Gender and police trust are categorical variables. Gender has ‘male’ and ‘female’ levels. The levels in trust police are ‘not at all’, ‘just a little’, ‘somewhat’ and ‘a lot’ when the participants recorded their level of trusting the police. In this analysis, we would like to test independence in the two categorical variables. We want to find out whether males and females differ in their levels for police trust. Since our variables are nominal, the chi-square test will be our best test. Chi-square is used to test for a relationship between two nominal variables, with each variable having two or more groups (Turner, 2014).
Research questions and test hypothesis
Our primary research question is whether males and females differ in their levels for police trust. The follow-up question would be to ask how the variables are dependent on one another if at all there is a significant relationship. We set our null and alternative hypotheses as;
Males and females do not differ in trust levels for police
Males and females differ in trust levels for police
Chart 1; clustered bar chart
From (chart 1), those who reported to trust the police somewhat, record the highest number across the levels. Therefore the chart shows the differences between the levels of trust in the police (McHugh, 2013).
Table 1; cross-tabulation table
|Q59h. Trust police * Q101. Gender of respondent Cross tabulation|
|Q101. Gender of respondent||Total|
|Q59h. Trust police||Not at all||Count||5803||5454||11257|
|Just a little||Count||6345||6275||12620|
The (Table 1) cross-tabulation table displays the observed and expected counts. The expected count refers to the values that we will expect to observe if the test is not significant (Michael, 2001). For example, if gender does not relate to the level of trusting the police, we expect to observe 6696 males who trust police a lot. The expected and observed values are a little bit different. However, the difference seems to be very small. We will, therefore, look at the chi-square table to determine if the difference is significant (Turner, 2014).
Table 2; chi- square table
|Value||df||Asymp. Sig. (2-sided)|
|N of Valid Cases||50485|
|a. 0 cells (0.0%) have expected count less than 5. The minimum expected count is 5600.74.|
The footnote under (Table 2) shows that 0.0% of cells have expected count less than 5. The latter means that the assumption of having at least 80% of the expected count to have more than five has not been violated (McHugh, 2013). Looking at the Pearson chi-square value, Since p-value (sig) < α=0.05, we reject the null hypothesis. We, therefore, deduce that there is a significant association between gender and level of trusting police. Males and females differ in their levels of trusting police.
Table 3; symmetric measures table
|Nominal by Nominal||Phi||.016||.006|
|N of Valid Cases||50485|
|a. Not assuming the null hypothesis.|
|b. Using the asymptotic standard error assuming the null hypothesis.|
(Table 3) Shows the size of the effect is (0.016). The latter shows a very weak association between gender and trust levels in the police. The significance for this effect is p-value (sig) =0.006 < 0.05. We, therefore, conclude that there is a positive and very weak association between gender and police trust. Males and females a little bit differ in trusting the police.
McHugh, M. L. (2013). The chi-square test of independence. Biochemia medica: Biochemia medica, 23(2),
Michael, R. S. (2001). Crosstabulation & chi square. Indiana University, Bloomington, IN. URL http://www. indiana. edu/~ educy520/sec5982/we ek_12/chi_sq_summary011020. pdf (Visited 2010, June 15).143-149.
Turner, G. (2014). Is it statistically significant? The chi-square test. In UAS Conference Series.