Robert Jervis – Why the CIA Doesn't Do Better


Event Date: 29 April – 1 May 2011
East Midlands Conference Centre
University of Nottingham  
University Park
Nottingham NG7 2RJ

Landscapes of Secrecy: The CIA in History, Fiction and Memory

Robert Jervis specializes in international politics in general and security policy, decision making, and theories of conflict and cooperation in particular. His book Why Intelligence Fails: Lessons from the Iranian Revolution and the Iraq War was published by Cornell University Press in March 2010. Among his previous books are American Foreign Policy in a New Era (Routledge, 2005), System Effects: Complexity in Political and Social Life (Princeton 1997); The Meaning of the Nuclear Revolution (Cornell 1989); Perception and Misperception in International Politics (Princeton 1976); and The Logic of Images in International Relations (Columbia 1989). Jervis also is a coeditor of the Security Studies Series published by Cornell University Press. He serves on the board of nine scholarly journals, and has authored over 100 publications.

Dr. Jervis is a fellow of the American Association for the Advancement of Science and the American Academy of Arts and Sciences. He has also served as the president of the American Political Science Association. In 1990 he received the Grawemeyer Award for his book The Meaning of the Nuclear Revolution.

Professor Jervis earned his BA from Oberlin College in 1962. He received his PhD from the University of California, Berkeley in 1968. From 1968 to 1974 he was appointed an assistant (1968-1972) and associate (1972-1974) professor of government at Harvard University. From 1974 to 1980 he was a professor of political science at the University of California, Los Angeles.

Contact details: Robert Jervis, International Affairs Building, Room 1333, Adlai E. Stevenson Professor of Political Science, 116th Street and Broadway, New York, NY 10027 USA




I. Introduction

1. US-centered, although I suspect some of the remarks apply to other countries. A great deal of value has been written that is relevant but my remarks will not be heavily footnoted.

2. Although I work with the Intelligence Community (IC) more than most academics (and this may introduce a bias), by exposure is actually limited.

3. The CIA is the object of much public fascination and scorn. Here let me just take a paragraph from my book:

For the general public, intelligence is not popular for the additional reasons that its two prime characteristics of secrecy and covert action clash, if not with American traditions, then with the American self-image, and even those who applaud the results are likely to be uncomfortable with the means. As I noted in the introduction, it is telling that discussions of interventions in others’ internal politics, and especially attempts to overthrow their regimes, are couched in terms of CIA’s interventions despite the fact that CIA acts under instructions from the president. Critics, even those on the left, shy away from the correct label, which is that it is a U.S. government intervention. Political leaders see little reason to encourage a better understanding.

4. Interactions with policy-makers often are difficult. A particularly important and misunderstood episode is the November 2007 NIE on the Iranian nuclear program (for details see the excerpt from another paper that I’ve included below as an Annex.)

II. Methodological Problems with Judging Performance

1. Information about CIA’s performance is limited. Part of broader question of how much intelligence is still the “missing dimension,” which relates in large part to declassification policy (& also to what is popular among historians). Budiansky is very pessimistic in Intelligence and National Security, 12/10).

2. Hard to know what the right baseline is for judging how well CIA is doing.

    • a) What is it’s “batting average”?
    • b) What should we expect it to be?
    • c) How does it compare to other services? To other units in the IC?
    • d) How does it compare to policy-makers and the media?
    • (a) is the simplest question, but really can’t be answered–controversies, e.g., on how well it did with the end of the Cold War and the collapse of the Soviet Union. Some of the arguments on this question are rooted in political and ideological disputes. The other questions are even harder to answer–and I certainly can’t answer them.

3. Should we concentrate on the important cases in judging performance?

4. If we really care about policy, can we tell how influential estimates were? Gaddis (Diplomatic History, vol 13, Spring 1989) and Immerman (ibid, vol. 32, Jan. 2008; also see Herman, McDonald, and Mastny, eds., Did Intelligence Matter in the Cold War?) were skeptical, but it is easy to overlook the biggest contributions like a) convincing leaders in late 1940s that Stalin didn’t want war and that a long Cold War was possible (Mitrovich, Undermining the Kremlin); not being surprised by new Soviet systems, and giving leaders confidence of this.

III. Why CIA Doesn’t Do Better

1. It is hard–people are taken by surprise all the times in their personal lives (e.g., when they find that their spouse is cheating on them).

    • 1a. Deception.
    • 1b. Others may not know what they’ll do or how they’ll act and can (and do) change their minds. If we had day-to-day access to Chinese decision-making after the US crossed the 38th parallel in Korea we might not have been better off than we had just tried to focus on what Chinese interests were.
    • 1c. Most obviously, the world is very complex and contradictory. If we are still debating the causes of World War I should we expect intelligence services to be able to understand and predict (and the relations between the two are worth exploring–but not here).
    • 1d. Empathy is very difficult.
    • 1e. As discussed in my book, the “Betts Paradox” is central–we are strongly and intelligently influenced by our generalizations and theories, and while these serve us well in most cases they mislead us in unusual ones, such as the revolution that overthrew the Shah and the Iraq WMD case.

2. When and where does intelligence have great advantages over policy-makers? Latter have often met their opposite numbers and have a finely-developed sense of politics. The former presumably have

    • a) more detailed knowledge;
    • b) more perspective both in knowing the history and in not being wedded to a policy;
    • c) better analytical training, tools, and habits.

3. But today for many questions secret information is less relevant. Intelligence may still have a comparative advantage over policy-makers, but less so over informed observers who are relaying on “open source” information. I’m not sure the IC has fully come to grips with this.

4. Much of the American IC, especially the CIA, lives in a bubble. On the good side, this insulates it from a surprising amount of politics, but may lead it to miss a great deal.

5. Political pressures are blamed for many errors, such as the Iraq WMD fiasco. But my previous indicates that I think this is often exaggerated and, as I argued in my book, is largely incorrect in the Iraq case. Often present and important in some cases (Yellow Rain may be an example–see Rod Barton, The Weapons Detective), but there usually is room to dispute (e.g., did Gates politicize when he was DDCI?). The issues, both of substance and of definition are complex, and I’ve discussed them (and referred to the literature) in the last chapter of my book.

6. The organizational culture is important, (somewhat) changeable, and hard to define, let alone study. But most of us feel it is a significant part of the problem. Specifics will be discussed below. In general, the ethos is driven by current intelligence with the attendant incentives to write for the PDB (the establishment of the DNI notwithstanding, the PBD remains largely a CIA product). Research is scanted; time on account is limited (although recently efforts have been made to rectify both these deficiencies); analysis is closely tied to reports and sources (also see # ? below); “speculation” is frowned upon; attempts to step back and analyze the “larger picture” are rare; presentations of alternative possible future are more common than are presentations of alternative explanations for others’ behavior. In fact explanations are not the preferred form of analysis, and digging deeply is not encouraged (in part because most intelligence analysis is brief or, in the case of NIEs, covers a great deal of territory.

7. Bad methodology–discussed in detail in my book. Sherman Kent: “The main difference between professional scholars or intelligence officers on the one hand, and all other people on the other hand, it that the former are supposed to have had more training in the techniques of guarding against their own intellectual frailties.” Key problems include:

    • a) Tendency for estimates to be driven by plausibility without anyone realizing this, which means that evidence arguably consistent with the prevailing view is taken as strongly supporting it in sense of contradictory other interpretations.
    • b) Ignoring of hypothetico-deductive method and so not seeing the significance of “negative evidence” and things that don’t occur (dogs that don’t bark). Analysts told me that they did not realize they had fallen into this trap until I pointed it out to them. Barton claims that he was ignored when he argued that if Yellow Rain were the Toxin that CIA claimed, it would have produced numerous eye injuries (p. 21).
    • c) Tendency for individuals and organizations to reach premature cognitive closure. This happened with the aluminum tubes.

8. Over-reaction to Iraq and greater hesitation to go beyond/against reports from the field. Of course speculation should not be unrestrained, but neither should we take seriously the Senate Select Committee’s admonition that intelligence should not go beyond specific reports. That is what intelligence is all about. Perhaps “B&A” (background and analysis) was over-used and became a justification for sloppy thinking, but it is how intelligence earns its keep.

9. Most review is done by (too many) managers up the line; review by peers remains insufficient; outsiders (not only academics) are rarely seriously involved (not that they have the answers).

10. Inhibitions against analyzing US policy can be a major problem. Of course intelligence should not second-guess policy, but when others’ behavior is based in significant measure on what they think the US has done or will do, then intelligence must take US policy into account.

11. There are particular problems with the age/experience distribution of analysts today. After the end of the Cold War there was little recruiting; after 9/11 there was a great deal. Now a large percentage of analysts are inexperienced, and there are too few skill and experienced ones to manage, monitor, and mentor them.








<<== back to main conference page

share this entry: