Is it unusual for analysts to downplay the progress of a nuclear program?
I spent several years digging through government archives and filed declassification requests to find answers to this question. My research compared how accurately U.S., British and West and East German intelligence tracked nuclear programs during the Cold War. U.S. intelligence performed best, followed by the British.
Nuclear espionage is always a challenge. Governments conceal their nuclear activities from curious foreign eyes. And even the biggest intelligence budget and the latest spy gadgetry do not guarantee omniscience.
Nuclear weapons are both a function of a state’s ability to acquire them — and its wish to do so. Knowing which technologies and materials a country has is enough for estimates of ability. Whether a country wants to build weapons, and how fast, might not even be known to its leaders.
Despite the difficulties, Western intelligence agencies have a remarkably good record in tracking the spread of nuclear weapons technology. By my calculations, British and U.S. intelligence estimates were correct well over 90 percent of the time. Continental European intelligence assessors erred in half of their proliferation estimates.
That U.S. intelligence is among the best in the world at tracking proliferation may come as a surprise. Scholars, journalists and politicians tend to look at big unrepresentative intelligence failures — such as Pearl Harbor or the Sept. 11, 2001, attacks — and draw their conclusions about the intelligence community’s analytical capacity.
Underestimation is the norm
While large errors might be rare, they recur in patterns. My research shows intelligence agencies are more prone to underestimate the progress a country is making than to exaggerate it. Since the 1960s, underestimates were 10 times as common as overestimates. During the 1962 Cuban missile crisis, for example, the CIA failed to provide strategic warning of deployment of missiles to Cuba.
But the 2003 exaggerations of Iraqi weapons of mass destruction are also seared into history. It was a major exception to have U.S. intelligence overestimate another country’s nuclear weapons potential. An underestimate of North Korea would be more typical of historical trends.
My research revealed four other trends:
1) Size stumps spies: The larger and more complex a nuclear program, the less correct the proliferation assessments were. Programs that produced nuclear energy were more likely to be mischaracterized by intelligence agencies. Simply collecting more information about bigger programs does not seem to offset the problems with tracking complex capabilities.
2) Safeguards for civilian technology improve intelligence: Many of the same technologies and materials that can generate nuclear energy are also needed to build nuclear weapons. The International Atomic Energy Agency (IAEA) monitors civilian nuclear programs to keep this technology from building nuclear weapons. More stringent IAEA inspection protocols tended to produce better intelligence assessments about the nuclear programs. Even if inspectors are only checking the accounting of open nuclear installations, their presence makes concealing secret work on weapons much harder.
3) Trade is not transparency: Intelligence assessments of import-heavy programs are typically less accurate, however. This is counterintuitive because it should be easier to track international supply chains.
Intelligence about the A.Q. Khan illicit nuclear supply network in 2004, for example, blew the lid off Libya’s mail-order nuclear program. Khan started his career as an importer of goods under the watch of British and U.S. intelligence — but this fact did not save them from underestimating Pakistan’s nuclear progress.
4) Words work wonders, at times: Sometimes leaders announce whether they want nuclear weapons. That helps intelligence analysts, at least most of the time.
In my research, formal diplomatic commitments proved to be information-rich for intelligence assessors. Intelligence analysts do not naively accept promises of nuclear abstention, but looked at the restrictions governments were unwilling to agree to for clues about which options governments wanted to save for future use.
Governments bluff, too, making it difficult to know when to take their protestations of peaceful intentions seriously. Pakistan and North Korea weren’t shy about their desire for nuclear weapons, and yet U.S. intelligence underestimated their ability to do so.
What’s worse — an intelligence over- or underestimation? Overestimates can lead to unnecessary confrontation, as was the case with Iraq. Underestimates, as the reporting on North Korea suggests, risks the sudden realization that a country has a clear nuclear weapons capacity.
But even the best intelligence does not help U.S. policymakers keep the nuclear club limited and exclusive. That is a devilishly difficult problem, not least because nuclear weapons and the materials needed to build them are no longer the secret engineering challenge they were in the 1950s.
Even as intelligence agencies failed to warn of specific nuclear and missile tests, there are decades-old intelligence warnings that North Korea would gain the ability to target the U.S. mainland right around now.
A better summary of U.S. intelligence performance would then consider the successful strategic warning to policymakers, rather than a failure of tactical warning of the quarterly progress of North Korea’s nuclear missile program.
In 1962, although the CIA was late to the party, U.S. analysts identified the Soviet nuclear-armed missiles before they were operational. This timing probably raised the odds of a peaceful resolution to the standoff. In 2018, though, policymakers will not have that advantage against Kim Jong-un.
Alex Bollfrass is a Stanton Nuclear Security postdoctoral fellow at the Harvard Kennedy School’s Belfer Center.