When Forecasts Fail
When Forecasts Fail
Charles Kurzman and Aseem Hasnain, “When Forecasts Fail: Unpredictability in Israeli-Palestinian Interaction,” Sociological Science, 1:239-259, June 2014. “This article explores the paradox that forecasts may be most likely to fail during dramatic moments of historic change that social scientists are most eager to predict. It distinguishes among four types of shocks that can undermine the predictive power of time series analyses: effect shocks that change the size of the causal effect; input shocks that change the causal variables; duration shocks that change how long a causal effect lasts; and actor shocks that change the number of agents in the system. The significance of these shocks is illustrated in Israeli–Palestinian interactions, one of the contemporary world’s most intensely scrutinized episodes, using vector autogression analyses of more than 15,000 Reuters news stories over the past three decades. The intervention of these shocks raises the prospect that some historic episodes may be unpredictable, even retrospectively.” More…
Supplemental materials:
File name: SocSci_v1_239to259_supp.pdf
Replication data:
Data file name: When_Forecasts_Fail_forecast_data.txt
Data file format (79 comma-separated variables, 11,218 observations):
date: Stata format, 7044 (April 15, 1979) to 18261 (December 31, 2009)
saturday: 1 if Saturday, 0 otherwise
top13: 1 to 13 on date of onset of major historical episodes, as determined by history textbooks, 0 otherwise
ip: Israeli actions toward Palestinians, -21 to +7
ipgoldstein: Israeli actions toward Palestinians, using Goldstein scale, -68.7 to +32.8
ipviolence: 1 if Israeli violence toward Palestinians, 0 otherwise
ipverbal: Israeli verbal actions toward Palestinians, -4 to +8
ipmaterial: Israeli material actions toward Palestinians, -8 to +3
ipbtselem: Israeli killings of Palestinians, B’Tselem data, 0 to 357
ipbbinary: 1 if Israeli killings of Palestinians, B’Tselem data, 0 otherwise
ipbcapped: Israeli killings of Palestinians, capped at 71, B’Tselem data, 0 to 71
ip_insample_prediction: In-sample vector autoregression prediction of Israeli actions toward Palestinians, -9.32 to +1.90
ipgoldstein_insample_prediction: In-sample vector autoregression prediction of Israeli actions toward Palestinians, using Goldstein scale
ipviolence_insample_prediction: In-sample vector autoregression prediction of binary of Israeli violent actions toward Palestinians
ipverbal_insample_prediction: In-sample vector autoregression prediction of Israeli verbal actions toward Palestinians
ipmaterial_insample_prediction: In-sample vector autoregression prediction of Israeli material actions toward Palestinians
ipbtselem_insample_prediction: In-sample vector autoregression prediction of Israeli killings of Palestinians, B’Tselem data
ipbbinary_insample_prediction: In-sample vector autoregression prediction of binary of Israeli killings of Palestinians, B’Tselem data
ipbcapped_insample_prediction: In-sample vector autoregression prediction of Israeli killings of Palestinians, capped at 71, B’Tselem data
ip_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli actions toward Palestinians, -8.64 to +1.59
ipgoldstein_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli actions toward Palestinians, using Goldstein scale
ipviolence_outofsample_predicti: Out-of-sample vector autoregression prediction of binary of Israeli violent actions toward Palestinians
ipverbal_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli verbal actions toward Palestinians
ipmaterial_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli material actions toward Palestinians
ipbtselem_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli killings of Palestinians, B’Tselem data
ipbbinary_outofsample_predicti: Out-of-sample vector autoregression prediction of binary of Israeli killings of Palestinians, B’Tselem data
ipbcapped_outofsample_predicti: Out-of-sample vector autoregression prediction of Israeli killings of Palestinians, capped at 71, B’Tselem data
bayesmeanip: mean of Bayesian forecasts of Israeli actions toward Palestinians
bayesmedianip: median of Bayesian forecasts of Israeli actions toward Palestinians
bayessdip: standard deviation of Bayesian forecasts of Israeli actions toward Palestinians
bayesquantileip: quantile of observed Israeli actions toward Palestinians among Bayesian forecasts
pi: Palestinian actions toward Israelis, -19 to +6
pigoldstein: Palestinian actions toward Israelis, using Goldstein scale, -58.5 to +37.7
piviolence: 1 if Palestinian violence toward Israelis, 0 otherwise
piverbal: Palestinian verbal actions toward Israelis, -5 to +10
pimaterial: Palestinian material actions toward Israelis, -6 to +2
pibtselem: Palestinian killings of Israelis, B’Tselem data, 0 to 27
pibbinary: 1 if Palestinian killings of Israelis, B’Tselem data, 0 otherwise
pibcapped: Palestinian killings of Israelis, capped at 71, B’Tselem data, 0 to 27
pi_insample_prediction: In-sample vector autoregression prediction of Palestinian actions toward Israelis, -5.17 to +1.49
pigoldstein_insample_prediction: In-sample vector autoregression prediction of Palestinian actions toward Israelis, using Goldstein scale
piviolence_insample_prediction: In-sample vector autoregression prediction of binary of Palestinian violent actions toward Israelis
piverbal_insample_prediction: In-sample vector autoregression prediction of Palestinian verbal actions toward Israelis
pimaterial_insample_prediction: In-sample vector autoregression prediction of Palestinian material actions toward Israelis
pibtselem_insample_prediction: In-sample vector autoregression prediction of Palestinian killings of Israelis, B’Tselem data
pibbinary_insample_prediction: In-sample vector autoregression prediction of binary of Palestinian killings of Israelis, B’Tselem data
pibcapped_insample_prediction: In-sample vector autoregression prediction of Palestinian killings of Israelis, capped at 71, B’Tselem data
pi_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian actions toward Israelis, -5.69 to +0.96
pigoldstein_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian actions toward Israelis, using Goldstein scale
piviolence_outofsample_predicti: Out-of-sample vector autoregression prediction of binary of Palestinian violent actions toward Israelis
piverbal_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian verbal actions toward Israelis
pimaterial_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian material actions toward Israelis
pibtselem_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian killings of Israelis, B’Tselem data
pibbinary_outofsample_predicti: Out-of-sample vector autoregression prediction of binary of Palestinian killings of Israelis, B’Tselem data
pibcapped_outofsample_predicti: Out-of-sample vector autoregression prediction of Palestinian killings of Israelis, capped at 71, B’Tselem data
bayesmeanpi: mean of Bayesian forecasts of Palestinian actions toward Israelis
bayesmedianpi: median of Bayesian forecasts of Palestinian actions toward Israelis
bayessdpi: standard deviation of Bayesian forecasts of Palestinian actions toward Israelis
bayesquantilepi: quantile of observed Palestinian actions toward Israelis among Bayesian forecasts
il: Israeli actions toward Lebanese, -16 to +5
ilgoldstein: Israeli actions toward Lebanese, using Goldstein scale, -53.4 to +22.9
ilviolence: 1 if Israeli violence toward Lebanese, 0 otherwise
ilverbal: Israeli verbal actions toward Lebanese, -3 to +4
ilmaterial: Israeli material actions toward Lebanese, -5 to +2
li: Lebanese actions toward Israelis, -12 to +5
ligoldstein: Lebanese actions toward Israelis, using Goldstein scale, -37.0 to +22.0
liviolence: 1 if Lebanese violence toward Israelis, 0 otherwise
liverbal: Lebanese verbal actions toward Israelis, -3 to +3
limaterial: Lebanese material actions toward Israelis, -4 to +2
pl: Palestinian actions toward Lebanese, -8 to +4
plgoldstein: Palestinian actions toward Lebanese, using Goldstein scale, -24.0 to +18.0
plviolence: 1 if Palestinian violence toward Lebanese, 0 otherwise
plverbal: Palestinian verbal actions toward Lebanese, -2 to +3
plmaterial: Palestinian material actions toward Lebanese, -3 to +1
lp: Lebanese actions toward Palestinians, -7 to +3
lpgoldstein: Lebanese actions toward Palestinians, using Goldstein scale, -22.0 to +10.0
lpviolence: 1 if Lebanese violence toward Palestinians, 0 otherwise
lpverbal: Lebanese verbal actions toward Palestinians, -2 to +2
lpmaterial: Lebanese material actions toward Palestinians, -2 to +1