,221.57
      South Africa 1978: $4,174.07
      USSR 1949: 2,623.27
      Pakistan 1998: $1,782.34
      North Korea 2006: $1,133.37
      India 1974: $843.42
      China 1964: $644.82

  5. George Herbert (History)

    I prefer to ascribe an earlier date to Pakistan for these types of discussions ; but it’s a good question of when, precisely.

    • Joshua Pollack (History)

      Yeah, that would make sense. What year do you suppose would work? 1985, maybe?

    • George Herbert (History)

      For the sake of argument, yes, that works.

  6. Allen Simmer (History)

    To continue with the who’s next, I ask the frightening question of who’s the next (first) non-state actor to acquire this capability, whether by acquisition or development?

    • Joshua Pollack (History)

      Your guess is good as mine. So far, no such group has demonstrated the level of organizational competency required to seriously test the proposition.

    • Brenton (History)

      As Alex Wellerstein pointed out in his post Doomsday on the Cheap, the main threat with non-state actors is the risk of them hijacking existing infrastructure from a nuclear state. Joshua is correct to point out that a rare NGO would be able to pull off a development program. I’m also cautiously optimistic that 28 yrs later, we can be reasonably assured that none of the Soviet stockpile is in clandestine hands after the collapse. Not that it couldn’t happen, but there’s really no way to predict such an event.

    • Joshua Pollack (History)

      I’d forgotten about that post! Interesting reading.

      When you say “NGO,” I have this weird vision of the Nuclear Threat Initiative building its own arsenal.

    • Allen Simmer (History)

      Enlightening, thank you. Some elements inside Pakistan concern me in this regard.

  7. Jonah Speaks (History)

    If you are talking total GDP (the right number in my view), North Korea is beyond number 100 on the list. An equally relevant number is the percent of GDP needed to achieve “success” in developing nuclear weapons. If the cost of developing nuclear weapons is about the same for all countries, it will be proportionally more costly for a poor country than a rich country. The North Korean regime evidently has a very high “demand” compared to most countries at the same or higher level of total GDP.

    Do you have any estimates of total cost for development, and of cost as a percentage of GDP? Has this cost gone down over time as the nuclear technology has become better understood, or as complementary technologies (e.g., computers) have come down in price? Is the total cost of development about the same (or higher or lower) for high per-capita countries as for low per-capita countries?

  8. Allen Simmer (History)

    Events of the day have provided an answer to my question of who’s next: Japan.

  9. Alistair (History)

    Time is important: late starters have the advantage of both better technology and more importantly knowledge of how to do things.

    When you control for the time variable under a simple linear model, then a lot of the above variance disappears. The std deviation of the logs down from a factor of 4 in your basic list to a factor of about 1.5. If I use Ln(GDP), which is arguably a better measure, then I can get the residual std dev down to a factor of 1.1. That’s really quite tight.

    The coefficients work out as 1 year-post-1945 is equivalent to about 5% extra GDP.

    • Alistair (History)

      Basically, a model of Ln(GDP) and time as a proxy for “technology” predicts a tight “threshold” for nuclear capability. Even though there’s 2 parameters and N=10 it’s a sufficient improvement on your 1 parameter GDP model to be statistically preferred.

      In economics terms, and I’m afraid contrary to your initial thoughts, the conclusion is that “supply side” rather than “demand side” explains nearly all the variance here.

      (Though to be fair, obviously states have to want to get nukes. There are scores of states above the threshold without nukes. So that can properly be said to be demand side variance)

    • Joshua Pollack (History)

      Indeed, this is only the states that are known to have gone nuclear. Making a judgment about the correlates of proliferation based on these 10 out of about 190 would be selecting on the dependent variable.

  10. robgoldston (History)

    Much though I am in favor of the Iran Deal, I think you have to count Iran as in the NPT, but potentially interested in getting out.