8/10/2019 Temperature and Riots Paper

1/8

J ou rn al o f P ers on al ity a nd S oc ia l P sy ch ol og y

1 97 9, V ol. 3 7, N o.3, 3 37 -3 44

Ambient Temperature and the Occurrence of Collective Violence:

ANew Analysis

J.

Merrill C arlsm ith and C raig A . Anderson

Stanford Universi ty

Prevalent folklore suggests that riots tend to occur during periods of very hot

w eather. Baron and Ransberger examined 102 major riots in the U nited States

betw een 1967 and 1971 and concluded that the frequency of collective violence

and ambient temperature are curvilinearly related. The present article points

out that the Baron and Ransberger analysis did not take account of the dif-

ferent number of days in different temperature ranges. The artifact is elim-

inated, and the proba bility of a riot, conditional upon temperature, is estim ated.

When this is done, the evidence strongly suggests that the conditional probabil-

ity of a riot increases monotonically w ith temperature. Some general implica-

tions of such data ana lyses are discusse d.

In a recent article in this journal, Baron

and Ransberger (1978) presented an analy-

sis of the re lation ship between the frequency

of major riots and the ambient temperature

occurring during the riots, To do this, they

studied 102 major riots in the United States

betw een 1967 and 1971. The hypothesis they

wished to test, and for which they claimed

confirmatory evidence, is the existence of a

curvilinear relationship between the likeli-

hood of a riot and the maximum ambient

temperature at the time of the riot. This

hypothesis contra sts w ith the prevale nt folk-

lore that riots tend to occur during periods of

very hot weather. Specifically, Baron and

The activities reported herein. were supported in

part by funds from Boys Town. However, the

opinions expressed or the policies advocated do not

necessarily reflect those of Boy s Tow n. T he research

was carried out while the first author was a member

of the Boys Town Center for the Study of Youth

Development at Stanford University and while the

second author was supported by a National Science

Foundation Fellowship. W e are indebted to Bradley

Efron, Lee Ross, and Amos Tversky for their com-

ments and suggestions on this work, and to Joy

Davis for her assistance in the seemingly unending

task o f tran scrib in g 5 8,0 00 tem peratu res.

Requests for reprints should be sent to

J.

Merrill

Carlsm ith , D epartm ent of Psychology, S tanford Uni-

v ersity , S tan fo rd , C alifo rn ia 9 43 05 .

Ransberger conclu ded that th e likelihood of a

riot increases with temperature up to the

range of 81 -85 F and then decreases

sharply with fu rther increas es in tempera tu re .

The evidence that they presented to support

this relationship is a frequency d istribution

of the number of riots plotted against tem-

peratu re . This fr equency polygon does indeed

peak in the interval 81 -85 F, falling off

sharply on either side.

We contend that this relationship is an

artifact of the particular way the data were

exam ined and that an appropriate reanalysis

suggests a monotonically increasing func tion

relating the p robab ility of riots and tempera-

ture. Basically, we argue that the Baron and

Ransberger results stem from their having

not taken account of base-rate differences in

temperature. For example, if days in the 81 .-

85 F range are more common than days in

the 91 -95 F range, there may well be more

riots in the former range. There are, after all,

many more opportunities for riots. But an

appropriate analysis may well show that riots

are relatively more common in the higher

temperature range. To be sure, Baron and

Ran sberger did con sid er this possibility, b ut

they rejected it. In our view, their rejection

w as prem ature; w e consider their arguments

and the w eaknesses therein at greater length

below.

Copyr ight 1979 by the Ame ri ca n P sycholog ic al Assoc ia ti on , I nc . 0022- 3514 /79/3703 -0337 00. 75

8/10/2019 Temperature and Riots Paper

2/8

338

J. MERRILL CARLSMITH AND CRAIG A. ANDERSON

18

0

UJ

>-

15