¹ It ought to be recorded that several assistants did most of the work. Layman Allen prepared an extensive survey of competition for legislative office in Missouri while Stanley D. Hopper and Frank Munger made other contributory analyses. I am also indebted to several colleagues for most helpful criticisms of an earlier draft.

² All this argument avoids the genetic question—which came first, the formal or informal organization. I suggest that the sequential relationship is not invariably the same.

³ Part I of this analysis largely follows Turner's, Julius study of congressional primaries, “Primary Elections as the Alternative to Party Competition in ‘Safe’ Districts.” The Journal of Politics, Vol. 15, pp. 197–210 (May, 1953)CrossRefGoogle Scholar.

⁴ The combination of five primaries in Figure 1 smoothes the curve but the fundamental relation shown appears at each election. Throughout, the combination of primaries usually smoothes a curve or pictures a more regular set of relations than appears from the data of a single primary. This effect is most marked in the cells of very small numbers. Enlargement of the sample by the cumulation of data on several primaries would probably iron out most of the irregularities that appear in cells of small numbers in some of the tables that follow. In the choice of data for presentation for illustrative purposes somewhat irregular tabulations were often used lest an exaggerated impression of the uniformity of the correlations be conveyed.

⁵ For relevant comments on Indiana experience, see Hyneman, Charles S., “Tenure and Turnover of the Indiana General Assembly,” this Review, Vol. 32, pp. 51–67, 311–31 (Feb., April, 1938)Google Scholar.

⁶ The Missouri primaries of 1948 and 1950 parallel this expectation more closely than do the primaries analyzed in Table III.

⁷ Some of the irregularities of the distribution in Table III would probably be ironed out if there were a better measure of expectation of general election outcome than the vote itself.

⁸ If the argument may be anticipated, the pattern here points toward our main problem. A situation in which competition prevails in both primaries and the general election may represent equilibrium most difficult to maintain. Does such a state of affairs represent so unstable an equilibrium that over the long run it tends to be destroyed by a movement of decision and voter interest to one or the other party primaries?

⁹ The percentage in the incumbent curve in Figure 2 for the general election range of 30 to 40 per cent is based on only seven cases. All were in Republican primaries and six fell in 1948 when all the contests occurred. The chances are that the rout of the Democrats in 1946 resulted in the candidacy of Republican incumbents from traditionally Democratic districts in the 1948 primaries when the bleak Democratic prospects gave rise to rosy estimates of the Republican chances of holding districts they had won in 1946. See the analysis of the 1946–1948 shift in Table I.

¹⁰ On this point, see Key, , “Partisanship and County Office: The Case of Ohio,” this Review, Vol. 47, pp. 525–32 (June, 1953)Google Scholar; A Primer of Statistics for Political Scientists (New York, 1954)Google Scholar, Ch. 5.

¹¹ An element that may influence urban rates in Ohio is the practice of electing many legislators at large from the metropolitan counties. This practice seems to be, at least in such areas, conducive to a multiplicity of candidacies for nomination.

¹² The absence of competition seems most noticeable at the extreme of ruralism. Thomas Page, in commenting on the low temperature of competition in rural Kansas districts, observes: “The cold statistics hint that a subtle, informal, and highly personal process of co-optation often goes on in the smallest legislative districts. This leaves no important role for the formal election but the registration of the previously implicit consent.” See the suggestive analysis in his Legislative Apportionment in Kansas (Lawrence, Kansas, 1952)Google Scholar, Ch. 4.

¹³ These remarks make relevant an exception to the theoretical proposition about the fixation of political practices and structures by social and demographic characteristics. When one pores over data, such as that processed for this article, he sees, for example, pairs of metropolitan counties with fairly stable and radical differences in political practice. Such contrasts suggest that political practices, developed through the application of human effort and hardened by custom, may exist independently of demographic and social determinants.

¹⁴ No entirely satisfactory explanation suggests itself for the sharp fall-off in primary competition in the high ranges of party strength which remains even in the analysis reported in Figure 3, although ruralism seems to be relevant. In the 1948 Missouri general election 34.5 per cent of the completely rural House districts went without contest; 20.3 per cent of the other districts up to 50 per cent urban, and only 6.7 per cent of the districts over 50 per cent urban went without party competition. Of the seats Republicans won by default, 95.2 per cent were from districts under 50 per cent urban, while the comparable figure for the Democrats was 83.3. The horrid thought has been suggested that districts that go by default may be controlled by so effective an organization that the certainty of being counted out discourages competition both in the machine party primary and from the opposite party. Undoubtedly such situations are concealed within the aggregate figures, yet for this factor to take on importance for the main argument that follows it would have to be assumed that this type of machine control has become more widespread over the past half-century, an assumption contrary at least to common supposition.

¹⁵ The difference between the two Missouri distributions is accounted for in large measure by the growth in the number of seats uncontested at the general election. From 1908 to 1916 between 2 and 4 per cent of the House seats went by default at each election. In 1918 a sharp rise in the proportion of uncontested seats occurred and then the proportion of seats won by default settled down to a new plateau between 5 and 10 per cent of the total. In 1938, another abrupt upturn in the percentage of uncontested seats took place—to a peak of slightly over 30 per cent—and again the level settled down to another plateau in the neighborhood of 20 per cent. Missouri maintained a fairly even state-wide two-party competition over the period. In other states, with wider swings in party strength, it is probable that the cyclical factor would manifest itself in changes in the proportions of seats uncontested, i.e., apart from any long term decay in party leadership the proportions of seats uncontested might rise with an abrupt short-term decline in the strength of one of the parties. Or a sharp rise in minority party strength might increase the proportions of seats contested. Examination of the records of such states might result in the identification of variables other than nominating procedures with a significant bearing on competition and such effects might obscure or offset those factors on which attention is centered here.

¹⁶ Undoubtedly in both states there occurred nominations by primary which resulted from draft or designation by party committee. The Indiana procedure segregates at least some of such nominations.

¹⁷ If this theory is correct, the 1908 and 1948 distributions of New York legislative districts would resemble those of Indiana, for New York also nominates for important state-wide offices by convention. The 1948 New York distribution curls up at the extremes of the scale of party strength but this turns out to be the result of dual nominations mainly in the Bronx, presumably by inter-party negotiation. The New York data, hence, conform to expectation.

¹⁸ Compare, on the national scene, the historic reasons for being of Republican “machines” in the Southern states: to control national convention delegations and to dispense patronage. Or, consider the role of Democratic organizations in the northern New England states.

¹⁹ That is, a pair of distributions of the Connecticut districts according to their presidential vote in 1908 and 1948 resembles the distributions according to legislative vote shown in Figure 6.

²⁰ The 1948 Connecticut figure reflects in part the fact that in some rural towns the Democrats could muster only one candidate when two seats in the House were to be filled. Comparable New York figures for the two years are zero and 13.3 per cent. The uncontested 1948 seats were principally the result of dual nominations. If it is assumed that without the availability of this nomination procedure such seats would have been contested, 0.7 per cent of the 1948 seats went by default.

²¹ This step, it ought to be made most explicit, represents the crucial stage of the analysis. If it is conceded that inter-party competition occurs quite generally when state-wide nominating conventions are employed and considerably less generally when both state and local nominations are made by primary, the question remains whether some factor other than nominating procedure accounts for the differences. Since no means exists to determine whether all the other possibilities have been excluded, findings based on such reasoning must remain provisional. The entire analysis illustrates neatly the basic difficulties of comparative government. Even in the comparative analysis of American states, which should hold constant a great many factors that would complicate comparisons among nations, it is extraordinarily difficult to know when significant variables have been identified. For a treatment of the problem of method, see Macridis, Roy and others, “Research in Comparative Politics,” this Review, Vol. 47, pp. 641–75 (Sept., 1953)Google Scholar.

²² The Connecticut convention system is said to persist in part because of the fact that delegates in the Republican state convention are apportioned in the same manner as the members of the lower legislative house. The small towns, greatly overrepresented, would lose advantage by the adoption of the state-wide primary.

²³ While the governments of American states are notable for their surface similarities, even the most cursory examination unearths persistent pecularities in their informal political organization and practice. The basic question here comes down to whether some feature of Connecticut, independent and separate from the convention procedure, operates to maintain an informal political order different from that of the primary states. One reader suggests a variation on the demographic theory tested in the text, viz., that the changing demography of the primary states oriented the rural districts toward anti-metropolitanism. Given a degree of independence between national politics and state legislative politics, that sentiment of anti-metropolitanism reflects itself in a lack of party competition for legislative seats in these districts even though they may have a substantial minority vote in gubernatorial and presidential elections. In Connecticut, on the other hand, the demographic fact of a sprinkling of smaller cities rather than a dominant metropolis or two has not provided the conditions for so high a degree of metropolitanism. The hypothesis is difficult to test. If it were correct, the Indiana and New York 1908-1948 distributions would behave as do those of Missouri and Ohio. Another reader suggests that the pay of legislators may be a factor. Their real compensation has probably declined, but it may be doubted that this factor bears significantly on our problem, which in a way is to explain why parties at one time or under some circumstances put up candidates without much of a chance to win low-paying offices and do not do so at other times or under other circumstances.

²⁴ Cortez A. M. Ewing concludes, from his extensive analysis of Southern nominations, that local constituencies show “distinctive characteristics which belie the facile assumption that a large constituency is only an overgrown small one.” Primary Elections in the South (Norman, Oklahoma, 1953)Google Scholar.

²⁵ Related questions on which the data throw no light are whether the shift from competition between candidates of parties to that between candidates within parties brings with it alterations in the nature of the issues of campaigns, in the conduct of campaigns, in the relation of legislators to their party group in the legislature, in the nature of the choice presented to the electorate, and so forth.

²⁶ Institutions are hard to kill but survival bereft of function tends to be associated with a metamorphosis of mission, as when an ancient and honorable volunteer fire company lives on to sponsor an annual ball and to appear in festive costume on ceremonial occasions.

²⁷ Schattschneider, Elmer E., Party Government (New York, 1942), p. 82 Google Scholar.

²⁸ One is puzzled why in small constituencies forces set in motion by the primary do not draw practically all voters into the primary of one or the other of the parties at least for local purposes. The organizing influence of gubernatorial and presidential politics may maintain a bi-polarization that has its effects in a modicum of party identification vis-à-vis local candidates even in the smallest constituencies.