the georgia primary system 1898–1970: factional patterns and political regions

15
THE GEORGIA PRIMARY SYSTEM 1898-1970: FACTIONAL PATTERNS AND POLITICAL REGIONS William Collins Introduction This note examines the pattern of political behavior which emerged from the adoption of the direct primary in Georgia. The note proposes a model to account for primary voting behavior, and then proceeds to test the empirical consequences which can be derived from it. The Southern Primary Research on political behavior in southern' primaries emphasizes 1 the importance of personality and the significance of localism. Empirical studies of southern primary voting focus upon factional systems dominated by particular personalities2 The implied view in these studies is that only the presence of a dominant political personality provides important political alternatives in the primary. However, Key has pointed out the limitations of factional leadership: the lack of continuity which makes information difficult to disseminate to the voter, the difficulty of a fight between the "ins" and the "outs", all of which produces in the operations of government itself a high degree of in~tability.~ Key and other analysts of southern primaries thus argue that the primary electoral process makes it difficult to choose between political alternatives. Electoral groupings are transient and devoid of any substantive political meaning. 157

Upload: william-collins

Post on 28-Sep-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

THE GEORGIA PRIMARY SYSTEM 1898-1970:

FACTIONAL PATTERNS AND POLITICAL REGIONS

William Collins

Introduction

This note examines the pattern of political behavior which emerged from the adoption of the direct primary in Georgia. The note proposes a model to account for primary voting behavior, and then proceeds to test the empirical consequences which can be derived from it.

The Southern Primary

Research on political behavior in southern' primaries emphasizes 1 the importance of personality and the significance of localism.

Empirical studies of southern primary voting focus upon factional systems dominated by particular personalities2 The implied view in these studies is that only the presence of a dominant political personality provides important political alternatives in the primary. However, Key has pointed out the limitations of factional leadership: the lack of continuity which makes information difficult to disseminate to the voter, the difficulty of a fight between the "ins" and the "outs", al l of which produces in the operations of government itself a high degree of in~tabi l i ty .~ Key and other analysts of southern primaries thus argue that the primary electoral process makes it difficult to choose between political alternatives. Electoral groupings are transient and devoid of any substantive political meaning.

157

Page 2: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

To deal with southern primaries the paper postulates a hypothetical structure for the primary system. It i s assumed, other things being equal, that a one party competitive setting will tend to aggregate individual political choices strictly on the basis of local effects.4 If a primary setting does act on the basis of local effects, it would follow that the counties will group themselves along a theoretical contour viz. the county vote would tend to be heaviest toward a center defined by the location of t h e candidate and this vote then tends to drop-off as a county is located further from the center. A set of such contour lines, one for each candidate, might, in the ideal setting, with

no intervening factors define a set of political regions. These regions, in effect, become summaries of how information about the candidate has diffused throughout the political surface. The major empirical question is, do such regions persist over time? This question is important to the analysis of primaries because the persistence of a regional effect under the assumption made about the basic structure of the primary will provide critical information about the character of the primary system. In other words, the specific character of the regional effects, assuming they can be demonstrated, should detail how the primary is operating to convert local effects into votes:

Research Hypotheses, Data and Technique

The major idea of our approach is that Southern primaries act to convert information in the form of physical distance into vote distributions. An implication following from this is counties in general should cluster around the home county or major location of the candidate, in effect forming regions. This is a logical outgrowth of the distance concept For any given election under this assumption the regional effect should be present and act as a means of ordering the significance of the vote distribution. Another idea follows from this. Key in his lntroducrion to state Politics presents a table which displays how the number of democratic party nominees over time tended to come from Boston.5 Key remarks "That under the direct primary marked alteration in the location of the controlling power points is mirrored in the changing geographical distribution of the party nominees."' I f this is true it would seem to follow that geographical distributions of voters should also shift from one set of locations to the other. Further there should be some sense of persistence over time reflecting the fact that successful ca'mpaigns tend to follow the same

158

Page 3: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

pattern. The argument is this.

Assumption: Without the party label primaries covert information about candidates into votes in the simplest possible fashion through distance i.e. proximity equals knowledge. Point f n ~ : Counties under this assumption will tend to cluster along a geographic continuum centered on the location of the candidates. Point three: Such dusters tend to persist over time depending variously upon the rules of the game and the degree of power concentration. Successful candidates tend to come from certain areas thus reinforcing the impact of "friends and neighbors." Change in the system may be measured by how these clusters breakup and reemerge around other points.

The import of these assumptions is to provide a model by which to account for primary voting behavior. Previous literature has emphasized the role of idiosyncratic factors, the lack of continuity from one election to another and the importance of dominant political personalities in the Southern primary system. If a regional factor can be demonstrated it may be assumed that while such factors as those previously mentioned do exist they will have impact beyond the immediate outcome of single elections.

Data and Problem Statement

Our data consist of vote returns by county for every Gubernartorial primary in Georgia extending from 1898 to 1970 (1 902 is not available: see Appendix 1 1 ) ; missing data account for about three percent of the entire data set. The missing data tends to be concentrated in the elections before 1916. As a starting point, the primaries will be ordered into three periods. The periods center around the role of the Talmadge faction. Period one extends from 1898 to 1930 and is labelled the pre-Talmadge faction. The second centers on the Talmadge elections from 1932 to 1950 and the third period goes from 1954 to 1970 and is labelled the post-Talmadge era. This approach has ample precedent being in effect employed by Key in Southern hlitics in his discussion of Georgia.

I f our basic idea about primaries is correct, then it would follow that candidates should tend to cluster or group themselves into geographical patterns. The use of Talmadge as a pivotal point makes it easy to assess the importance of region from one set of elections to the

159

Page 4: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

next. me consequence of the m e e l therefore i s thatJhere are regional patterns centering around clusters of candidates.

Analysis

In order to develop a convincing test for the persistence of regional effects we employ the technique of principal components analysis, and attempt to extract a set of simple factors which will account for the variation within each set of elections for every candidate. This a proach has been adopted widely in the comparative voting literature! Wildgen has used this approach to measure critical election change in one-patty Louisiana and Todd and Ellis have analyzed faciional patterns in Texas from 1944-1972. Our idea is to cluster the candidates from one election to the next for the three periods. This will present a basic pattern of the interrelationships of county vote patterns across time. As a second step a set of composite indicators are

9 developed which then permit us ro compare counties wirh each other. We then used these composite scores to see of regional patterns have emerged and persisted during each of the three periods. The approach tends to be relatively cumbersome in that a great deal of within period change is obscured. However, the approach i s justifiable in the sense of choosing a set of predefined elections and examining whether or not such elections do tend to demonstrate regional effects over the respective periods.

Tables one through three present the results of the principal components analysis. For period one 1898-1 930 the first five factors account for only 55.1 percent of the total variance within the set of elections. Factor one accounts for only 19.0 percent. There is some data reduction, but the results are not spectacular. Factor one can generally be said to account for the differences between the Hoke Smith and HowelVBrown factions in first years of the Georgia primary. Smith and Smith related candidates load negatively while Brown and Brown related candidates load positively. The distribution of the composite scores based upon factor one is displayed on figure one. What we did was to take the first twenty positive factor scores and compare them to the top twenty negative factor scores. A rough indicator of regionalism is the number of contiguous counties which emerge in this figure, based upon the clustering of candidates over a thirty-two year period, no regional or general clustering effect is apparent. The exceptions are small clusters of counties in the northwest and middle eastern section of the state.

8

160

Page 5: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

1898 E l e c t i o n

4. Chandler

B . B e m e r

c. Atk inson

1906 E l e c t i o n

4.. H. Snith

B. t i w e l l

c. Russel l

0. E s t i l

E. J. SnJth

1908 E l e c t i o n

A. Ermrn

8. Smith

1910 U e c t i o n

A. l lmm

B. S m i t h

1912 E l e c t i o n

A. Alexander

8 . H a l l

t. S l a t o n

1914 E l e c t i o n

A. Pndenon

6. Hardman

C. H a r r i s

1016 E l e c t i o n

A. Dorsey

8. Hardman

C. H a r r i s

c. P o t t l e

-. 106

-. 326

.491

-. 552

.035

.253

.403

-.OK1

.931

-.931

.898

-.a96

-.266

.061

.144

.519

.122

- .326

.007

. I17

.OI4

-.150

-.056

.056

-. 158

.158

.213

-. 103

,022

-.111

- .022 -. 128

.090

TABLE 1

FACTOR SCORES P m n n ONE irjye t o 1 9 ~

PRE-TALMlnGE PEFIOD

1925 E l e c t i o n

A. 3rwn

8. g a r h i c k

C. H o l d e r

0. .Jalker

A. l i a r h i c k

9. U a l k e r

1922 E l e c t i o n

A. B a y l o r

8. t 'a rcb ick

C. :Jalker

1926 E l e c t i o n

A. Carswel l

B. Hardman

C. H o l d e r

.492

- .S84

.262

. l o1

.454

-.537

.413

.244

-.389

v.263

,249

-. 150

,089

,061

- .055 ,050

-. 153

.014

-.w5

,249

-. 19 .089

,061

-.OM

. 0 50

,281

-. 759

.766

.025 - .5&

- . l a2 ,482

0. j lood .079

E. Hardman .141

F. H o l d e r -.075

1928 E l e c t i o n

A. Hardman .075

6. R i v e r s - .209

1930 E l e c t i o n

A. Carswel l - . 151

8. H o l d e r -.039

C. P e r r y .052

D. R i v e r s .281

E. Russell -.092

F. Carswel l .'159

G. Q u s s e l l .059

-. 107

.394

.022

.321

-. 182

-. 574

.353

.373

.115

,371

-.597

.597

161

Page 6: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

1932 Election

A. Edrards

8. Hardrick

C. Holdcr

D. Kelly

E. H l x

F. O'Kelly

6. Stmen

H. Talmadgc

1934 Election

A. l a l i d g e

8. Gfllian

C. Pfttman . 4936 Election

A. Fortson

8. Redine

C. R i v e n

I938 tlectqon

A. Howl1

8. knal lan

c. Rivers

b. Uwd

1940 Election

A. H i r

8. Robtrts

C. T i l m a d p

.044

.I02

.la7

-.041

.594

. oao

.094

- .696

-.695

,165

.697

.396

- . a 9

.407

- .635

.179

.517

.05G

.489

-.I76

-.692

TPBLE I1

FACTnR SCORES PERIOD TIiO 1932 to 1950

TAWAPGE PERIOD

- . X ) E

- .231

- .346

.339

. 3 9 3

.033

-.647

.023

.*72

.067

-.412

-.422

- .222

. y35

-.i77

- .284

.419

.047

-.I19

-.014

. lo2

1912 E lection

A. Amall .a59

8. Tallladoe -.a59

1946 Election

A. Charmichael .579

8. O'Kelly .173

C. Rivers .095

D. T i l ~ d g c -.737

1948 Election

A. O'Kelly .283

8. R a b m ,335

C. TrlMdge -.715

0. Thanpron .679

E. WIlllr .450 1950 Election

A. h e y - .053

8. Baker .274

C. Jenkins -.OM D. Talwdp -.la

E. Thoapsm .143

.182

-. 101

-.399

-.163

.7??

-.a3

-.m .051

-.193

.242

-.145

.179

-.015

,250

.053

-.062

162

Page 7: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

1954 Elect ion

A. Banf ield

8. Garland

C. Gwen

0. G r i f f i n

E . Hand

F. Uinder

.G. Nelson

H. Thomas

'1. Tl\ompsm

1958 Elect ion

A. Abemathy

6. Bodenhauser

C. Vandiver

1962 Elect ion

A. G r i f f i n

B. Yingham

c. O'Kel ly

0. Sanders

E. T h m s

- .a47

.no8

,416

-.I71

. 141 - . S l , q

.315

. 4 e ~

,152

-.092

-.295

.346

-. 374 -.391

-. 185 ,367

.062

TPBLE I l l

FPCTOR SCORES 1954-1970 PmT TPL!IPDGE EPP

1966 E l e c t i o n

-.974 P . P r n a l l ,831 .051

.164 C . Dyrd -.%1 .007

.203 C . t a r t e r ,076 .695

.lo5 C. Gray -.194 -.694

-.956 E. hddox -.667 .357

-. 226 F. O'Kel ly - .095 .011

- . 253

. 2 4 7

-.039

.611

-. 739 .567

-.627

.395

.546

.590

,021

t . A m a ' l

H. rsddcr

1970 E l e c t i o n

P . Carter

8. cox

C. hagretts

0. I r w i n

E. King

F. EcatL+ews

G. Saunders

H. Stoner

.036

-.936

-.782

.055

-. 144 .I15

.317

.loo

.522

-.316

. 166 -. 166

.195

-.'I61

.036

-.zm -.406

-.34

.20?

- .277

I . S w i t qqq -.050

J. Carter -.764 -.010

K. Saunden .784 -.010

This same approach was used for both Talmadge and post-Talmadge periods. The principal components analysis yielded basically the same results. The first five factors for the Talmadge period accounted for 52.5 percent of the overall variation and the first five factors for the post-Talmadge period explained only 49.5 percent of the variation. Figure two is based upon the factor which accounts for the Talmadge, anti-Talmadge factionalism characteristic of this period. This factor in turn accounts for 23.2 percent of the overall variation in the period under consideration. The figure represents a plot of the top 20 composite scores for both the negative and positive loadings. When a l l the elections of the period are considered together no pronounced regional effect is observable. Regional effect here is being viewed as a cluster of scores grouping together and tending to decrease as one m v e s from the center. Fulton, Coweta, Dekalb and Troup counties do

163

Page 8: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

tend to cluster but the appearance of Fayette county belies any overall regional effect.

The third period also reflects l i t t l e to no clustering over the period from 1950-1970. The figure is based upon composite scores derived from the first factor which accounts for only 19.6 percent of the total variance. This factor is identified as the Maddox-Griffin versus the Arnall-Sanders faction. The loadings are not high and reflect only a small degree of interrelationship among the candidates across the races.

Consideration of the entire set of elections using the three figures as the basis of comparison suggests that factionalism in Georgia politics cancot be accounted on the basis of within state regional factors. This hypothesis was tested in a very broad sense using clusters of candidates from each of three periods, the periods being defined around the dominant figure of Eugene Talmadge. This set of results leaves us in an equivocal sitions with respect to providing a structure for primary elections.lgOWe have contrasted two interpretations of primary voting behavior. On the one hand is the view that primaries obscure information and tend to reinforce the influence of ad-hoc and idiosyncratic factors. From this perspective there is l i t t le continuity from one election to the next and each election tends to have l i t t le, if any effect, on the subsequent election. On the other hand we attempted to build a model based on the idea that primaries tend to equate candidate information with physical proximity. This assumption led to the postulation of a regional effect which would provide a sense of continuity terms of which a series of elections codd tk analyzed. Two sets of findings emerged. First the relatively low amounts of variance accounted for by the three sets of principal components analysis coupled with the generally low sets of loadings makes it hard to reject the idea that there i s l i t t le continuity between elections. Secondly, the geographical distribution of opposing factor scores make it very difficult to postulate any sort of regional effect in the gubernatorial primaries based upon an extension of the "friends and neighbors effect." From these results the obvious response to the problem of how to interpret primary voting is to say that you cannot reject the initial interpretation of the primary voting as a random and idiosyncratic behavior. However, we are convinced that this i s not the proper interpretation. Our evidence unfortunately does not make it possible to reject the traditional view. However, it seems to us a t this time that Southern primaries do act to convert important and

164

Page 9: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

165

Page 10: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

Clusters o f Composite Scores based on analysis o f the Deriod 1932-1950 a =Anti-Talmadge Scores

W =Talmadge Scores

166

Page 11: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

Cluster o f comoosite scores tased on analysis o f the Period 1950-1970

=Arnall/Sanders Factlo- IZI = Haddox/Griffin

Facticn

167

Page 12: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

significant information into vote distributions and the fact that such information cannont be shown a t this time to reflect the idea of distance does not alter the fact that the information is being processed in some consistent and patterned way. The problem of the empirical and normative implication of the primary system is thus to our minds s t i l l very much an open question.

'The author wishes to thank Mn. Susan Collins for her most valuable contribution in the preparation of both the manuscript and tabl- in this paper.

' Refer to:

a. V. 0. Key. Southern Politics in State and Nation, (N.Y. Vintage Books,

b. L M. Holland, The Direct Primary in Georgia, (Urbana University of

c J. M. Kousser, The Shaping of Southern Politics, (New Haven Yale

1949).

Illinois Press, 1949).

University Press, 1974).

'Refer to:

a. A. P. Sindler, "Bifactional Rivalry as an Alternative to Two Party

b. V. 0. Key, op. cit., pp. 106-1 29. c Joseph L Bernd, "Georgia Static and Dynamic" in W. Havard (ed.)

The Changing Politics of the South, (Baton Rouge: LSU Press, 1974). pp. 294-365.

d. Earle Black and Merle Black, "The Wallace Vote in Alabama: A Multiple Regression Analysis," Journel of Politics 35 (1973) pp. 730-736.

Competition in Louisiana," American Political Science Review.

3V. 0. Key, op. cit , p. 89.

4This phenomena has been described in R. Tatalovich, "Friends and Neighbors Voting: Mississippi 1943-1973", Journal of Politics, 37 (1975): pp. 807-81 4, as vrrell as in Black and Black, op. cit.

5V. 0. Key, An lntroduction to State Politics, (New York: Knopf, 1967). p. 155.

61bid., p. 186.

7V. Capecchi and G. Fanini, "A Classification of Italian Regions According to Electoral Behavior," Qualify and Quantity, 2 (1969): pp. 116-23, and V. E. McHale and J. E. Mclaughlin, "A Further Note on the Classification of Political Regions According to Electoral Behavior." Qualify and Quantity, 9 (1975) pp.

Alfred A.

168

Page 13: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

303-325, are two of the better studies employing this approach.

8J. Wildgen, "The Detection of Critical Elections in the Absence of Two Party Competition," Journal of Politics, 36 (19741, pp. 465-75.

'This technique is described in R. J. Rummel, Applied Factor Analysis (Evanston: Northwestern University Press, 1970) Chapter 13.

"The existence of Georgia's county unit system would seem to be a mmplicating factor obscuring the outcomes of this analysis. However, this system ms no more than a vote migh t ing scheme impacting more on candidate strategies than upon the idea that votes transform simple distances.

APPENDIX 1: Data Sources

1898 Primary The Atlanra Consrirurion 1906 Primary The Atlanta Constitution 1908 Primary The Arlanta Constifurion 1910 Primary The Atlanta Constifurion 1912 Primary The Atlanta Constitution

1914 Primary The Atlanta Constirution 1916 Primary The Atlanta Constifurion

The A tlan fa Journal

9, June 1898 26, August 1906 7, June 1908 25, August 1910 23, August 191 2 23, August 1912 20, August 1914 13, September 1916

19201948 Primaries: Alexander Heard and Donald S. Strong, Southern Primaries and Elections 1920-1949. Freeport: Books for Libraries Press, 1970.

1950-1970 Primaries: Consolidated Vote: State Democratic Primary €lections 1950- 1970. Office of Secretary of State. Atlanta, Georgia.

APPENDIX 2: Chronology of Georgia Primary Rules, Regulations, and Laws

1898 1904 1907 1907 1908 1917 1922 1946

1948 1963

1st Statewide Primary (Democratic Party). All primaries must be held within sixty days o f regular election. Temporary suspension of county uni t system. Legal requirement of run-off election. Reinstatement of county uni t system. Neil1 Primary Act (County uni t system required be law). Australian Ballot Law. End o f White Primary in Georgia. Chapman v. King. 154 F. (2nd) 460

Mandatory Secret Ballot Law. (Georgia Laws 1949, pp. 1291-93). End o f County Unit System.

( 1946).

Gray v. Sanders 372 V.X. 368 (1963) ( In 1962 district court ruled the unit system unconstitutional, thus giving Georgia i t s f i r s t primary elec- t ion since 1908 that w a s held o n a popular vote basis; 203 F. Supp. 158 (1 962).

Adapted from Hynwood Holland, "The Direct Primary in Georgia" and The Changing Polirics of rhe South. William C. Howard Est. Baton Rouge: Louisiana State University. Pres. 1969.

169

Page 14: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

APPENDIX 3: List of Candidates

No. elections - 32 1926 No. of candidates = 119

'Designates winner

1898

1902

1904

1906

1908

1910

1912

1914

1916

1918

1920

1920

1922

1924

*Chandler 1926 Berner Atkinson

Data not available 1928 'Tellell

NO primary held this year 1930 *Tell el I

OH. Smith Howell Russell Estill J. Smith

- .

*Brown H. Smith

'H. Smith Brown

'Slaton Alexander Hall

1930

1932

1934 Harris Anderson Hardman

1936 'Dorsey

Hardman Harris Pottle

No primary held this year. *Dorsey 1940

Run-off

1942 Hardwick Holder Walker

'Hardwick Holder

'Walker

1946

Baylon Hardwick

No primary held this year. *Walker

1948

Carswell Hardrnan Holder Wood

Run -0 f f Hardman Holder

Hardman Rivers

Carswell Holder Perry Rivers Russell

Run-off 'Russell Carswell

*Talmadge Edwards Hardwic k Holder Nix O'Kelly Kelly Summens

OTalmadge Gillian Pittman

'Rivers Howell Mangham Wood

Talmadge Nix Roberts

Talmadge

O'Kelly Carmichael Rivers

H. Talmadge O'Kelly Rabun Thompson Willis

'Arnall

'Talmadge (Died before taking office)

170

Page 15: THE GEORGIA PRIMARY SYSTEM 1898–1970: FACTIONAL PATTERNS AND POLITICAL REGIONS

1950 'H. Talmadge Avery Baker Jenkins Thompson

Thompson 7 other candidates

Abernathy Bodenhasen

1962 'Sanders Griffin O'Kelly 2 other candidates

1966 'Maddox Arnall 4 other candidates (Maddox winner in run-off)

Sanders 7 other candidates (Carter winner in run-off)

1954 'Griffin

1958 'Vandiver

1970 'Carter

1 7 1