FOR decades the state of Kansas has been of special interest to all those concerned with the problems of politics and especially of elections. This interest has not been limited by the boundary lines of the commonwealth, but has extended from the Atlantic to the Pacific. Furthermore, in making another study of voting behavior,  it was noted that Kansas in the period under consideration always cast its electoral vote for the presidential candidate who won. Beginning with McKinley's election in 1900 up to and including the election of Franklin D. Roosevelt in 1932, Kansas has always helped to elect the winner. The only other states possessing such a record are Ohio and North Dakota. And, finally, having made a number of studies of voting behavior,  especially of western states and subdivisions thereof, it was thought wise to include Kansas as a unit in this larger and more comprehensive study of voting behavior in the West.
Forty-four counties were included in this analysis, representing the different districts or geographic sections of the state, the various economic interests and activities, and the large and small units, considered both from the standpoint of area and the size of the population. Table I presents itemized information concerning each county included in this study.
Except for background purposes, the elections analyzed in the study were limited primarily to the first three decades of the twentieth century. The study was limited also to a consideration of voting for President, for congress, for governor and the other state executive officers, for the state senate and the state house of representatives.
The results and conclusions  which emerged from this study were
arranged under two general topics: (I) material related to party victory; and (II) material related to voting behavior, and are presented according to this major classification. The first of these was further subdivided into national and state or commonwealth, and the second was broken into time differences, size differences, and location differences. In each case, the results are indicated and then the conclusions presented.
In the introduction of this article it was pointed out that the people of Kansas have voted for the presidential winner at each election in this century. Table II presents a graphic picture of this behavior.
Instead of the expression, "As Maine goes so goes America," it might well be said, "As Kansas votes, so goes the election." However, even after all these years of success, one hesitates to rely too much upon the political sagacity of the people of Kansas; the next election may find the record broken.
During this period Kansas has always had at least one Republican United States senator. In 1912 William H. Thompson, Democrat, defeated Gov. Walter R. Stubbs, Republican, for this high office. In 1930 George McGill, Democrat, defeated Henry J. Allen, Republican, and in 1932 Senator McGill defeated Ex-Governor Ben S. Paulen, Republican, for the senatorship. Consequently, out of thirteen United States senators chosen directly or indirectly by the people of Kansas, ten have been Republican and three Democratic, or, in other words, for more than two-thirds of the first thirty-three years of this century, Kansas has been represented in the senate by Republicans only, while during the remainder of the period the representation has been divided. Therefore, Kansas can be thought of as Republican in its relationship to the United States senate.
The analysis of the contests for election to the United States house of representatives is limited to the period 1904-1930. In 1904 Kansas was represented by seven congressmen from as many districts and one congressman at large, while in all subsequent elections, including 1930, the eight congressmen were selected from as many districts. Table III gives a picture of the party representa-
tion in the house of representatives as a result of the congressional elections held.
Throughout the period the first district, located in northeast Kansas, has elected Republicans, as has the third, which is located in the southeastern corner of the state. The sixth district, which is made up of the counties in the northwest corner, has been Republican, except when John R. Connelly, Democrat, was elected in 1916. The eighth district, established between 1904 and 1906 to take the place of the congressman at large as a result of redistricting the state, is composed of a narrow band of counties extending north from the Oklahoma boundary. They include Sumner, Sedgwick, Harvey, and McPherson with Butler off to the east. This district took care of Victor Murdock until 1914 when W. A. Ayres, Democrat, captured the district from Ezra Branine, the Republican candidate. Aside from one term, when R. E. Bird was elected, 19211923, the eighth district has been Democratic since 1914.
Turning next to the state ticket, it was found that during the period 1904-1932, twelve of the fifteen governors have been Republican. In 1912, 1922, and 1930, the Democrats were successful. No Democrat was able to secure reelection.
In the selection of the other elective state officers, the time period extended from the election of 1910 to include the election of 1930. The results for these two decades are very significant. Table IV
and the state executive offices.
gives a clear picture of the election results for these offices and for President and governor.
The election of Woodrow Wilson in 1912 and in 1916 and the election of Democratic governors in 1912, 1922 and 1930, seemed to have had no effect whatever upon the selection of men to the elected other state executive offices. All were Republican.
The state senate is composed of forty members chosen from as many districts. They are elected for terms of four years and all are up for election in presidential election years. This is quite different from that system used in California and the one used in selecting the United States senators. Table V shows the party strength in the upper house for the period 1908-1930, inclusive.
Thus the senate was clearly Republican for twenty of the last twenty-four years. In filling the two hundred and forty offices (40 offices X 6), 49 (20 percent) were Democratic. The senate was eighty (80) percent Republican during this twenty-four year period. Omitting the 1912 election, which appears to have been an exceptional situation, twenty-eight of two hundred were held by Democrats, thus giving the Republicans eighty-six (86) percent of the voting strength in twenty of the twenty-four years. For four years (1912-1916) the Democrats had fifty-two (52) percent of the voting power. However, the upper house of the Kansas legislature is distinctly Republican and the Democrats will have to capture and hold the upper chamber for several four-year periods before another evaluation will be in order.
The Republicans have a distinct advantage as a result of electing all forty state senators at the presidential elections. Either three out. of five, or four out of five times in current history, the Republican party has been successful in electing the President. This is of great help in successfully carrying state elections. One Democratic governor out of the three has had a friendly senate, while only one Republican out of nine has had an unfriendly upper house.
The lower house is distinctly Republican as measured by the number of victories obtained in the twenty-four year period under consideration (twelve elections, 1908-1930). Table VI presents the party strength as a result of the various elections held.
The Republicans have controlled the lower house for twenty-two of the twenty-four years under consideration, and in only two periods (1910-1912 and 1914-1916) was that control seriously challenged.
When the analyses of elections of governor, of members of the upper house, and of members of the lower house were combined, it became apparent that in nine of the twelve periods the three sections of the state government were united politically and that, in the remaining three periods, one party controlled two while the other party was in possession of one of the sections. During eight of the nine periods when unified control was present, the Republican party controlled. Only in the 19121914 period did the Democratic party control the three sections of the government. During each of the periods when the power was divided, the Republicans controlled two of the three sections: In 1914-1916 the governorship and the lower house, in 1922-1924 both houses, and in 19301932 both houses. Without adding the fact that in at least twenty-two of the twentyfour years included in this portion of the study all the elected members of the so-called state cabinet were Republican, it is quite evident that for all practical purposes and during the great. part of the time under consideration, the state officials have been Republican.
In this section the central question is, "What is the behavior of the unit or the comparative behavior of units under consideration?" It is not "Who won?" Consequently, the forty-four counties become the main feature. The state of Kansas, as such, is a factor only when "Time Differences" are being presented.  "Time Differences" will be presented under two headings: (1) the behavior of the state of Kansas, and (2) the behavior of the counties of Kansas.
Table VII gives a picture of the voting behavior of the state of Kansas when electing the President of the United States. Two measuring stickspopulation and voting population-are included in the table, as well as the absolute vote cast, so additional information can be developed in the process of presentation. It should be noted that the population and voting population estimates for 1932 are extremely temporary and will be revised as soon as the returns from the next census are available. 
Even at first glance, it is evident that the votes cast did not vary directly with changes in the voting population or the population. On three occasions when the population and voting population were continuing to increase, the absolute vote cast was less than in the preceding election. In the period prior to woman suffrage, the population increased about twelve (12) percent while the vote cast for President did not change appreciably. During the period since the adoption of woman suffrage for national elections, the popula-
tion has increased about ten (10) percent and the total vote cast about twenty (20) to twenty-five (25) percent. As one examines the behavior for the entire period, two points stand out prominently and call for consideration. With the introduction of woman suffrage in the election of President, the voting population was, for all practical purposes doubled, the actual increase was 101.2 percent, and, assuming equal interest and equal training or ability, one might. have anticipated that the vote cast in subsequent elections would have been approximately twice as great, but such was not the case. The mean * of votes cast in the four elections prior to the adoption of universal suffrage was 355 (in 1,000) while the mean for the period subsequent was 672, and it should have been 710 to 712. The same results appeared when analyzing the vote cast per 1,000 of the population. The mean prior to 1914 was 225, the mean since 1914 was 351, and it should have been about 450. The increase was fifty-six (56) percent instead of one hundred (100) or one hundred one point two percent. This may have been due either to a general lack of interest or to an undeveloped interest on the part of the women, or to a continued and serious loss of interest on the part of the men, or to a combination of these. The loss of interest was evident from the beginning of the period down to and including the election of 1920. Woman suffrage may not have contributed to this decline, but it certainly did not succeed in stopping the decline until after 1920-if then. In the second place, the last column, "Vote cast per 1,000 of the Voting Population," indicated the appearance of a "U" curve with the minimum point at 556 in 1920. These increases since 1920 are not as great in magnitude as the comparable decreases prior to 1920. These increases may be due, in part at least, either to the existence and growth of actual issues, or to developing interest on the part of the women of the state, or to a renewed interest on the part of the men which, in fact, means a developing interest on the part of the new generation of men, or it may be the product of a combination of these and other factors. In California similiar results were discovered. The mean of votes cast per 1,000 of population for President prior to the adoption of woman suffrage was 183 and the mean for the period subsequent was 275,  while the mean should have been about 360 to 370, if
doubling the voting population should double the number of people participating. When voting population was the basis of the California study, the 1912 election for President was the low point in the series, and it was also the first election in which the women of the state participated, possibly indicating, as in Kansas, that either the women did not immediately rush to the ballot box, or that, when the women were allowed to vote, a considerable number of the men stayed away, or it may have been a combination of both. This similarity of behavior is significant especially when the dates are not identical, when the states are of different sizes from the standpoint of population and when they are in distinctly different geographic regions. Nine general state officials are elected every two years. These nine are the governor, the lieutenant governor, the secretary of state, the auditor, the treasurer, the attorney general, the superintendent of public instruction, the superintendent of insurance, and the state printer. The time series showing the voting behavior as regards the election of governor and secretary of state are given to illustrate the general behavior pattern along with the results already presented.
Similar results appeared in these series and in the series for the other state offices as in the series for President. That is, prior to 1914, the votes cast did not change appreciably from one election
Just at this point in our discussion, another set of differences make their appearance. These might be labeled "office differences." The votes cast for the other general state offices are practically without exception fewer than the votes cast for the chief executive of the state and the votes cast for the governor of the state are generally fewer in number than the total vote cast for the presidential electors. Furthermore, one may infer that there is a definite relationship between the size of the vote cast for an office and its relative location on the ballot. Would the total vote cast for the first office appearing on the ballot continue to be larger than the second, and so forth, or would the total vote cast for President and governor continue to be relatively large regardless of position?
Measuring the differences between offices from election to election and from period to period, gives additional information and conclusions concerning time changes. Table XI gives the differences in votes cast per 1,000 of the population and per 1,000 of the voting population for governor and lieutenant governor, and between governor, at the head of the list, and the office of state printer, at the end of the list.
TITUS: VOTING IN KANSAS, 1900-1932 301
From an inspection of this table it is not only evident that the differences are greater as the differences in political rank increase and as the place on the ballot is relatively prominent or inconspicuous, but also there is a fourfold increase in differences based on population following the adoption of woman suffrage and more than a twofold increase in the differences when voting population is the base. In the third period, the differences are almost doubled when comparing the governor and the lieutenant governor and they are more than doubled when comparing the governor and the state printer. This increasing loss of interest on the part of the Kansas voters-the California voters express the same feeling, whether from the same causes or not it is not now known-forces one to consider the advisability of selecting some of the state executive officers by some method other than election.
The following conclusions are apparent when the state of Kansas is analyzed as a single political unit and its voting behavior is determined from the votes cast for the President and the nine state executive offices: (1) Prior to the adoption of woman suffrage in general elections, the voting behavior was more or less horizontal in its general appearance. (2) Subsequent to the adoption of universal suffrage, the voting behavior has been gradually increasing in its general appearance. (3) By plotting the values of these series of votes cast in percentages relatives to population and voting population, it was immediately seen that the angles of change from election to elec
tion became more acute or sharper as one moved from 1904 toward 1930. (4) The differences between the various lines, indicating the relative positions of the plotted values of the series, became greater as one moved from 1904 toward 1930.
The second section under the heading of "time differences" pertains to the voting behavior of the counties in Kansas. As it was out of the question to present the twelve time series for each of the forty-four counties, the more or less representative counties shown in Tables XII, XIII, XIV and XV have been selected to give a picture of some of the results obtained in this study.
The counties included in this study have similar behavior to that of the state as far as time differences are concerned. The general confusion in voting prior to the adoption of woman suffrage has produced a more or less horizontal pattern. The decline until the period following 1920, and then the increase in the past decade, are all in accord with the characteristics of state behavior. The increase in differences between the various offices is also apparent as one examines the county series.
As the so-called rhythmic factor was examined, the one-two or up-down beat was quite apparent when the office of governor was
under consideration. In order to eliminate the factor of the introduction of woman suffrage, the analysis was made of votes cast per 1,000 of the voting population. The range of behavior could extend from 1 (complete agreement with expected behavior) to 0 (complete disagreement). The extent of this agreement is presented in the form of fractions with the denominator indicating the number of counties included in the particular set of comparisons. The accompanying table indicates to what extent the counties behaved in harmony with our theoretical expectations.
In addition, it is important to note that eleven counties (25 percent of those included in this study) behaved completely in accord with the theoretical expectations, while thirteen of them deviated once and eight of them twice. Out of thirteen possible deviations, almost three-fourths of the counties deviated two times or less.
When one turns from considering the votes cast for governor to those cast for President, the factors are found to be more complicated. When the absolute vote cast was classified, it was found that., in 1908, thirty-eight counties cast a larger vote than in 1904, five cast a smaller vote, and one the same vote. In 1912 twelve went up, twenty-five down, and seven remained the same. In 1916, due partially at least to the introduction of woman suffrage, all fortyfour cast a larger vote. In 1920 eight followed the upward trend and thirty-six turned downward, while in 1924, without the stimulus of woman suffrage, all forty-four counties cast a larger vote than in 1920. In 1928 thirty-seven continued upward, five declined, and two remained the same. In 1932 forty-three increased and one showed a decline. Thus, when absolute vote cast is analyzed, the elections of 1908, 1916, 1924 and 1932 indicate a strong upward or major beat and the 1912 and 1920 elections produce the downward or minor beat. The 1928 election indicates a downward beat
in relation to the 1924 election, but it is not as pronounced as the other downward beats.
Using votes cast per 1,000 of the population as the basis for analyzing changes from one election to the next, similar results are obtained.
Here, again, one finds strong upward or major beats in 1908, 1916, 1924 and 1932, when compared with the minor beats of 1904, 1912, 1920 and 1928. The election of 1928 does not have as pronounced a downward break except when comparing it with the surrounding elections.
Combining the analysis of behavior when voting for President with the analysis of behavior when voting for governor, the following situation becomes apparent for the period under consideration. The behavior pattern for the election of governor is a "W" eightyear cycle pattern-the outer wings of the "W" being elongatedwhile the pattern for the election of President is a "V" eight-year cycle pattern superimposed over the "W" ( v) . If, on the other hand, one wishes to think of the behavior pattern for the election of governor as a "W" eight-year cycle pattern-the outer wings of the "W" being seriously shortened, then the pattern for the election of President becomes an inverted "V" "A" superimposed over the ***
From the information presented, it is immediately seen that major beats are not associated with a particular major party. In 1908 and in 1924, the Republican candidates were successful, while in 1916 and 1932, the Democratic standardbearers were victorious. These major beats are not related to candidates seeking election or those seeking a second term. In 1916 and in 1924 Presidents sought reelection and were successful, while in 1908 and in 1932 those seeking first terms were successful. Furthermore, there seems to be no close relationship between major beats and economic depressions
or periods of business activity. The elections of 1908 and 1932 follow periods of economic stress and the elections of 1916 and 1924 are in the midst of periods of business activity. The election of Republican and Democratic governors seems to have little in common with these patterns. Of the three Democrats elected, one was with a Democratic President (Wilson, 1912), two were carried into office in a bi-election (1922 and 1930), and none was elected at a major point on the presidential pattern. Republican candidates were successful at major points and at minor or low points on the presidential pattern and at major and minor points on the gubernatorial pattern.
Analyzing this problem of possible rhythm when votes cast per 1,000 of the voting population are used as the basis for the study, other results appear than those in the preceding paragraphs. The following summary tells the story.
Furthermore, eleven of the forty-four counties behave as the summary indicates; i.e., up, down, down, down, up, up, up. Thus a "V" twenty-four year cycle pattern presents itself when voting population is used as the measuring stick. Looking back over these paragraphs presenting material which pertains to rhythm, one is puzzled concerning the significance of these observations, and asks whether any general propositions are to be evolved or extracted from these behavior patterns.
Would it be entirely absurd for one to expect or anticipate the 1934 vote for governor to be down when compared with the 1932, the 1936 vote for President to be down when compared to 1932 and the vote for governor to be up when compared with 1934? It will be interesting to note to what extent these anticipations are realized. The theory here presented has been upset neither by the 1932 election nor by the 1896 election (when Kansas was treated as a single unit), but has been further verified. With only one cycle available,
when voting population is the basis, it would be unwise to extrapolate beyond 1936. However, it would be interesting if the 1936, 1940 and 1944 presidential elections should prove to be down, down, down, when voting population is the measuring stick applied to the votes cast.. Consequently, from the analysis of "time differences" for the state of Kansas and for forty-four fairly representative counties of the state, certain uniformities are discovered, such as (1) the possibility of rhythmic behavior between the various elections; (2) an increase in the amount of difference between votes cast for the different offices as one moves from early elections to more recent ones, and (3) either a reticence on the part of the newly enfranchised voter to participate immediately upon being given the right to vote or the refusal on the part of an element among the men to participate in the first few elections after the adoption of the amendment, or both of these factors working together.
As attention was turned to the consideration of "size differences," the material was reclassified and the results analyzed in the light of the new relationships. For each election beginning with 1904 and continuing through the election of 1930, the counties were ranked from the one having the largest population to the one having the smallest, and in a second analysis they were ranked on the basis of voting population. Seven classes were established similar to the arrangement used in other studies. The classification is as follows:
Class A-Population over 100,000
The same system was used when "voting population" was the basis of operations. It should be noted that in one or two of the early elections there were no counties in Class A and in the latter elections no counties in Class G. Table XIX presents the means of votes cast for President per 1,000 of the population by classes.
This classification of the counties of Kansas further validates a possible scientific law of voting behavior which was first suggested in March, 1928, namely, the larger the population of a political unit the smaller the vote cast relative to the population. By com-
bining Classes E, F and CT, there would be only one exception to the rule for these counties. Of thirty-nine possibilities there were seven exceptions to uniform behavior in voting for President. In voting for governor, there were eleven deviations from uniformity of a possible seventy-seven, and in voting for congressmen there were thirteen deviations of a possible seventy-seven. When the counties are ranked on the basis of voting population, the results obtained are presented clearly by analyzing Table XX.
In this table the so-called law of voting behavior manifests itself even more clearly than in the table presenting the material based on the population. The larger the voting population of a political unit, the smaller the vote cast relative to the voting population is a statement of human behavior relative to voting activity which is applicable in Kansas and in California for the periods considered. From this and other studies partially completed, one is justified in suggesting that this statement of behavior may be universally applicable where a relatively large proportion of the population does have an opportunity to participate in the selection of governmental officials by means of the Australian ballot. Kansans and
Californians may be peculiar when it comes to voting activities, but up to the present no objective evidence has been introduced to substantiate such a position, and, until such evidence is introduced, it ought to be considered sound to assume that the voters in these two commonwealths are reasonably representative of voters in general and particularly of Anglo-Saxon voters. Tables XXI and XXII present the behavior of the forty-four counties when voting for governor and for congressmen.
In measuring and analyzing the votes cast for governor and for congressmen,
as was the case with the President, the results further validate the suggestion
that the rule of voting behavior-the larger the population and the voting
population of the political unit, the smaller the relative vote cast-may be
universal in extent. When this possible law of voting behavior was first
suggested, an important problem presented itself which up to the present time has
not been solved; namely, are democracy and popular control of government through
systems of elections compatible with metropolitan areas and rapidly growing
political units? If there is further develop
went in the "back to the farm" movement, this problem may solve itself, but if the urbanization of America persists and cities continue to grow, can we expect democratic control to be established, or, if established, to be maintained over government?
Some years ago Prof. William Bennett Munro suggested that there might be some relationship between voting behavior and the area or size of the political unit or units being studied. Since that. time, the author has been watching for an opportunity to follow up this suggestion. As a result, the counties included in this study were classified on the basis of acres contained within their boundaries. Five classes were established:
The results obtained from analyzing six elections for the President on the basis of this classification of counties is indicated in Table XXIII.
Recognizing the meagerness of information and the absence of a distribution compatible with the classification, the uniform behavior exhibited on the part of the counties in these elections is not to be taken too seriously at this time. The presentation merely indicates another method of analyzing the possible effect that "size differences" may or may not have upon voting behavior.
Finally the results of analyzing the statistical data on the basis of "location differences" are presented and briefly compared with
results obtained in other studies. For this study the state was divided into twelve districts. The districts and the number of counties contained in each are shown in the accompanying diagram.
The counties included in each district are listed in the following table:
The two block patterns of Kansas which follow indicate, on the basis of population and voting population respectively, the voting behavior by geographic districts. The values of "M" (arithmetic mean) and of "b" (quadrennial change) in the equations of lines of best fit to votes cast for President per 1,000 of the population are