This was a methodological critique of a study called Race, Ethnicity, and Political Participation: Competing Models and Contrasting Explanations by Jan E. Leighley and Arnold Vedlitz. This was written for a polisci research methods class. There has been some discussion between myself and other individuals on dailykos about some of the issues involved, and I have posted this by request.
Methodological Critique - "Race, Ethnicity, and Political Participation: Competing Models and Contrasting Explanations"
By: Nathan Jaco
In "Race, Ethnicity, and Political Participation: Competing Models and Contrasting Explanations" Jan E. Leighley and Arnold Vedlitz explore the causes for the differences in political participation between Anglos and African-Americans, Mexican-Americans, and Asian-Americans. They contend that there are empirical limitations in the models developed by other researchers in this area as those researchers merely compared Anglos and African-Americans and assumed that data about the African-American populations would apply to other minority groups. Their assertion is that though the models they use apply differentially to the individual groups that they constitute a valid account for the differences in political participation among ethnic groups.
Socio-economic status (SES) is widely regarded as an important, if not primary, determinant of political participation and it has been held that this was also a primary determinant in the level of African-American participation and hence, discrepancies in the participation level of African-Americans was due to discrepancies in SES. However, this model was found lacking when it was determined that African-Americans participated at higher levels than Anglos when controlling for SES. So, a different model was needed to account for the greater efficacy of the African-American population in spite of the fact that that population had lower SES and SES is regarded as having a positive effect on efficacy. A number of alternative models were developed and tested, such as the impact of political empowerment or the local social context, but Veighley and Vedlitz hold that none of the studies conducted were comprehensive enough, meaning they included variate relationships from all the competing models, to determine the significance on one model relative to the significance of all other rival alternative models. Futhermore, until the late Nineties none of the published studies about ethnicity and political behavior included other minority groups.
After finding the current state of the research to be lacking, Leighley and Vedlitz determined to test the validity of five competing participation theories against one another in a single experiment across four prominent ethnic groups. The five theories tested are:
1. Socioeconomic Status: The basic assumptions of SES is that higher levels of SES aspects -
a. Education
b. Income
c. Occupational Status
- facilitates participation via "psychological orientations." Factors of the SES model apply differentially to the various respective ethnic groups. SES is not determined to work in a exclusively positive or negative way across ethnic groups, perhaps because minorities realize a lack of political empowerment which negates the characteristically positive effect on efficacy that typifies the response of Anglos with higher SES.
2. Psychological Resources: Concepts of this model include -
a. Political Interest
b. Political Efficacy
c. Trust in Government
d. Civic Duty
This study tested both group-oriented and system-oriented psychological indicators to determine the influence on participation.
3. Social Connectedness: Involves the relationship of the individual to mass society and concepts like:
a. Organizational Involvement
b. Church Attendance
c. Home Ownership
d. Marital Status
When individuals feel connected to a larger social context they are more inclined to participate in the political process.
4. Group Identity/Consciousness: How much an individual identifies with a group of which they are a representative is said to affect the participation of minorities more than Anglos. The aspects of Group Identity/Consciousnes are:
a. Group Closeness
b. Out-Group Closeness
These are anticipated to have a positive impact in the participation levels of less enfranchised segments of the population.
5. Group Conflict: Ironically the most controversial model because of contradictory data sets demonstrating positive impact on Anglo participation using aggregate level data and negative impact on Anglo participation using individual level data in samples reacting to pressures from areas where minority ethnicities represent majority populations. Concepts include:
a. Division of Resources (Competition)
b. Application of Political Power
c. Fundamental Cultural Values
d. In-Group Identification
e. Out-Group Hostility
The four ethnic groups are:
1. Anglos
2. African-Americans
3. Mexican-Americans
4. Asian-Americans
The novelty of this study is that it simultaneously tests the validity of the competing ethnicity-participation models relative to one another among the four major ethnic groups. The experiment was conducted in Texas using a statewide public opinion survey which is geographically appropriate in that it is a very large area, yet may be sociographically inappropriate because it is more ethnically diverse than most other states and thus the minority groups have been oversampled (the data analyzed and presented should be weighted for this against the size of each community in the overall U.S. population, however an additive index is used rather than a weighted index) and that may affect factors of the Social Connectedness, Group Identity, and Group Conflict models. The telephone survey was conducted between 29 Oct 1993 and 23 Feb 94 using random digit dialing and this sample design was developed to avoid biasing the overall sample though minorities were largely oversampled. Furthermore, the ethnicity of the household was determined by self-report for the data rather than the spatial (meaning implicitly the geographic locality for African- and Mexican-Americans and the surname listed for the household for Asian-Americans) origin of the sample. The dependent variables were the level and type of political participation that was reported. The first participation measure was an additive index of reported political activities over a 12 month period with overall scores coded as 0 (no activities reported) through 8 (individual engaged in all political activity). This index combines voting and other political activities and is valid since they are all measures of the underlying variable participation. However, some activities are more substantial than others in terms of their impact and that could be a distinction in level of participation. Also, if the amount of time invested in one activity by one individual is greater than the amount of time invested in the sum of political activities by another individual then it could be said that they are more involved, this study simply uses dummy variables looking at whether or not the units of analysis engage in different types of political activity. Furthermore, there is certainly a distinction in level of involvement between someone merely going to vote on election day and someone who takes the time to research the issues and candidates and makes a well-informed decision when he/she votes. Plus, this study has the same essential problem that much of the research in social science does in that it relies on self-reports, hence all it really tells is how much different ethnic groups report engaging in various political activities. The results of the index are (in descending order from greatest reported participation to least reported participation):
1. Anglos
2. African-Americans
3. Mexican-Americans
4. Asian-Americans
For their empirical measures, Leighley and Vedlitz operationalized the 5 models into the following variables
1. Socioeconomic Status
a. Education: a demographic indicator
b. Income: a demographic indicator
2. Psychological Resources
a. Political Interest: self-reported 4-point scale measuring interest in public affairs
b. Political Efficacy: individuals perceived influence on public affairs
3. Social Connectedness (the problem with the variables for this model is that none of them necessarily facilitate social ties to a community since an individual's tendency to develop social ties is highly dependent on their personality and those of their neighbors as well as a myriad of other complex personal factors)
a. Marital Status: binary dummy variable with 1 representing being married
b. Length of Residence: years lived in a community
c. Home Ownership: binary dummy variable with 1 representing owning a home
4. Group Consciousness
a. Group Closeness: 5-point scale of self-reported closeness to members of the same ethnicity; 5 being "very close" and 1 being "not close at all"
b. Intergroup Distance: 5-point scale of self-reported closeness to the other 3 ethnic groups; 5 being "very close" and 1 being "not close at all" (one does not understand why this is referred to as Intergroup "Distance" rather than "Closeness" as the polarity of the scale is not reversed from the Group Closeness variable)
5. Group Conflict
a. percentage of population comprised of other ethnic groups from the individual's own in their zip code (the zip codes used are self-reported and were matched with 1990 and 1993 census data and those with new zip codes were reported as missing rather than included using their survey reported zip code, this reduced the number of cases N and caused an approximately 8% reduction in the sample size between some of the tables presented; an analysis was replicated by the researchers excluding the threat variable with no substantive differences, however a reduction in N can cause a reduction in the statistical significance because it directly affects the calculation of degrees of freedom); the indicator of acculturation used is whether the respondent reports English as the primary language of use in their household, however political and culturally significant discourse about America can certainly take place in another language and it is conceivable for a family that speaks English as a primary household to not have attained a degree of citizenship that allows them to participate in every type of political activity.
Prior to the present study, the consensus hypothesis was that when specifically controlling for SES, Psychological Resources, and acculturation African- and Mexican-Americans were as likely to vote than Anglos and Asian-Americans were less likely to vote than Anglos. This hypothesis was based on studies that used dummy variables for each minority group represented, primarily African-Americans since the other minority ethnicities were not considered in every study, and compared the polarity of the coefficient estimates to Anglos (used as a reference group, not a dummy variable) to reach their conclusions. The present study contends the validity of the consensus hypothesis on the basis that they are based on the assumption "that in estimating the theoretically relevant variables, the population parameters are equal across ethnic groups (1101)." (Admittedly, Leighley and Vedlitz do the same thing in this study with the Asian-American population, though at the time the Asian-American population of Texas consisted primarily of Chinese and Vietnamese; Asian-American was a self-reported parameter, and individuals were given the opportunity to specify an alternative ethnicity, yet the authors do not say what they did with that data which is somewhat suspect, but the primary issue is intergroup applicability rather than intragroup applicability and this will not necessarily skew the validity to a significant degree) This assumption may be false and parameters may not be coequally applicable across ethnic groups but to get an empirical understanding of how the parameters apply across these ethnic groups is beyond the scope of this study. One cannot determine how the coefficients were weighted for parametric differences in cross-ethnic applicability without the authors knowing specifically the difference in how, for example, income would explicitly affect the participation level of African-Americans vis-a-vis that of Anglos. Simply testing the participation models separately, as the authors do, does not determine how the specific parameters differentially apply to the respective ethnic groups.
Since this study compares competing models, it is appropriate that they use model specification, or multiple regression to assess the validity of each model. The authors use OLS regression to test each theory independently against each individual ethnic group. The models are specified accurately enough and there are no indicators of highly significant error in the measurement of the variables, which are interval level. However, there are correlations among some of the models and some of the variables in the same model. SES positively affects political efficacy and political interests, which are both variables in the Psychological Resources model. Involvement in political organizations is a concept in the Social Connectedness model though it is not used as a variable, perhaps because it is one of the parameters for the dependent variable, though for a fair test of the model an important concept should not be left out. Group Closeness and Intergroup Distance could be correlative, quite plausibly, because the more closeness a unit of analysis feels to their own ethnicity will most likely affect the distance they feel to other ethnicities because a greater proportion of members of their social network will be their own ethnicity as opposed to others. One cannot determine there to be any perfect multicollinearity, yet there is arguably quite a bit of correlation that might affect the reliability of the multiple regression analysis. The SES and Psychological Resources models are both strongly supported as indicators of political participation, which does not differentiate the correlation that is anticipated between them by this analysis. The Social Connectedness model receives respectable support by these tests. The Group Consciousness model receives little support, though Intergroup Distance is significant for African- and Asian-Americans. The Group Conflict model is partially supported. With regards to the Group Threat variable, the present study clearly refutes the studies of Giles et al. but may suffer from reliability problems and is counter-intuitive. This study says that larger out-group populations have a negative effect on the participation of Anglos. This may be due to the fact that in a lot of areas where Anglos reside with larger out-group populations there is lower SES for that Anglo population. Though this is a problem with reliability, the present study uses individual level data while the Giles et al. studies use aggregate data for drawing conclusions about individual behavior, which runs the risk of ecological fallacy, so the impact on the results is probably less significant in this study than that in the other studies. Yet, as stated before, this seems to be counter-intuitive because there is currently a racist and xenophobic blowback to affirmative action and immigration policies and an influx of out-group populations that is occurring in many states, particularly California. The conclusions of this study, drawn from using individual level data, may only be applicable to the local level. Aggregated data may give different responses because the Group Conflict model may apply differentially to more or less mass levels of social organization.
Leighley and Vedlitz develop an interactive model which compares the overall affect of each variable from the 5 models on 4 different, specific parameters of the political participation/involvement dependent variable:
1. Participation
2. Voting
3. Contacting
4. Contributing
This is designed to determine the impact of each parameter on each group, where the minority groups are used as dummy variables and the Anglo group is used as a comparative index. This was designed to compensate for the parametric differentials among the categories of units of analysis and, as aforementioned, is not satisfactory in this regard and does not account for the correlation between some of the variables, both inter- and intra-model. This study challenges prior consensus hypotheses about ethnicity and political participation as well as intuitive percepts about the issue. The methods are satisfactory enough to lend some credence to the relevance of the data developed. The particularly counter-intuitive results of the Group Conflict model should, however, be replicated in order to determine their validity on different levels of social organization.
Bibliography
Leighley, Jan E. and Arnold Vedlitz. 1999. Race, Ethnicity, and Political Participation:
Competing Models and Contrasting Explanations. The Journal of Politics. Vol. 61, No. 4. November 1999. 1092-1114.