A Palestinian man runs in an area damaged in an Israeli airstrike in Gaza City in the Gaza Strip on July 24. (Hatem Moussa/AP)
A month ago, three Israeli teenagers were found murdered near Hebron, presumably by Palestinians. Then a Palestinian teenager was found murdered in Jerusalem, presumably by Israelis. Within days, war had broken out between Hamas and Israel. Was this conflict predictable? In a general sense it seems almost inevitable. Hamas and the state of Israel have long vowed to destroy each other, and their resumption of violence can hardly be considered a surprise. But the timing of the conflict, sparked by murder and revenge – how could that be predicted?
Some political scientists believe it can – that statistical analysis can forecast conflicts months, even years, in advance. This field of political forecasting has grown into a cottage industry, with scholars opening consulting businesses and joining government projects like the CIA’s Political Instability Task Force, whose goal is “to develop statistical models that can accurately assess countries’ prospects for major political change and can identify key risk factors of interest to U.S. policymakers.”
This ambition lays claim to one of the original visions of modern social science – Auguste Comte’s 19th century dream of a “social physics” that would “enquire into the present, in order to foresee the future, and to discover the means of improving it.” Historic events like the escalation of conflict and the achievement of peace, in the view of political forecasters, are just as predictable as more routine phenomena like election results or traffic patterns. They all obey the laws of political and social life, analogous to the laws of the natural world – or do they?
I explored this question with my colleague Aseem Hasnain in a recently published paper examining Israeli-Palestinian interactions from April 1979 through December 2009. We found that sophisticated vector-autoregression (VAR) models predicted routine events fairly accurately, but were far less accurate in predicting historic episodes like wars, uprisings and peace accords – the very events that political forecasters are most eager to anticipate.
Spikes in prediction errors, shown in the graph below, are particularly visible at the outbreak of the First Intifada in 1987, the Madrid Conference of 1991 (especially for Israeli actions), the Oslo Accord of 1993 (especially for Palestinian actions), the Second Intifada in 2000, and the Israeli-Hamas war of 2006.
This graph plots squared prediction errors for in-sample daily forecasts, using machine-coded news reports from the Reuters news service. Similar patterns emerge with dichotomized variables and other operationalizations, and with forecasts of fatalities from a separate data source.
Many of those historic moments involved “structural breaks,” a technical term that indicates shifts in the underlying parameters of the statistical model. These shocks to the system could not be extrapolated from prior data – they could only be identified as they occurred. All of this suggests that major historic events may not obey the same laws as the more routine events that precede them. Instead, major events can dissolve seemingly permanent laws of political and social life, initiating new patterns of interaction, for better or for worse.
Complicating matters further, four distinct types of shocks seem to have occurred at different moments in Israeli-Palestinian history. Each element of the statistical model can shift without warning: An “effect shock” to the coefficients, “input shock” to the independent variables, “lag shock” affecting the optimal lag structure, and “agent shock” affecting the number of actors in the system.
In earlier work, I explored a similar phenomenon with qualitative methods, using oral histories and contemporaneous accounts of the Iranian Revolution of 1978-79 to examine how the urge to revolt crept up on participants as protests began to multiply. Many respondents reported that watching others risk their lives changed their aspirations and calculations. They felt that the old rules of routine social and political life no longer applied. When this sense of novelty becomes widespread, it can erase aspects of prior patterns of interaction, catching everybody by surprise. That is what happened during the Iranian Revolution. It happened again during the “Arab Spring” uprisings of 2011.
These momentous breaks from routine mark the limits of social-scientific knowledge. They stubbornly resist domestication in social-scientific models. What remains, I have proposed, is to study the experience of wildness. What does it feel like to live through such moments, to participate or avoid participation, to make history?
“I saw my friend in the street shouting, ‘Death to the shah,’ and my fear left me,” one young Iranian told me when I was researching the Iranian Revolution. Staying away from protests felt emotionally impossible, as another young protestor told his wife before heading out to a demonstration: “I’m no different from the others, am I? My blood’s no redder than theirs, is it?” In a lighter vein, humorist Hadi Khorsandi recalled, “I was tolerating the presence of people whom I had hated all my life. Why? The first reason was that I could see that people who had hated me all their lives were smiling at me and were tolerating my presence.”
Even as the old order appeared to be collapsing, nobody could tell whether it might recover. “Everyone is asking everyone what is going to happen,” the Iranian security police reported. People sought out conversations with strangers to gather evidence on who was planning to protest and opinions on what might ensue. “Whenever two or three people got together, they would start a discussion,” author Mahmud Golabdareh’i recalled in his memoirs. “Words, views, advice all differed from top to bottom. It was unclear what would happen.”
At moments of historic change, Israelis and Palestinians also struggled to understand whether old patterns of interaction still held or new ones were forming. In the early days of the First Intifada, for example, the Israeli defense minister reassured a colleague that “the army will assert control very quickly,” while a foreign ministry official drew the opposite conclusion that “this was the beginning of something big.” Palestinian leaders were also disconcerted by the uprising. The head of intelligence for the Palestinian Liberation Organization recalled that “when the intifada broke out, we were at first afraid” that it would amount to nothing, because “nobody had been calculating on such an intifada, with its force and power. The one who was most in touch with the occupied territories was Abu Jihad [Khalil al-Wazir], but even he didn’t expect it to be like that.”
Acknowledging the emotional, confusing experience of these moments runs counter to the enterprise of political forecasting. Reviewers of our paper on Israeli-Palestinian interactions noted this as well. One complained that the findings were “unsatisfying” and not “optimistic” enough, and said “there must be some other useful conclusions/directions for forecasting analysts” that the paper could have explored. Another reviewer said there was “no benefit” in pointing out shocks that disrupt statistic models unless the paper “go[es] on to correct for these effects.” Another said that because the paper “does not offer steps to improve the forecast accuracy of models,” it flouts “the well-known first rule of wing-walking. You don’t let go of something until you have something else to hang on to.” This colorful adage about wing-walking encapsulates the daredevil spirit of political forecasting. Enshrined in a textbook on international relations by political scientist Bruce Bueno de Mesquita, the self-proclaimed “predictioneer,” the metaphor implies that forecasters believe they are flying. Our evidence suggests, by contrast, that they are standing on the ground with the rest of us.
Charles Kurzman is a professor of sociology at the University of North Carolina at Chapel Hill and co-director of the Carolina Center for the Study of the Middle East and Muslim Civilizations.
