How the Hispanic Vote Has Changed in Modern Elections

According to an exit poll research, Hispanics voted for President Barack Obama over the Republican presidential candidate Mitt Romney by 71% to 27% and in terms of raw votes, Democrats won handily over the Republicans by a huge margin vote of 53 to 42 Percent (Sanchez, 2015). It was for this reason that a statistical analysis was carried out, specifically a regression analysis to study the actual voting patterns exhibited by Hispanics. Regression analysis is a useful statistical tool for determining the relationship between two variables. Some relationships between variables are not easily determined and therefore careful analysis of data is required in order to be certain whether there is a significant relationship between two sets of data or not. Regression analysis is, therefore, a useful method for investigating functional relationships among variables ( Chatterjee, & Hadi, 2012 ). Statistical modeling was used to run regressions separately on Hispanic views toward Obama and Romney. 

Importance of the study 

There has been a growing Latino political influence especially in some states such as New Mexico which was largely considered as a swing state for many years. The state of New Mexico, for instance, has the most number of Mexican-Americans (Hispanics) in the United States and due to their high levels of achieved political power, the state has gained the reputation of being considered a swing state in the recent elections (Wayne, 2011). The Hispanic votes have become a battleground and this is evident from the presidential elections held in the year 2000 and the year 2004. In 2000, the Democratic presidential candidate Al Gore won the Hispanic vote over his Republican rival George Bush by a mere 366 votes which are considered as the narrowest margin ever. However, in the following general elections held in the year 2004, The incumbent President George Bush won the Hispanic vote by 5,988 votes over his Democratic rival showing just how much of a swing vote the Hispanic vote could be (Sanchez, 2015). 

These facts and figures are what drove both presidential candidates Barack Obama and Mitt Romney to make several trips to the predominantly Hispanic States such as New Mexico to seek votes. According to the National Association of Elected and Appointed Officials (NALEO), the Hispanic votes were crucial to Obama's eventual victory as he won overwhelmingly compared to his rival (Sanchez, 2015). 

In order to better understand the voting patterns exhibited by the Hispanic voters, a linear regression analysis was put into use as an accurate and reliable relationship among various variables could be easily developed. It was noted that the relationship between variables such as age, income, and education had not been examined fully for the benefit of political analysis, media organizations and learners in general. 

Method 

This study was aimed at examining the relationship between the independent variables such as age, income, and education of the Hispanic voters and a simple random survey was conducted to obtain the data. Before a regression analysis is carried out, it is necessary to identify the dependent and the independent variable(s). In this study, the dependent variable is Obama/Romney rating since it is the variable that this analysis is interested in predicting. It is also known as the response variable and is mostly denoted by “Y”. The independent variables are the ages, income and education level of the Hispanic voters. This is because these variables are the variables being used to predict other dependent variables. These variables are also known as the predictor or explanatory variables and are denoted by "X" (Sen & Srivastava, 2012). The analysis was conducted using the SPSS software. 

Linear Regression Analysis 

Linear regression between Obama rating and the variable dem_hisp (whether one is Spanish, Hispanic or Latino) 

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate	Change Statistics
					R Square Change	F Change	df1	df2	Sig. F Change
1	.133 a	.018	.018	34.039	.018	98.437	1	5467	.000
a. Predictors: (Constant), PRE: R: Are you Spanish, Hispanic, or Latino

ANOVA a
Model		Sum of Squares	df	Mean Square	F	Sig.
1	Regression	114052.796	1	114052.796	98.437	.000 b
	Residual	6335214.257	5468	1158.640
	Total	6449267.053	5469
a. Dependent Variable: Obama Rating
b. Predictors: (Constant), PRE: R: Are you Spanish, Hispanic, or Latino

Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	Correlations
		B	Std. Error	Beta			Zero-order	Partial	Part
1	(Constant)	84.261	2.843		29.633	.000
	PRE: R: Are you Spanish, Hispanic, or Latino	-14.714	1.483	-.133	-9.922	.000	-.133	-.133	-.133
a. Dependent Variable: Obama Rating

Analysis 

After a regression analysis is performed, the first table obtained is the model summary table. This table contains the R, R-squared, and the Adjusted R-squared values. The R value denotes the correlation which is given as 0.133 in the R column and this signifies a low degree of correlation. This indicates that the variables Obama rating and the race of voters (whether Hispanic, Spanish or Latino) has a weak correlation and therefore whenever one variable increases or decreases, there is a low probability that there will be a relationship with the other variable under investigation. The R-square, on the other hand, indicates how much of the total variation being witnessed in the response variable, Obama rating, can be explained by the explanatory variable dem_hisp (Sen & Srivastava, 2012). In the above results, 1.8% can be explained which is very small. 

In analyzing the ANOVA table (Analysis of Variance), how best the regression equation line fits the data is obtained/realized. In order to interpret the ANOVA table, the Sig column is looked at and the value is very low; .000. This, therefore, implies that there is statistical significance and the regression model fittingly forecasts the response variable (Obama rating) significantly well. A small value here corresponds to a lower p-value (probability of making type I error) which is less than the standard α level of significance .05. 

The table for Coefficients gives the information required to predict Obama rating from voter race variable. This table simply provides the coefficients which form the linear regression equation line. The values from the B column which is under the Unstandardized Coefficients column are what are used in developing the regression equation which in this case can be given as 

Obama rating= -14.714x+84.261. 

This equation can be used to predict future voting pattern and relationship between these two variables, assuming President Obama was to run for yet another term. 

b). Multiple Linear regression between R vote, 2012 and the independent variables age, income, education and party ID 

Key 

R vote, 12- Presidential vote in the 2012 presidential elections 

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate	Change Statistics
					R Square Change	F Change	df1	df2	Sig. F Change
1	.788 a	.621	.621	.307	.621	1444.547	4	3525	.000
a. Predictors: (Constant), Party ID, Education, Age Groups, Income quintile

ANOVA a
Model		Sum of Squares	df	Mean Square	F	Sig.
1	Regression	546.144	4	136.536	1444.547	.000 b
	Residual	333.267	3526	.095
	Total	879.412	3530
a. Dependent Variable: R Vote, 2012
b. Predictors: (Constant), Party ID, Education, Age Groups, Income quintile

Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	Correlations
		B	Std. Error	Beta			Zero-order	Partial	Part
1	(Constant)	.746	.021		35.305	.000
	Age Groups	.008	.002	.054	5.216	.000	.103	.088	.054
	Education	-.017	.007	-.029	-2.571	.010	.002	-.043	-.027
	Income quintile	.010	.004	.027	2.385	.017	.113	.040	.025
	Party ID	.175	.002	.780	74.475	.000	.785	.782	.772
a. Dependent Variable: R Vote, 2012

Analysis/Results 

After a regression analysis is performed, the first table obtained is the model summary table. This table contains the R, R-squared, and the Adjusted R-squared values. The R value denotes the correlation which is given as 0.788 in the R column and this signifies a high degree of correlation (Cameron & Trivedi, 2013) . This indicates that the variables Respondents presidential vote in the elections and the independent variables age, income, education and party ID has a strong correlation and therefore whenever one variable increases or decreases, there is a high probability/chance/likelihood that there will be a similar outcome in the other variables being analyzed. The R-square, on the other hand, indicates how much of the total variation being witnessed in the response variable, R vote 12 , can be explained by the explanatory variables age, income, education and party ID . In the above results, 62.1% can be explained which is slightly high. 

In analyzing the ANOVA table (Analysis of Variance), how best the regression equation line fits the data is obtained/realized. In order to interpret the ANOVA table, the Sig column is looked at and the value is very low; .000. This, therefore, implies that there is statistical significance and the regression model fittingly forecasts the response variable R vote 12 significantly well. A small value here corresponds to a lower p-value (probability of making type I error) which is less than the standard α level of significance .05. 

The parameter estimates in the table for Coefficients gives the information required to predict R vote 12 from the independent variables age, income, education and party ID. This table simply provides the coefficients which form the linear regression equation line. The values from the B column which is under the Unstandardized Coefficients column are what are used in developing the regression equation which in this case can be given as 

Y = b0 + b1*x1 + b2*x2 + b3*x3 + b4*x4 

Age Groups - in this case, the coefficient for Age groups is .008 which means that for every u nit increase in  Age groups , a 0.008-unit increase in  R vote 12  is predicted, whilst keeping all the other analyzed variables constant. Education - For every unit increase in Education level of the Hispanic voters, there is an expected -.017 unit decrease in the R vote 12  (number of votes cast in the 2012 presidential elections. Since Obama votes were coded as 1 and Romney's votes coded as 2, the interpretation would be that of all the total votes cast based on the levels of education of the Hispanic voters, the votes cast for Mitt Romney would be .017 times lower than those cast for Barack Obama. 

Income quintile- The income level has a coefficient of .010. So, for any unit increase in the income levels of the Hispanic voters, we would expect an increase of about .01 in the votes cast in the presidential elections. Since Obama votes were coded as 1 and Romney's votes coded as 2, the interpretation would be that of all the total votes cast based on the levels of income of the Hispanic voters, the votes cast for Mitt Romney would be .01 times higher than those cast for Barack Obama. 

Party ID- the coefficient for party ID is 0.175. This, therefore, implies that for every unit increase in the number of Hispanic voters who are affiliated with a certain party, there would be a 0.175-unit increase in the total votes cast in the Obama-Romney presidential election. 

  Std. Error  – St. Err are the relevant standard errors which are associated with the individual coefficients obtained above. 

  t  and  Sig . – in these columns are the t-statistics and their associated p-values obtained in the regression analysis to test (while using α level of significance of 0.05) whether any given coefficient is significantly different from zero. 

The coefficient for Age is .008 and is statistically significant/different from 0 since it corresponds to a p-value of .000 which is lower than .05. 

The coefficient for Education is -0.17 and is statistically significant/different from 0 since it corresponds to a p-value of .010 which is lower than .05. 

The coefficient for Income is 0.10 and is statistically significant/different from 0 since it corresponds to a p-value of .017 which is lower than .05. 

The coefficient for Party ID is 0.175 and is statistically significant/different from 0 since it corresponds to a p-value of .000 which is lower than .05. 

Charts 

Interpretation 

From the histogram, the chart obtained it is evident that the distribution is normal since the shaped obtained is that of a bell-shaped distribution. The graph is neither skewed to the right or to the left but instead symmetric. The Normal P-Plot is a graphical technique of ascertaining whether or not the data being observed is normally distributed (Donnelly, & Abdel-Raouf, 2016) . Looking at the P-Plot graph, it is evident that the data points form nearly a linear line pattern and this implies that the normal distribution is an ideal model for this Hispanic voting trend data set. 

Implications and Conclusion 

The findings from this regression analysis study supported the notion that there is a relationship between the Hispanic views towards Obama and Romney, which affect voting in the presidential elections with independent variables such as age, income, education, party ID. It is evident that factors such as education level largely determined who the Hispanics would vote for in the elections as the results of the analysis showed that based on the education level alone, Obama would get more of the Hispanic votes compared to his rival Romney. Age was also a significant variable since the results showed that for every unit increase in the age of the Hispanic voter, there would be a 0.008 increase in the votes cast. 

The results of these analyses suggest that the Hispanics largely favor Democratic presidential candidates, Obama in this case. These figures, therefore, can be used in future elections to predict the Hispanic people voting patterns by use of the regression line equation obtained from this study. 

References

Cameron, A., & Trivedi, P. (2013).  Regression analysis of count data . Cambridge: Cambridge University Press. 

Chatterjee, S., & Hadi, A. (2012). Regression Analysis by Example (5th ed.). New Jersey: John Wiley & Sons. 

Donnelly, R., & Abdel-Raouf, F. (2016). Statistics (1st ed.). Indianapolis, Indiana: Alpha, a member of Penguin Random House LLC 

Sanchez, G. (2015).  Latinos and the 2012 election . East Lansing, Michigan: Michigan State University Press. 

Sen, A, & Srivastava S. Muni (2012). Regression Analysis. Chicago: Springer 

Wayne, S. (2011).  Appendix to, The Road to the White House, 2012  (9th ed.). Washington, DC: Cengage Learning.