IV. The Data
First, there is no question that very extreme weather conditions affect crime patterns, just as they affect all other human activities. If a hurricane strikes a city with 100-mph sustained winds, burglary will go down during the hours that those winds are present. This is not because there are no motivated burglars in the city, or because there are no unprotected homes or businesses with valuables in them (in fact, there are probably more unprotected homes, because individuals with resources may have evacuated the area). The simple fact is that when it is impossible to walk on a street, the burglars cannot get to the homes. However, in a study of the impact of Hurricane Hugo, James LeBeau found that, once the hurricane had passed, there was a significant increase in calls to police for burglary as well as to report a “man with a gun,” suggesting a possible increase in defensive gun use (LeBeau, 2002). When the motivated offender is able to move about, when suitable targets are available, and when the activities of guardians such as the police are directed elsewhere, crime increases.
In general, criminologists do not look at such extreme weather events. Instead, they conduct their research on the range of normal variations in weather factors, seasonality, or climate, and look for changes in crime patterns that relate to changes in those factors.
Although variations in climate and its effect on people served as explanations of differing crime rates in much of the early literature on weather and crime, the impact of climate on crime has been largely discounted. The earliest observations that led criminologists to suggest climatic impacts on crime were geographic differences in crime rates. In the United States, this was in particular the consistently higher murder rates found in the South. It is a fact that murder rates in the South have been higher than in any other region of the United States since data on crime have been collected. Examinations of the correlation of the South with homicide rates has became progressively sophisticated over the past hundred years, and a significant debate has developed in criminology as to whether the association of murder and “Southern-ness” is due to cultural differences (particularly among minority populations) or to structural differences along economic lines. What is significant, however, is that in the dozens of scientific articles published on this question since the 1960s, climate is no longer considered as a possible explanation.
Research conducted by DeFronzo (1984) near the end of the 20th century may have effectively laid the climate and crime argument to rest. After controlling for nonclimatic variables, climate had only weak and indirect associations with crime rates. DeFronzo found that economic conditions, urbanization, and population demographics remain the primary predictors of overall crime rate (again, with the debate continuing over exactly which of the three carries the primary explanatory power).
From this, the research shifted initially toward seasonal effects on crime. Then, as computers gave us the ability to do increasingly complex research on increasingly large databases, criminologists turned to the examination of more precise, short-term weather factors.
Quetelet’s thermic law of delinquency argued that heat and violent behavior were related, such that violent crimes should be higher in the summer months and property crimes higher in the winter. A casual examination of Uniform Crime Reports data indicates that there is a considerable seasonal effect for both violent and property crime in the United States. However, when one looks at the Uniform Crime Reports for 1990 through 2003, it is obvious that the months in which the property crimes of burglary and larceny are highest are not in the cooler seasons but in July, August, September, and June (Hipp et al., 2004). Thus, a simple examination of the data challenges the thermic law. Research in other locations has also found seasonal patterns for particular crimes, although some of these patterns are not the same as those found in the United States. In England and Wales, robbery and burglary both increase in the winter, whereas personal crimes peak during the summer; in Ghana, the personal crime of assault is also highest between June and September. Landau and Fridman (1993) examined the seasonality of robbery and homicide in Israel and found that robbery followed a strong seasonal pattern of higher rates in the winter but that homicide, although somewhat more common during certain months, displayed no seasonal pattern (August was high, but July was low; March was high, but April was low). In the Southern Hemisphere, where the seasons are reversed, the patterns also are reversed. Studies of sexual abuse in Chile indicated that the months in which the number of cases were highest were November, October, and December (late spring and early summer in Chile), with the lowest number of cases appearing in late autumn and winter during the months of May, June, and July (Tellez, Galleguillos, Aliaga, & Silva, 2006).
Indeed, in the United States, a temporal pattern in which certain months are significantly high and certain other months significantly low is apparent in almost all crimes. The problem is to determine whether the pattern variation is seasonal or monthly. This may sound like the same thing, but there is a difference, and it is a very important difference for theory, research, and policymakers. If one looks at American data over the past five decades, one can see that a number of crimes are highest in the warmer months of June, July, August, and September (Hipp et al., 2004). Larceny (theft), burglary, aggravated assault, and rape all have peaks in July and August, with the third highest month most commonly being either June or September. So, looking at American data we see a seasonal pattern for rape, assault, larceny and burglary, with all being higher during summer months. Because two of these are personal crimes and two are property crimes, they obviously do not support Quetelet’s thermic law of delinquency.
Robbery and murder, on the other hand, show significant monthly patterns, but the months involved do not appear in any one season. July and August are also among the three highest months for murder in the United States, but the third most common month for high murder rates is December. This indicates a significant monthly pattern for homicide, yet it is clearly not a seasonal pattern. Also, robbery, which is considered a personal crime by the Uniform Crime Reports, is consistently at its highest during the months of December, January, October, and August (Cohn & Rotton, 2000).
What these data lead us to suspect, and research confirms, is that in any country the seasonal or monthly patterns characteristic of any crime are determined in large part by cultural patterns. In one of the most pronounced cultural effects, Zimring and associates (Zimring, Ceretti, & Broli, 1996) discovered that crime of all types drops by half in Milan, Italy, during the month of August, the month in which a large proportion of the Italian population goes on holiday. They noted that the opportunities for crime (suitable targets and the absence of capable guardians) do not decrease; in fact, there should be more unguarded homes and fewer guardians in place. The only conclusion that seems to fit their data is that “social processes unknown in American cities reduce criminal activity in Milan almost in half during the vacation month of August . . . Crime takes a holiday in Milan during August apparently because criminals take a holiday” (Zimring et al., 1996, p. 277).
In sum, the relationship between the hotter months of summer and a peak in rape and assault seems to be almost universal. However, although there is no such universal seasonal pattern for property crimes, robbery, or murder, many locations do show a pronounced monthly pattern for those crimes. Oddly, it appears that the thermic law of higher personal crimes in the summer and higher property crimes in the winter may hold for other locations, including parts of Europe (Rotton & Cohn, 2002, p. 487), but not for the United States.
When we examine the data on season and specific crimes in more detail, the results strengthen the idea that the seasons act on crime by bringing about changes in routine activities and increases in interactional stress. Recall that July and August are the most common months for murder but that the third month in which murder is most likely to occur is December. Obviously, July, August, and December have drastically different weather patterns. However, these are the months in which we tend to take vacations as well as the ones during which we interact more frequently with friends and family. As many of us are aware, those people who can cause us the most stress— who can really push all our buttons—are the same people to whom we are closest. These months also see an increase in alcohol consumption. Alcohol releases inhibitions and can increase aggression, and it is a common drug of abuse among young males, who constitute the group with the highest murder rates (as both offenders and as victims). Increased interactions among people with strong emotional ties, which produce increased stress, and the increased use of alcohol by high-risk groups during the summer and over the holidays should be expected to increase the incidence of homicides, and that is what we see. It is not the weather characteristic of the season but the nature of social interactions that are influenced and changed by that weather.
This proposition is further supported when one compares seasonal patterns of murder and assault. Criminologists frequently make the argument that aggravated assault and homicide are “sibling” crimes; that is, they are the same behavior—an attack by one person with the intent to do serious bodily harm to another—and the only real distinction is whether the victim lives or dies as a result of the attack. If this is correct, then we would expect to see the same pattern of July, August, and December being the most common months for aggravated assault, just as we did for murder. However, that is not the case. The months with the highest reported cases of aggravated assault are July and August, but then June and September tie for third. Not only is December not one of the highest, but also it is consistently the month in which the lowest number of aggravated assaults are reported. On the one hand, that could mean that murder and aggravated assault are not the same behaviors at all, but it is also possible that assaults occurring around Christmastime are less likely to be reported because, as our theories suggest, they are more likely to occur between friends or members of an extended family. And that is exactly what happens when criminologists look at the reports made to police. The reporting of assault goes down near the holidays, often because fights between friends or family are hidden, but fights that result in a death cannot be hidden, and the number of murders increases. As Anderson noted as early as 1989, “It is probably the case that within families, assaults are relatively unlikely to be reported to the police. Obviously, within family [or within-friend] homicides cannot be correspondingly underreported” (p. 84).
In sum, it is not the weather characteristic of the season that directly increases aggression in individuals; instead, it is the social behavior characteristic of the season for each culture that changes the probability that criminal behavior may result, and this seems to apply across cultures. Most of the findings on the subject of weather and crime are based on American data, but research conducted in other nations suggests that although seasonal and weather effects on crime appear to be universal, the form they take is shaped by unique cultural patterns.
Although the results of research testing the impact of the full range of weather variables vary from study to study, there is fairly solid evidence for a relationship between crime and temperature, with lesser support for such a relationship between crime and humidity, precipitation, or changes in barometric pressure. A few isolated studies have found some impact of cloudiness, precipitation, or barometric pressure on crime, but most research using those variables does not. Only temperature seems to produce relatively consistent findings over the years (Rotton & Cohn, 2002, provided a good review of this research). Furthermore, temperature has been examined in a number of different ways. Basic raw temperature, as well as the discomfort produced by the addition of humidity in the temperature humidity index (originally called the discomfort index), have been tested. The temperature recorded at periods of time as short as 3 hours has been examined, as has the effect of consistently high temperatures over a number of days. Regardless of how it is tested, in one study or another, an increase in temperature has been found to correlate with increases in assault, homicide, rape, robbery, burglary, larceny, and domestic violence.
Studies that have examined the combination of weather effects and time effects on crime patterns have found that they are related. In line with routine activities theory, street crimes are consistently higher during weekends, when people engage in more leisure activities, have more time on their hands, are more likely to use alcohol, and are more likely to leave their homes for other entertainment venues. As a result, weekends are more likely to reduce guardianship for crimes such as larceny and burglary and are more likely to place demographic groups with higher propensities for aggressive crime (young males) in entertainment situations involving alcohol and in contact with suitable targets for aggression (anything from other young males to young females), with a concomitant increase in murder, aggravated assault, rape, and robbery. Temperature not only has a general impact on crime, but it also appears to compound or accentuate the impact that day of the week has on crime. Crime consistently increases on weekends, but research shows that it increases more on hotter weekends than on cooler weekends. Taking this even further, LeBeau and Langworthy (1986) made the insightful observation that the increase in both crime and police calls for service during the summer months should thus not be considered unusual, “since vacations from work and school are primarily extended weekends” (p. 139).
So, routine activity models are strongly supported by both seasonal and weather data, but what about stress? Researchers looking at assaults in Dallas during the early 1980s divided neighborhoods into low status, medium status, and high status and found that the link between the discomfort produced by a combination of heat and humidity during the summer and an increase in aggravated assaults during that time was significantly more pronounced in low-status neighborhoods. This fits neatly with the argument that it is the ability to cope with increased stress and discomfort over time that provides a key to understanding the relationship between weather and assaultive crime. Increases in assault were associated with increases in the temperature humidity index across Dallas. All three classes of neighborhoods showed calendar variations, with some increase during summer months and a peak in assaults during the weekend, but the increase was significantly more pronounced in neighborhoods where economic disadvantage limited residents’ options to accommodate increased discomfort. In those low-status neighborhoods, during periods of increased heat, assault increased at a higher rate than in the more affluent neighborhoods because the poor “are less able to control the comfort of their home and work spaces and are perhaps more susceptible to the complex manifestations of heat stress” (Harries, Stadler, & Zdorkowski, 1984, p. 598).
In sum, most factors of weather—rain, snow, fog, weather fronts, barometric pressure, or wind—do not display consistent results when tested for their impact on crimes. Only temperature seems to be related, and the relationship is both robust and consistent across most studies of weather and crime or season and crime. Higher temperatures, or higher temperatures combined with higher humidity, produce such discomfort that our adaptations to stress are stretched to their limits. This discomfort also changes our patterns of routine activities in ways that place us at higher risk of both property and personal crimes. At the societal level, the impact of weather is further mediated by day of the week, the demographic structure and cultural matrix of the population, and the socioeconomic structure of the population. As with so many other things, our cultural, economic, and physical environments modify how we are affected, and how we respond to, everything we encounter.