The aim of this study is to know if there is a relationship between the race and the perception of a change in the state of the economy in the US a year before the 2012 elections. This was the last year of the first Obama presidential term. As Obama was the first black US president, and his Democratic political speech was particularly addressed to minorities, like black and Hispanic people, question arises if these people have perceived the changes in the economy differently from white people, specially just before the elections in 2012, where their vote was so important.
Data come from the American National Elections Studies (ANES: www.electionstudies.org). These studies, supported by the National Science Foundation, Stanford University, and the University of Michigan, have been conducted since 1948 in the years of presidential elections, before and after the elections, to have a better knowledge about the eligible voters, principally their voting behaviour and their opinion about important questions.
Only the ANES 2012 Time Series Study (www.electionstudies.org/studypages/anes_timeseries_2012/anes_timeseries_2012.htm) is used in the present study. This survey was conducted in the US two months before the presidential elections in 2012 and two months later. Data were collected by face-to-face interviews or by internet, using web-based questionnaires. It was the first time that on-line resources were used by ANES, and this permitted a larger sample than in the previous studies, 5916 respondents. In fact, data analyzed here are an adaptation of these data (http://bit.ly/dasi_anes_data), with a total of 5914 cases and 205 variables.
The variables analyzed are race or ethnic group (DEM_RACEETH in ANES terminology) and opinion about the state of the economy in the US in the previous twelve months (CSES_ECON), both of them collected during two months before the elections in November 2012. Both variables are categorical, race with four categories (white non-Hispanic, black-non-Hispanic, Hispanic and other non-Hispanic) and the economy opinion with three (gotten better, stayed the same and gotten worse).
The population of interest is made up by all the people in the US who could vote in the 2012 presidential elections (US citizens age 18 or older). Nevertheless, as the questionnaires are written only in English or Spanish, it could be a little bias in this sense.
The study is observational because the variables of interest are measured in some subjects selected among the population. As in this case researchers do not assign treatments randomly to the subjects, the division into groups is not controlled. For this reason, causation cannot be established, so we will just talk about association or not between race and opinion about economy.
Results are supposed to be generalizable, because it is an important survey developed in the US to be used in order to know the opinion of voters. So individuals will have been selected representing adequately the different people present in the population. Indeed, the ANES data included sufficient number of minorities (black and Hispanic people) to enable a study of these groups.
The first step in summarizing the information about the relationship between race and economy opinion is the contingency table:
table(anes$cses_econ, anes$dem_raceeth)
##
## White Non-Hispanic Black Non-Hispanic Hispanic
## Gotten Better 924 439 272
## Stayed The Same 1044 377 421
## Gotten Worse 1281 134 224
##
## Other Non-Hispanic
## Gotten Better 106
## Stayed The Same 112
## Gotten Worse 129
As we want to ascertain whether the opinion about the change in the economy in the last year varies according to race or ethnic group, a suitable graphic is the mosaic one:
mosaicplot(anes$dem_raceeth ~ anes$cses_econ, main = "Race and opinion on the economy",
color = TRUE, las = 2, cex.axis = 0.8, xlab = "race", ylab = "Opinion on the economy")
The corresponding percentages of people thinking that the US economy in the last year has gotten better, stayed the same, and gotten worse, by ethnicity are:
WhiteNonHispanic <- table(anes$cses_econ, anes$dem_raceeth)[, 1]/sum(table(anes$cses_econ,
anes$dem_raceeth)[, 1])
BlackNonHispanic <- table(anes$cses_econ, anes$dem_raceeth)[, 2]/sum(table(anes$cses_econ,
anes$dem_raceeth)[, 2])
Hispanic <- table(anes$cses_econ, anes$dem_raceeth)[, 3]/sum(table(anes$cses_econ,
anes$dem_raceeth)[, 3])
OtherNonHispanic <- table(anes$cses_econ, anes$dem_raceeth)[, 4]/sum(table(anes$cses_econ,
anes$dem_raceeth)[, 4])
cbind(WhiteNonHispanic, BlackNonHispanic, Hispanic, OtherNonHispanic)
## WhiteNonHispanic BlackNonHispanic Hispanic
## Gotten Better 0.2844 0.4621 0.2966
## Stayed The Same 0.3213 0.3968 0.4591
## Gotten Worse 0.3943 0.1411 0.2443
## OtherNonHispanic
## Gotten Better 0.3055
## Stayed The Same 0.3228
## Gotten Worse 0.3718
Accordingly, the most frequent opinion among white non-Hispanics and other non-Hispanic was that the economy was getting worse, whereas almost half of black non-Hispanics believe that it was getting better. Among the Hispanics, the most frequent opinion was that the state of the economy in the US was the same as a year before.
Expressed in figures, 39.42% of the white non-Hispanics interviewed thought that the US economy had gotten worse versus 14.11% of the black non-Hispanics and 24.43% of Hispanics. In contrast, 28.44% of white non-Hispanics interviewed thought that the US economy had gotten better versus 46.21% of the black non-Hispanics. These data suggest a relationship between ethnicity and opinion about the change in the economy.
As we want to evaluate the relationship between two categorical variables, both with more than two categories, the appropriate method is the chi-square test of independence.
The null hypothesis is that opinion about the change in the economy in the last year and ethnicity are independent. So the alternative hypothesis says that there is a relationship between these two variables.
The conditions for the application of the independence chi-squared test are verified: the sample size is big enough, each expected frequency is at least 5 (see next R calculations), each case belongs only to one cell of the contingency table and the observations are nearly independent (although the sampling is without replacement, the sample size is less than 10% of the population size).
indeptest <- chisq.test(anes$cses_econ, anes$dem_raceeth)
indeptest$expected
## anes$dem_raceeth
## anes$cses_econ White Non-Hispanic Black Non-Hispanic Hispanic
## Gotten Better 1035 302.8 292.2
## Stayed The Same 1162 339.8 328.0
## Gotten Worse 1051 307.5 296.8
## anes$dem_raceeth
## anes$cses_econ Other Non-Hispanic
## Gotten Better 110.6
## Stayed The Same 124.1
## Gotten Worse 112.3
indeptest
##
## Pearson's Chi-squared test
##
## data: anes$cses_econ and anes$dem_raceeth
## X-squared = 286.8, df = 6, p-value < 2.2e-16
The chi-square statistic, which compares the observed frequencies and the expected frequencies if the null hypothesis (independence) is true, takes the value 286.8077. Taking into account the degrees of freedom (6=2*3), the p-value is P(chisquared_6>286.8077)=2.2e-16. So, if race and opinion were independent, the probability of obtaining a value of the chi-statistic at least as extreme as that obtained with the data would be very small. Consequently, there is significant evidence against the null hypothesis of independence, and it is rejected.
The conclusion is that there exists a significant association between ethnicity and perception of the change in the economy. To have a better understanding about which ethnic groups have significant related to which opinions, a logit regression could be carried out. However, in view of the conditional percentages in the exploratory analysis, it seems that black people had the most optimistic view of the changes in the American economy during the last year of the first Obama presidential term (approximately half of them thought that the economy had been getting better), whereas white non-Hispanic and other non-Hispanic people had the most pessimistic (the most frequent opinion among them was that the economy had been getting worse). Hispanics were in the middle, with the greatest percentage thinking that nothing had changed.
Nevertheless, as this is an observational study, we cannot conclude causality, that is, that race determines the opinion about changes in the economy. For example, the level of income could be a possible confounder, as it is feasible that ethnicity and opinion about the economy changes could be related with income.
The American National Election Studies (ANES; www.electionstudies.org). The ANES 2012 Time Series Study [dataset]. Stanford University and the University of Michigan [producers]. Adaptation from http://bit.ly/dasi_anes_data.
ANES. 2014. User's Guide and Codebook for the ANES 2012 Time Series Study. Ann Arbor, MI and Palo Alto, CA: the University of Michigan and Stanford University.
dataproject <- subset(anes, select = c(dem_raceeth, cses_econ))
print(dataproject[1:100, 1:2])
## dem_raceeth cses_econ
## 1 Black Non-Hispanic <NA>
## 2 Black Non-Hispanic Gotten Better
## 3 Black Non-Hispanic Gotten Better
## 4 Black Non-Hispanic Gotten Better
## 5 Black Non-Hispanic Gotten Better
## 6 Black Non-Hispanic Stayed The Same
## 7 Black Non-Hispanic Stayed The Same
## 8 Black Non-Hispanic Gotten Better
## 9 Black Non-Hispanic Stayed The Same
## 10 White Non-Hispanic Gotten Worse
## 11 White Non-Hispanic Gotten Better
## 12 White Non-Hispanic Stayed The Same
## 13 White Non-Hispanic Stayed The Same
## 14 White Non-Hispanic Stayed The Same
## 15 White Non-Hispanic Gotten Worse
## 16 White Non-Hispanic Stayed The Same
## 17 White Non-Hispanic Gotten Worse
## 18 White Non-Hispanic Gotten Worse
## 19 White Non-Hispanic Stayed The Same
## 20 Other Non-Hispanic Gotten Worse
## 21 White Non-Hispanic Stayed The Same
## 22 White Non-Hispanic Stayed The Same
## 23 White Non-Hispanic Stayed The Same
## 24 White Non-Hispanic <NA>
## 25 White Non-Hispanic Stayed The Same
## 26 White Non-Hispanic <NA>
## 27 White Non-Hispanic Stayed The Same
## 28 White Non-Hispanic Gotten Worse
## 29 White Non-Hispanic Stayed The Same
## 30 White Non-Hispanic Gotten Better
## 31 Black Non-Hispanic Gotten Better
## 32 Other Non-Hispanic Stayed The Same
## 33 Black Non-Hispanic Stayed The Same
## 34 Hispanic Stayed The Same
## 35 Hispanic Gotten Better
## 36 Hispanic Gotten Worse
## 37 Hispanic Stayed The Same
## 38 Hispanic Gotten Worse
## 39 Hispanic Stayed The Same
## 40 Hispanic Stayed The Same
## 41 Other Non-Hispanic Stayed The Same
## 42 White Non-Hispanic Stayed The Same
## 43 Hispanic Gotten Better
## 44 White Non-Hispanic Gotten Worse
## 45 White Non-Hispanic Gotten Worse
## 46 Other Non-Hispanic Gotten Worse
## 47 White Non-Hispanic Stayed The Same
## 48 White Non-Hispanic Gotten Worse
## 49 White Non-Hispanic Gotten Worse
## 50 Other Non-Hispanic Gotten Worse
## 51 Hispanic Gotten Better
## 52 White Non-Hispanic Stayed The Same
## 53 Hispanic Gotten Worse
## 54 White Non-Hispanic Stayed The Same
## 55 Other Non-Hispanic Gotten Worse
## 56 Hispanic Gotten Better
## 57 Hispanic Gotten Worse
## 58 Hispanic Gotten Worse
## 59 White Non-Hispanic Gotten Better
## 60 White Non-Hispanic Gotten Better
## 61 White Non-Hispanic Gotten Worse
## 62 Other Non-Hispanic Gotten Better
## 63 White Non-Hispanic Gotten Better
## 64 White Non-Hispanic Stayed The Same
## 65 White Non-Hispanic Stayed The Same
## 66 Other Non-Hispanic Gotten Worse
## 67 White Non-Hispanic Gotten Better
## 68 Hispanic Gotten Better
## 69 Hispanic Gotten Better
## 70 Other Non-Hispanic Gotten Worse
## 71 Hispanic Stayed The Same
## 72 Hispanic Gotten Better
## 73 White Non-Hispanic Gotten Worse
## 74 Hispanic Gotten Worse
## 75 White Non-Hispanic Stayed The Same
## 76 Hispanic Gotten Worse
## 77 White Non-Hispanic Gotten Worse
## 78 White Non-Hispanic Gotten Worse
## 79 White Non-Hispanic Stayed The Same
## 80 White Non-Hispanic Gotten Worse
## 81 Hispanic Gotten Better
## 82 White Non-Hispanic <NA>
## 83 White Non-Hispanic Gotten Better
## 84 Other Non-Hispanic Stayed The Same
## 85 White Non-Hispanic Gotten Worse
## 86 White Non-Hispanic Stayed The Same
## 87 Black Non-Hispanic Stayed The Same
## 88 White Non-Hispanic Gotten Worse
## 89 Other Non-Hispanic Gotten Worse
## 90 Other Non-Hispanic Stayed The Same
## 91 White Non-Hispanic Gotten Worse
## 92 White Non-Hispanic Gotten Worse
## 93 White Non-Hispanic <NA>
## 94 White Non-Hispanic Gotten Better
## 95 Black Non-Hispanic Gotten Worse
## 96 Black Non-Hispanic Stayed The Same
## 97 White Non-Hispanic Stayed The Same
## 98 White Non-Hispanic Stayed The Same
## 99 Other Non-Hispanic Gotten Worse
## 100 White Non-Hispanic Gotten Better