Popular Science Monthly/Volume 55/September 1899/Influence of the Weather Upon Crime


THE relation between general climatic conditions and the prevalence of suicide has been somewhat exhaustively studied by students of criminology, the result being a considerable accumulation of data and the formulation of a number of more or less tenable theories. From these studies we may safely conclude that the homicidal tendency, as shown by self-destruction (suicide) and the destruction of others (murder), is stronger in the temperate climatic zones than in the torrid or frigid, and that in the late spring and early summer months more of these offenses have been recorded than for any other period of the year. To these few facts the seeming effects of cosmical forces upon such tendencies has apparently been limited.

It fact, it was the oft-repeated statement that nothing was known of the exact relations of the more definite meteorological conditions with the prevalence of suicide-—a statement to be found in most treatises upon the subject—that has given rise to this paper. Realizing that the science of climatology must include, and in fact be based upon, a study of the meteorological conditions prevalent, and that the study of these definite conditions for the exact times when suicides or murders occurred might throw some light upon the question, this problem was undertaken.

In the preparation of the accompanying charts, from the study of which the conclusions herein stated were deduced, the record of crime for Denver, Colorado, for the fourteen years ending with June, 1897, was made use of. Superintendent Howe, chief of the city detective service, has kept such a record with the greatest care, and we wish here to acknowledge the many courtesies of his office.

No attempt has been made in this paper to compare the conditions for Denver, either meteorological or social—and each is somewhat unique—with such conditions elsewhere. In fact, such a comparative study is at present impossible since data are wanting.

In the actual preparation of the charts each murder, suicide, or attempt at suicide—which, for our purpose, is equally important—was set down chronologically in the left-hand columns of large sheets of paper ruled for the purpose. These sheets were then taken to the office of the United States Weather Bureau, F. H. Brandenburg, director, where were recorded in the proper columns the maximum and minimum barometer readings, maximum and minimum temperature, maximum and minimum humidity, maximum velocity of the wind, precipitation, and character of the day for each day during the fourteen years on which a crime of either class occurred. When several took place upon the same day the fact was taken into consideration. From the sheets thus filled out, the curves on the accompanying charts were plotted by computing the

PSM V55 D674 Graphs of the influence of weather on crime.png
Fig. 1.

per cent of crimes of each class committed under the definite meteorological condition indicated.

The curves marked "normal" were constructed by tabulating in a similar manner the conditions for every day in a sufficient number of days to secure a fair average. Five years were so tabulated for Figs, 2, 3, 4, and 5, and the records for nineteen years used in Figs. 1 and 6.

The whole number of suicides recorded is two hundred and sixty; murders, one hundred and eighty. It may be noted that this number of suicides, for a city averaging hardly one hundred thousand inhabitants for the fourteen years, is largely in excess of the rate recorded for American cities, but it must be remembered that some of these were unsuccessful attempts, and also that the social conditions of Denver tend to swell the number—containing, as it does, so many disappointed in the last struggle for health.

Fig. 1 shows the occurrence, in per cent, of crimes of both the classes considered for each month of the year, together with the monthly meteorological means, computed from the records for nineteen years. The expectancy curve in the occurrence table is based upon the supposition that the months of the year are all of the same length, and that the numerical expectancy would be one twelfth, or eight and a third per cent for each. It will be seen that the crime curves are for the most part below the expectancy for the winter months, and above it for the summer (except for April, and suicides for June), showing the maximum for the latter class in May and for murders in March. Morselli shows[1] that for most European countries suicides are at the maximum in June, though a considerable number show that condition for the later spring months. A study of the general meteorological means, shown upon the same plate as the occurrence table, fails to indicate any good reason for irregularity of the crime curves. The "month" columns read from the top to the bottom of the chart, and by following that for May, for instance, which month shows the maximum for suicide, we find that the meteorological condition for each class of data is about halfway between the extremes for that class for the year, while for January (minimum suicides) each class is by far more divergent. Yet a mean, like those considered in this table, is but the average of the extremes, and those months which show great per cents of crime also present great extremes of condition, which fact, interpreted in the light of those disclosed by the charts yet to be considered, make the occurrence curve more explicable.

Wind.—An explanation of the various curves in Fig. 2 may serve for the series following, so I give it somewhat in detail. The vertical distances from the base line indicate per cents, and the distances from left to right, divided into columns, the maximum velocity of the wind per hour for the days tabulated. In the "normal" curve every day for five years was considered, and it was found that seven per cent of the days for that period showed a maximum velocity of between one and ten miles (first column), forty-eight per cent a maximum velocity of between ten and twenty miles (second column), nineteen per cent a maximum velocity of between twenty and thirty miles, and so on, as indicated by the curve. Now, it can readily be seen that this normal curve may also be considered the expectancy curve—if the wind has no effect. That is, if forty-eight per cent of the days of the year show a maximum velocity of the wind, between ten and twenty miles an hour, PSM V55 D676 Relationship of wind velocity and crime rates.pngFig. 2. the law of probability would give us the same per cent of the crime for the year on such days if this meteorological condition were not effective. What we do find, however, is indicated by the other curves, and any increase of crime over expectancy may in this case be ascribed to the wind. We notice that for slight velocities (one to twenty miles an hour) the crime curves are below that of expectancy, but we can see that if the sum of all the per cents for any one curve is one hundred, and one is forced above the other at any part, there must be a corresponding deficiency at some other part. So we may, perhaps, with justice suppose that these mild velocities do not exert a positively quieting effect emotionally, but simply a less stimulating effect than the higher ones. For velocities of between twenty and thirty miles a marked effect is noticeable, and under those conditions the proportion of suicides to that expected is 37:29; velocities of from thirty to forty miles, 14:11; of forty to fifty miles, 7:2; of fifty to sixty miles, 0.4:2.6; of fifty to sixty miles, 0.2:2. The curve for murders shows the increase to be slightly less than for suicides, but the same general relation is preserved throughout. The value of such curves is, of course, somewhat proportional to the number of observations made and recorded, and we must confess that two hundred and sixty (suicides) and one hundred and eighty (murders) is a hardly sufficient number from which to deduce a definite law, but we can hardly doubt, even considering this somewhat limited number, that the wind is, in our problem, a factor of no mean importance.

Temperature.—Fig. 3 is intended to show, in a similar manner, the relation between expectancy curves, based upon conditions of temperature, and the actual occurrence of the crimes in question. With this class of data, as well as that for the barometric readings and humidity (Figs. 4 and 5), both the maximum and minimum readings are considered. This was done instead of taking the mean of both for the day, since in many cases the latter might be quite normal, while one or possibly both the former might exhibit marked peculiarities. All the curves were constructed precisely as in the chart just considered, and those marked "normal" are again the expectancy curves. An inspection of the chart shows no marked discrepancies till we reach the higher temperatures. For the lower the coincidence for all the maximum and all the minimum curves is not exact, but somewhat similar. When, however, we reach for the minimum curves, temperatures of from 40° to 50° and from 50° to 60°, which means that for the per cent of days indicated, the temperature did not go below those points, the per cent of crime exceeds that expected under the conditions in the proportions of 22:16.5 and 24:18 (suicides), and 21:16.5 and 29:18 (murders).

The same general relation exists between the maximum curves, where it is shown that for temperatures between 80° and 100° the actual crime is about thirty-three per cent in excess of the expected.

PSM V55 D677 Relationship of temperature and crime rates.png
Fig. 3.

These facts have their bearing upon the already noted statement that the summer months show a preponderance of homicide.

Barometer.—Fig. 4, disassociated from the others, shows but little. Naturally we should not look for very marked effects from variations of an inch or less in the barometric readings, when in the course of a journey from the sea level to Denver a change of six inches is brought about, and in going from the same point to the summit of Pike's Peak one of nearly twelve inches without producing any marked emotional abnormities, but we must take into consideration the fact that sudden barometric variations generally accompany or more frequently precede other important meteorological changes. In the latter case, though they might be the primary cause of factors considered in this study, they themselves could fail to show upon the tables.

Humidity.—This figure (Fig. 5) indicates in a very decisive manner that states of lower relative humidity, as shown by both maximum and minimum readings, are conducive to excesses in both the

PSM V55 D678 Relationship of barometric pressure and crime rates.png
Fig. 4.

classes of crimes studied. For instance, for maximum humidities between ten and twenty the proportion of actual crime to that expected is 1: 0.1; between twenty and thirty (suicide), 11: 1; between thirty and forty, 9.5: 4.5; between forty and fifty, 15: 8. The maximum curves show somewhat the same general relation though not with quite so marked divergences. To one who has experienced the general low humidities of our Colorado altitudes

PSM V55 D678 Relationship of humidity and crime rates.png
Fig. 5.

(Denver is one mile above the sea level) this result is not surprising. There is no doubt that a nervous tension much in excess of that common in the lower altitudes exists, due in part, perhaps, to the deficiency in barometric pressure and a consequent effect upon the respiratory processes, but probably, as shown by these curves, more largely to the dryness of the atmosphere, as indicated by low humidity. I hope at some future time to verify or disprove this supposition by a comparative study made at some lower altitude.

Character of the Day.[2]—Fig. 6 shows the relation between the expectancy of crime, based upon the actual per cents of cloudy, partly cloudy, and clear days (records of nineteen years), and its actual occurrence. The disagreements are very slight, although a slight excess of murders is shown for cloudy days.

Summary.—Fig. 1 shows at a glance no generally prevailing meteorological conditions to which can be ascribed, with any degree of certainty, the monthly variations of crime.

Fig. 2 shows that high velocities of wind seem to increase to a marked extent the tendency to crime. For the highest velocities increasing the probability twenty times (two thousand per cent). Fig. 3 shows that high temperatures seem to have the same effect, that of between 90° PSM V55 D679 Murder and suicide rates and the daily weather.pngFig. 6. and 100° increasing the probability one hundred per cent.

Fig. 4 fails to show that barometric changes are accompanied by any marked excesses in crime.

Fig. 5 shows that low conditions of relative humidity are attended with very marked excesses, those below thirty increasing the probability of suicides eleven times (eleven hundred per cent).

Fig. 6 fails to show that the character of the day has any considerable effect.

Considering briefly, in conclusion, the results of the foregoing study, and comparing them with a somewhat similar one for children,[3] we may safely conclude that the tendency to homicide varies with those meteorological conditions which bring about an emotional state necessitating a considerable discharge of motor stimulus. The same conditions which bring about irritability and unruliness on the part of the child accompany suicidal tendencies.

This supposition is upheld by the fact that suicide is less common in the colder climates, where the metabolic processes are slow, and in the torrid zone, where the heat produces a general depletion of energy for motor discharge, than in the temperate regions, where the climate is exhilarating. The study, from the social standpoint, too, leads us to the same conclusion. The excess of crime in the social whirlpools of our great cities is convincing, and especially the careful study made by Morselli of the prevalence of suicide in the different countries of Europe, interpreted in the light of what we know of their social conditions.

Yet, in considering the facts disclosed by the present paper, we must not dogmatically assert that each is of the importance that the figures indicate. In fact, it seems evident from a careful study of the sheets, which show all the conditions together for the same day—a thing impossible with the charts illustrating this paper—that the various conditions for the day mutually react and interact upon one another, certain combinations seemingly resulting in a re-enforcement of the tendency to crime, while certain others inhibit it. Space forbids any full discussion of this phase of the problem in the present paper, but it very probably will be made the subject of some future study.

Author's Note.—The above paper was written more than a year ago. Since that time the work of comparing the prevalence of crime with the meteorological conditions has been carried on upon a much larger scale in the city of New York. An immensely greater number of data have served to corroborate the earlier conclusions arrived at in this Denver study, only in minor points—and those directly traceable to the very different climates—proving at all in opposition to them.—New York, July 1899.

  1. Suicide, International Science Series.
  2. By the United States Weather Bureau days are characterized as "cloudy" when for 0.8 or more of the possible hours of sunshine the sun is obscured; "partly cloudy" when from 0.4 to 0.7 inclusive is obscured; and "clear" when 0.3 or less.
  3. See The Child and the Weather, Pedagogical Seminary, April, 1898.