Assignment 2: Interpreting Quantitative Findings
2764983Y
13.01.25
Word count: 2992.
Research Question: Did a person’s gender affect their income in Northern Ireland in 2012?
Introduction
Since 1987, women’s wages in Ireland have increased faster overall than men’s (Barrett et al., 2022). This factor, combined with a decline in the number of hours men work, has narrowed the gender earnings gap in Ireland, particularly at the lower end of the income spectrum. A key reason for this is the shift in societal norms. Women have entered more male-dominated fields in the past century, and the barriers to education for women have lessened, allowing women to obtain the same education and qualifications as men. Subsequently, this has helped to decrease the wage gap. However, women are still more likely to be stay-at-home parents, meaning that even if women and men enter the workforce at the same rate, Women more often leave for an extended period, sometimes not returning, to look after their families (Office for National Statistics, 2023). This prolonged absence from the workforce makes their chances of a promotion less likely, meaning that the men who enter the workforce at the same time as them can climb the company ladder quicker, subsequently increasing their salaries. Over the past couple of decades, the Northern Irish Government have introduced policies and initiatives to help improve equality within employment. For instance, in 1998, the Government imposed the Fair Employment and Treatment (NI) Order, which imposed that public bodies must aim to guarantee equal employment opportunities for many minority groups, including disability, race, and gender. Moreover, initiatives have been implemented to help recruit women in non-traditional gender roles. The Women into Non-Traditional Sectors (WINS) (2005) aimed to help deal with the horizontal segregation in the workforce, which had grouped women into lower-paid, lower-status employment. This scheme was implemented to help prepare women through training and development programmes to enter non-traditional jobs such as construction and horticulture. Giving women the same job opportunities as men should allow them to have the same income opportunities, hopefully closing the gender pay gap.
This lab report uses data from 2012 to examine the income gap between men and women, examining whether policies and economic and social changes have had an effect. The research question is, ‘Did a person’s gender affect their income in Northern Ireland in 2012?’. As more policies and initiatives are implemented, the gender wage gap should have significantly decreased; however, due to inherent societal barriers such as sexism and gender roles, this may not be the case. Unemployment in Northern Ireland peaked at 17.2% in 1986 but dropped to below 10% in the 2010s (DETI, 2014). Due to this, 2012 is an optimum period to examine the gender pay gap, as the largest number of people in the last few centuries were employed and earning an income. To further answer this question, it is broken down into two hypotheses:
H1: Men have a higher income than women
H2: Women have a higher income than men
Data and method
The dataset used in this study is the 2012 Northern Ireland Life and Times (NILT) Survey. It was provided by ARK Northern Ireland, a partnership between Ulster University and Queen’s University Belfast. ARK’s mission is to monitor the attitudes and behaviour of people in Northern Ireland and provide a time series of how attitudes and behaviour develop on various social policy issues (ARK, 2013). Overall, 1204 respondents were questioned for the survey.
The data was collected through face-to-face interviews with the respondents, which were completed using computer-assisted personal interviewing (CAPI). A benefit of this is accurate real-time results, which helps avoid human error. The respondents then completed a self-completion questionnaire using a computer-assisted self-interviewing questionnaire. This is reliable as the respondents are more likely to answer honestly, and there is no chance of forgery from interviewers. However, without an interviewer, the respondents lose clarification and quality control from the interviewer, which can help ensure that respondents understand the question and prevent incorrect answers (Bryman et al., 2021)
The dependent variable for this study is Annual Personal Income. This is measured as £ per year, computed from the midpoint of the income band before tax and National Insurance contributions. The responses have been measured on a scale. There were 897 responses to this question, which is 307 responses less than the overall survey. Respondents were chosen through systematic random sampling. Although this removes selection bias, it means that you can’t guarantee that every respondent can answer every question, meaning that you can end up with a sample that is not an accurate representation. Or have samples of unequal size for different questions (Gilbert, 2016).
Below is a Histogram and Summary Table displaying the responses:
The histogram is skewed to the right, meaning that the outliers present towards the maximum values. These outliers are respondents with high incomes, highlighting income inequality within Northern Ireland. This means that more respondents’ incomes lie in the lower ranges. When looking at a summary table of statistics of the dependent variable, it is evident that the central tendency can support this finding. The mean income is £16,395, which is low in the range of incomes on the histogram. Moreover, the median is £11,960. As the median and mean are not close together, this suggests that the data is not normally distributed. The responses are widely dispersed, with a range of £74,740, a minimum value of £260, and a maximum value of £75,000. The standard deviation is 13466. This means the respondents’ income deviates from the mean income of £16,395 by £13,466 on average (Fogarty, 2023). This is a significant deviation, highlighting how large the spread of responses is. The 25th quartile is £6,760, which means that 25% of respondents have an income equal to or less than £6,760. The 75th quartile is £22,100, meaning another 25% of respondents have an income equal to or more than £22,100. The other 50% of respondents have an income between £6,760 and £22,100. The large spread of data, with a majority in the lower range, can highlight a poor spread of a sample chosen for this society. It can also suggest the issue of income inequality in Northern Ireland, where income is unequally distributed, with a substantial proportion making less than £23k a year.
Below is a table of descriptive statistics for the variables included in the model:
| Variable | N | Mean | Std. Dev. | Min | Pctl. 25 | Pctl. 50 | Pctl. 75 | Max |
|---|---|---|---|---|---|---|---|---|
| persinc2 | 897 | 16395 | 13466 | 260 | 6760 | 11960 | 22100 | 75000 |
| rsex | 1204 | |||||||
| … Male | 537 | 45% | ||||||
| … Female | 667 | 55% | ||||||
| religcat | 1168 | |||||||
| … Catholic | 491 | 42% | ||||||
| … Protestant | 497 | 43% | ||||||
| … No religion | 180 | 15% | ||||||
| orient | 1191 | |||||||
| … I am heterosexual or straight | 1173 | 98% | ||||||
| … I am gay or lesbian (homosexual) | 14 | 1% | ||||||
| … I am bi-sexual | 2 | 0% | ||||||
| … Other | 2 | 0% | ||||||
| uninatid | 1183 | |||||||
| … Unionist | 348 | 29% | ||||||
| … Nationalist | 255 | 22% | ||||||
| … Neither | 580 | 49% | ||||||
| tunionsa | 1179 | |||||||
| … Yes | 440 | 37% | ||||||
| … No | 739 | 63% | ||||||
| rsuper | 883 | |||||||
| … Yes | 267 | 30% | ||||||
| … No | 616 | 70% | ||||||
| rage | 1201 | 50 | 19 | 18 | 35 | 48 | 64 | 97 |
The key independent variable used in the model is Sex (Female compared to the reference category Male); of the responses, 55% were female, whilst 45% were male. This is similar to the overall population of Northern Ireland, where 49% are male and 51% are female (NISRA, 2012). This is beneficial in allowing us to get an accurate sample. The control variables used in the model are Religion (Protestant and No Religion compared to the reference category Catholic), Sexual Orientation (Homosexual, Bisexual and Other compared to the reference category of Heterosexual); Constitutional View (Nationalist or Neither compared to the reference category of Unionist); Involvement in a Trade Union (No compared to the reference category yes); Supervisor (No compared to the reference category yes); and Age (measured in years). Furthermore, the control variable of age showed a similar trend to that of Northern Ireland (NISRA, 2012). The average age of the sample is fifty. The 25th quartile is 35, and the 75th quartile is 64; this means that 50% of respondents are between 35 and 64. This is beneficial in allowing us to obtain an accurate sample for age. Furthermore, middle-aged people tend to be in more continuous employment overall. This is because a large proportion of those in full-time education are younger people, and older people are more likely to be retired. Having a sample of majority middle-aged people will allow us a more accurate representation of people actively in the workforce.
Results and discussion
| Annual Personal Income (GBP) | |
| Sex: Female (ref.: Male) | -5,068.737*** |
| (994.748) | |
| Religion: Protestant (ref.: Catholic) | 465.188 |
| (1,458.367) | |
| Religion: No religion | 895.169 |
| (1,533.323) | |
| Sexual Orientation: Homosexual (ref.: Heterosexual) | -6,247.777 |
| (3,437.048) | |
| Sexual Orientation: bi-sexual | -2,826.980 |
| (8,698.806) | |
| Sexual Orientation: Other | 1,323.336 |
| (8,737.282) | |
| Constitutional View: Nationalist (ref.: Unionist) | 1,788.873 |
| (1,898.294) | |
| Constitutional view: Neither | 1,438.036 |
| (1,350.423) | |
| Trade union membership: No (ref.: Yes) | -5,277.978*** |
| (977.008) | |
| Supervisor: No (ref.: Yes) | -8,648.320*** |
| (1,037.559) | |
| Age | -84.369** |
| (29.430) | |
| Constant | 31,343.540*** |
| (2,488.067) | |
| Observations | 675 |
| R2 | .183 |
| Adjusted R2 | .170 |
| Residual Std. Error | 12,252.840 (df = 663) |
| F Statistic | 13.533*** (df = 11; 663) |
| Notes: | *P < .05 |
| **P < .01 | |
| ***P < .001 | |
The model is a multiple linear regression model showing how the independent and control variables affect Annual Personal Income (£). Unlike a simple linear regression, multiple linear regression adds control variables to ‘control for’ the influence of the other variables when examining how the chosen variable will affect someone’s annual personal income. This has two critical purposes: first, it allows us to isolate the effect of our independent variable (in this case, gender) to answer the research question accurately. Secondly, it allows us to have an unbiased model. As the independent variable is not the only factor that affects our dependent variable; if we do not include these in our model and ignore their effects, our model will be biased (Fogarty, 2023)
To understand the variables in the model, it is vital to look at five key statistics. Firstly, the P-value tells us whether or not there is a statistically significant relationship between, in this instance, an independent variable and the dependent variable of personal income. ‘Trade union membership: No (ref.: Yes)’, ‘Supervisor: No (ref.: Yes)’, and ‘Age’ all have P-values of less than 0.01, meaning their relationship with the dependent variable is statistically significant. ‘Sex: Female (ref.: Male)’ has a P-value of less than 0.001, meaning that it is statistically significant, and we can reject the null hypothesis that gender does not affect income. No other variables have a P-value of less than 0.5, meaning that according to the P-value they do not have a statistically significant relationship with the dependent variable.
Another way of determining statistical significance is using standard errors and T-test statistics. A standard error measures the amount of discrepancy that can be expected in a sample estimate compared to the actual value in the population. This can be used to find a t-test value, which, if greater than two, shows that the relationship is statistically significant. ‘Sex: Female (ref.: Male)’ has a standard error of £994.748, which is how much our point estimate of -£5,068.74 would vary in repeated sampling. The t-test value is -5.1, meaning the relationship between the variables is statistically significant.
‘Religion: Protestant (ref. Catholic)’ has a standard error of £1458.37. This results in a t-test value of 0.32, meaning the relationship with annual personal income is statistically insignificant.
‘Sexual orientation: homosexual (ref. heterosexual)’ has a standard error of £3437.05. This results in a t-test value of -1.82, which means that the relationship with annual personal income is statistically insignificant.
‘Constitutional View: Nationalist (ref.: Unionist)’ has a standard error of £1,898.294. This results in a t-test value of 0.94, meaning the relationship with annual personal income is statistically insignificant.
Trade union membership: No (ref.: Yes) has a standard error of £977.008. This results in a t-test value of -5.4, meaning the relationship with annual personal income is statistically significant.
Supervisor: No (ref.: Yes) has a standard error of £1,037.559. This results in a t-test value of -8.34, meaning the relationship with annual personal income is statistically significant.
Age has a standard error of £29.430, resulting in a t-test value of -2.87, meaning the relationship with annual personal income is statistically significant.
The results from the T-test support the P-values, as the four variables with P-values of less than 0.05 and that were statistically significant are the same four found to be significant from their t-test values. These are Sex, Age, Trade Union Membership, and Supervisor. These results are not surprising. As stated earlier, middle-aged people are more likely to be employed long-term. There is also the assumption that the longer you work, your salary will increase. So, it makes sense that there is a relationship between age and personal income. There is a relationship between Trade Union membership and income, as being a member takes a percentage of the member’s pay meaning that employees automatically take home less money than their colleagues simply by being in a Trade Union (NiDirect, 2019). Finally, supervisory roles often come with a pay increase, which answers why a positive relationship exists between supervisory and annual income.
Another key statistic is R2 and Adjusted R2, which explains how much variance our model explains in the dependent variable. The model’s R2 value is 0.183. This means that our model explains 18.3% of the variance in the dependent variable. The model explains very little of the variance in the model. Each time we add another independent variable to our model, the R2 value will increase regardless. The adjusted R2 penalises the model if it has variables in the model that do not contribute to explaining the variance in the dependent variable. If the R2 and adjusted R2 are far apart, it tells us that some of our independent variables poorly explain the dependent variable’s variance (Fogarty, 2023). For this model, the adjusted R2 is 17%. As R2 and adjusted R2 are close together, it tells us that some of our independent variables successfully explain the dependent variable’s variance.
The residual standard error shows how well our model fits compared to another model, whether there are fewer or more errors. The residual standard error is £12,252.840, which is the average amount of the actual values that differ from the predicted model. This is a significant variance, suggesting that the model might be inaccurate. The small number of responses in the sample supports this. There were 675 responses, almost half the number of responses the NILT survey had in total (NILT, 2012)
Finally, the F-test is whether our model explains the dependent variable better than a no-model or an empty model. The F Statistic is 13.533. We cannot interpret the actual F Statistic test value (13,533) by itself, so we must look at its p-value. The P-value is less than 0.05, which means that it is significant. We can reject the Null hypothesis and conclude that at least one of our independent variables explains some of the variation in the dependent variable.
The results from these statistical tests clearly show a statistically significant relationship between Sex and Annual Personal Income. This report will now examine the relationship in further detail. As gender is a categorical variable, and personal income is numeric, we will present this data as a boxplot.
## $y
## [1] "Personal Income (£)"
##
## $x
## [1] "Gender"
##
## attr(,"class")
## [1] "labels"The median is lower for women than men, which means that, on average, women’s income is lower than men’s. The 25th and 75th quartiles are also lower for women than men, making the interquartile range (IQR) smaller for women than for men. This means that there is less variety in women’s income. This could be due to fewer job opportunities and choices for women, connecting back to the issue of horizontal segregation in the job industry. Overall, the boxplot results highlight that gender did affect income in Northern Ireland in 2012. These findings support our first hypothesis:
H1: Men have a higher income than women
They show that, on average, men have a higher income than women. These findings are supported by statistical tests that show the relationship between Annual Personal Income and Gender is significant. As evidence was found to support Hypothesis one, hypothesis two can be discredited as it is the opposite:
H2: Women have a higher income than men
Conclusion
This lab report has sought to answer the research question: ‘Did a person’s gender affect their income in Northern Ireland in 2012?’. Overall, it has been found that gender does affect people’s income, with women being found to make less than men. The multiple regression model shows that women make, on average, £5,068.74 less than men. All five statistical tests completed on the model show that gender significantly relates to personal income. This supports Hypothesis One and denounces Hypothesis Two, as men have been found to have a higher income than women.
The results from this study support the previously discussed literature. Historically, women have had lower incomes due to barriers to education, fewer job opportunities, and traditional gender roles. However, societal norms have changed in the past few decades, and women have gained more job opportunities (Potter and Hill, 2009). The gendered gap in annual personal income makes sense in a historical context; however, in the present day, it needs to be looked at in a different light. This suggests some direction for future research. For instance, it could be beneficial to look at the three statistically significant control variables in relation to gender. These were age, supervisor position and Trade Union member. For example, how many men compared to women are supervisors or income by age and gender? Is there a larger income gap in entry-level work or executive-level work? Nowadays, due to gender equality laws and initiatives, the difference in pay is not due as much to outright job sexism but instead to other societal factors and norms. For example, parental roles and the effects of parental leave on employment. Do mothers still take more parental leave than fathers? Has there been a change in the number of stay-at-home parents? Furthermore, looking at the shifts in employment due to COVID-19 and inflation would be beneficial. Are more people experiencing financial struggles which give them no option but to work? Has there been an increase in remote and flexible work post-pandemic, and has this allowed parents the ability to continue working flexibly and remotely? To understand these income differences in a modern-day context, it could be beneficial to look at these areas.
Furthermore, the small sample size limits the findings. Random strategic sampling for a broad survey like this does not guarantee that every respondent can answer every question, which means that the sample size may vary for different sections. This causes an imbalance in the accuracy of responses to different questions. This is evident in the model, where the sample size was only 675 compared to the total of 1204 for the entire survey. This makes it less reliable than other sections with more responses and can result in an answer which is not representative of the population as a whole.
Overall, the NILT 2012 survey found that a person’s gender affected their income, with men generally making more than women.
Bibliography
Barrett, M., Doorley, K., Redmond, P. and Roantree, B. (2022). How Has the Gender Earnings Gap in Ireland Changed in Thirty Years? Social Sciences, 11(8), p.367.doi:https://doi.org/10.3390/socsci11080367.
Bryman, A., Foster, L., Sloan, L. and Clark, T. (2021). Bryman’s Social Research Methods. 6th ed. New York: Oxford University Press.
Davies, C. (1991). Women, employment and social policy in Northern Ireland: A problem postponed. Women’s Studies International Forum, 14(5), p.XII. doi:https://doi.org/10.1016/0277-5395(91)90066-q.
Department of Enterprise, Trade, and Investment (2014). Full Economic Overview.
Fogarty, B.J. (2019). Quantitative social science data with R : an introduction. London: Sage.
Gilbert, N. and Stoneman, P. (2016).Researching social life. Los Angeles ; London ; New Delhi ; Singapore ; Washington Dc Sage.
Legislation.gov.uk. (2011). The Fair Employment and Treatment (Northern Ireland) Order 1998. [online] Available at: https://www.legislation.gov.uk/nisi/1998/3162/contents/made.
NiDirect (2019). Introduction to trade unions | nidirect. [online] nidirect. Available at: https://www.nidirect.gov.uk/articles/introduction-trade-unions.
Northern Ireland Statistic and Research Agency (2011). 2011 Census. [online] Northern Ireland Statistics and Research Agency. Available at: https://www.nisra.gov.uk/statistics/census/2011-census.
Office for National Statistics (2023). Families and Households in the UK - Office for National Statistics. [online] Office for National Statistics. Available at: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/families/bulletins/familiesandhouseholds/2022.
Potter, M. and Hill, M. (2009). Women into non‐traditional sectors: addressing gender segregation in the Northern Ireland workplace. Journal of Vocational Education & Training, 61(2), pp.133–150. doi:https://doi.org/10.1080/13636820902933239.