is given in Table 3.5 The data in Table 3, presented in terms of the effect which a one unit increment in gun control will have on homicide rates, indicate that gun control probably has a negative effect on homicide by firearm and total homicide rates. In all five sets of equations the estimated gun control coefficient is negative and in only one case is there more than a fifteen percent chance that the coefficient's sign is positive (see total homicide rate for "States-1965"-Probability of Sign Error).

Table 4 presents the relationship of each of the independent variables to the death and crime rates. The results in Table 4 show that median income and population density are negatively related to homicides by firearm and total homicides; that the percentage of males, the number of police employees and the percentage of nonwhites have strong positive relationships with homicide rates; and that the correlation between education and homicide was negative in 1960 and positive in 1965.

"All relationships reported in this study are conditional because the effects of other demographic variables are taken into account.

Suicide

Table 5 reports the relationship between gun control and suicide by firearm and total suicide rates. For all five equations the sign of the estimated gun control coefficient is negative, significant and sizeable. For four of the five equations there is a less than ten percent probability that the sign is incorrect. In comparison with the results for total homicide, there is a much greater probability that the gun control coefficient of total suicide is negative and of a greater magnitude the estimated gun control coefficients of total suicide are approximately twice as large.

Under the heading Suicide-by-Firearm (SF) in Table 4; it can be seen that the number of years of school completed, median age, the number of licensed hunters, average temperature and the percentage of males have strong positive relations with suicides by firearm while population density has a strong negative relation. There is also an indication that median income is negatively related to suicides by firearm but not to total suicides.

Accidental Deaths by Firearm

Table 6, which shows the relationship between accidental deaths by firearm and gun control, indicates that the estimated gun control coefficients of accidental death by firearm are as significant (in terms of confidence interval and probability of sign error) as the suicide coefficients but only about the size of the homicide coefficients.

35-394 O 70 - 18

The relationship between accidental deaths by firearm and other variables as shown in Table 4 indicates that the only significant negative factor is income. Significant positive factors include the percentage of males, the percentage of non-whites, the number of police employees and the number of licensed hunters.

Aggravated Assaults

In Table 7 the relationship between gun control and aggravated assault is presented. There is a 67% probability that gun control is associated with a lower number of aggravated assaults by firearm (since the probability of sign error is .327), and about a 75% probability that gun control is related to a higher total of aggravated assaults (probability of sign error is less than .25).

Table 4 indicates that low income, a high percentage of nonwhites, a high number of police employees, lower population density and fewer licensed hunters are associated with a higher number of aggravated assaults.

The relation between robbery and gun control is set out in Table 8. The estimated gun control coefficient indicates that one unit of gun control will reduce the total number of robberies by less

than one half robbery per million population per year (or about 100 fewer robberies per year in the U.S.) and that there is only a 56% chance that the coefficient's sign is correct.

Returning to Table 4, robbery is shown to be negatively related to education, population density, licensed hunters and the percentage of males; and positively related to income, the number of police employees and the percentage of non-whites.

Observations

The following observations may be drawn from the data presented in the previous section.

One: The data indicate that gun control legislation is related to fewer total deaths by homicide, suicide and accident by firearm. The estimated gun control coefficient is negative in the five homicide equations, the five suicide equations and both accidental death by firearm equations. Moreover, in seven of these twelve equations the probability of error is less than ten percent and in only one of the equations does it exceed seventeen percent (total homicides-States -1965-31.0%).

The estimated gun control coefficients of total homicide, total suicide and accidental death by firearm for the states in 1960 were -.228, -.389 and -.196, respectively; and for the states in 1965 were -.095, -.286 and -.167, respectively. On the basis of these results it is estimated that one unit of gun control saves between .548 and .813 lives per million population per year. Thus it can be estimated that the gun control legislation of New Jersey (39 units) saves between 21 and 32 lives per million population per year. On a nationwide basis such legislation would save between 4200 and 6400 lives per year 57

The estimates on lives saved include lives already saved by existing gun control legislation. In 1965 in the United States there were approximately, per 100,000 population,