It is evident that the peak in building employment does not coincide with the peak of contracts awarded or of permits granted, but occurs some three or four months later. Consequently, though they have a wider geographical scope and extend over a longer period of years than the available direct statistics of employment or unemployment, it does not appear desirable to utilize the contract or permit figures as indices of seasonal variation in employ

ment.

In Fig. B, the curve for New York building and street labor is repeated, and curves are added to indicate the seasonal variation in Wisconsin building construction during the three years from 1922 to 1924, inclusive, and also employment as reported by contractors in twenty-six cities of the United States in 1922. (See footnotes to Table 57 for sources.) To obtain a composite estimate of seasonal fluctuations in building employment, weights of six, three, and one were assigned to the New York, Wisconsin, and 1922 series, respectively, and the result is plotted as Curve "h" in Fig. C of Chart 53.

Our index of seasonal variation in employment on highway construction (Curve "g", Fig. C, Chart 53) is also admittedly only an estimate based upon fragmentary data. The statistics used cover the years 1922, 1923, and 1924, and include (1) the number of common laborers, by months, on highway projects receiving Federal aid in the fourteen states included in the New England, Middle Atlantic, and East North Central sections, and (2) employment on highways in Wisconsin. In computing the index of highway construction these two series were weighted by the relative population of the states represented.

Finally, a composite index of employment in construction (Curve "i" in Fig. C, Chart 53) was computed by combining the building and highway indices, assigning a weight of six to the building and one to the highway series. Obviously, this index of construction employment must be taken as a rough approximation. It indicates small activity in the first quarter of the year; increasing employment in the second quarter; maximum activity in the third quarter; and then a decline in October, November, and December.

In some of the subsequent comparisons, the index for railway maintenance and the index for construction have been weighted by the estimated numbers employed in these occupations in 1909 and combined into an index of "Selected outdoor industries" (Curve "f" of Fig. C, Chart 54).

Using the same method of weighting, the indices for all the industries under consideration-that is, factory employment, bituminous and anthracite coal mining, railway maintenance, and construction, have been combined into an index of seasonal fluctuations in "all selected industries" (Curve "g" of Fig. D, Chart 54). Limitations.

Before proceeding further, it may be well to summarize the limitations of these indices of seasonal variation. In the first place, the object in mind has been to obtain evidence of seasonal fluctuations which may be applied to the years immediately preceding the Great War, since the indices of seasonal variation in migration are computed chiefly from data for these years. It has been necessary, however, to utilize some employment data applying to more recent years. Furthermore, these evidences of employment conditions have in some instances been fragmentary and indirect. While care has been taken to make the indices as representative of the actual conditions as possible, and we have no reason to believe that they are inaccurate in material respects, yet the existence of a considerable margin of possible error must be recognized.

A further source of possible misinterpretation of the significance of seasonal fluctuations in employment lies in the inadequacy of the available information concerning the extent of dovetailing of employment in various industries. When industries are separately considered, the aggregate account of seasonal fluctuation may be magnified by the failure to take into account that workers may shift from one industry to the other when periods of activity do not coincide. On the other hand, the consolidation of data for several industries may create the impression of a more uniform seasonal distribution of employment than actually exists for most workers. It is obvious from the data which we have been examining that factory employment is declining in midsummer while activity in the outdoor industries is increasing; but, unless idle factory workers shift readily to outdoor industries, a consolidation of the data for all the important industries may convey an exaggerated impression of the degree of seasonal regularity in employment. While such an index is useful for present purposes, it is not an adequate measure of the variation in employment for individual workers or groups of workers.

We have analyzed separately the seasonal fluctuations, first in migration, then in employment. We may now turn to a direct comparison of the degree of similarity in these seasonal movements.

COMPARISON OF SEASONAL TENDENCIES IN
IMMIGRATION, EMIGRATION, AND

EMPLOYMENT

The most significant of the several indices of seasonal changes in migration and employment which we have been discussing in the preceding pages are brought together in Chart 54. In each of the

Numerical data for Curve "a" are in Table 58, column A; Curve "b", Table 58-A; Curve "c", Table 56-C; Curve "d", Table 56-F; Curve "b", Table 56-H; Curve "f", Table 58-C; and Curve "g", Table 58-D.

four sections of this chart the curves for male immigration and for male emigration are repeated, and for comparison therewith there is given, in Fig. A, the index for factory employment; in Fig. B, the

index for bituminous coal and also for anthracite coal, both based on production; in Fig. C, the index for selected outdoor industries; and in Fig. D, the index for all selected industries. To avoid a possible misinterpretation of the relative violence of the seasonal fluctuations in the several industries, it should be noted that for each of the sections in Chart 54, Scale 2, for employment, has a different scale unit.

TABLE 58.-INDICES OF SEASONAL FLUCTUATIONS IN MIGRATION AND EMPLOYMENT Monthly average = 100

A. Male immigrants, January, 1893, to December, 1913. (See Table II, in Appendix).

B.

Male emigrants, January, 1910, to June, 1914.

C. A composite of indices for railway maintenance and construction (see Tables 56 and 57), weighted according to numbers employed in 1909.

D. A composite of the indices for factory employment, bituminous and anthracite coal mining, railway maintenance and construction (see Tables 56 and 57), weighted according to numbers employed in 1909.

Factory Employment (Fig. A, Chart 54).

Immigration agrees only moderately well with the seasonal fluctuations in factory employment. The peak in factory employment is reached earlier in the year, hence the bulk of the immigrants arrive after the maximum demand has passed. Both series exhibit a summer decline, a fall recovery, and a slump late in the year. For most of the year emigration is increasing while factory employment is decreasing, and vice versa. Emigration declines in February, due partly to the length of the month, while factory employment is increasing, and increases from March to July while factory em

ployment is diminishing; also the high level in November and December corresponds with declining factory activity.

Coal Mining (Fig. B, Chart 54).

The low point in bituminous coal mining comes in the second quarter of the year, at the period when immigration is greatest, and the peak does not occur until some six months later; hence the spring rush of immigrants is premature if their destination is the bituminous coal mines.

The fluctuations in immigration are somewhat better timed for anthracite coal mining, as the peak of activity in this industry is reached after, rather than before, the high tide of arrivals. In both types of mining, a decline occurs in November and December, coincident with increasing emigration and declining immigration. Selected Outdoor Industries (Fig. C, Chart 54).

The seasonal fluctuations in immigration are well timed for employment in the outdoor industries. The number of immigrants is small in the stagnant months of December, January, and February, increases with the spring rise in outdoor work, and reaches a peak early enough to make it possible for the bulk of newly arrived immigrants to participate in outdoor work throughout the extent of the summer boom. Then, as outdoor employment declines rapidly in the closing months of the year, the tide of emigration swells rapidly and the volume of immigration recedes.

All Selected Industries (Fig. D, Chart 54).

In general, the composite seasonal movement of employment in all the selected industries exhibits the same features as the index for outdoor industries, hence the comments just made for the outdoor industries can be applied to the combined seasonal fluctuations in employment in the entire group of industries under consideration. The heaviest immigration is when employment is increasing and somewhat in advance of the maximum in employment activity, and the decline in immigration and the increase in emigration at the close of the year coincides with the falling off of employment.

THE NET VOLUME OF ARRIVALS LESS DEPARTURES AND THE MONTH-TO-MONTH CHANGES IN THE NUMBER EMPLOYED

So far in this chapter we have been comparing the seasonal changes in employment and pre-war migration with reference to the time of year at which increases or decreases take place. There