Assignment 2: Interpreting Quantitative Findings

Word count: 3171

Introduction

The relationship between sexual orientation and personal economic outcomes represents a critical yet underexplored facet of socio-economic research, particularly in the context of Northern Ireland. In an era characterized by an escalating consciousness of LGBTQ+ rights and issues, the exploration of how sexual orientation influences annual personal income transcends economic analysis alone, evolving into a crucial question of social equality (Duggan 2017). Furthermore, scholars have highlighted that even after the Equality Act 2010, there is evidence to support that pay disparity exists among different sexual orientations (Bryson 2016). This type of study is therefore integral in a global landscape increasingly striving towards a paradigm that both embraces and celebrates diversity. My analysis seeks to address the notable gap in existing literature by investigating the relationship between sexual orientation and personal income levels within Northern Ireland, a region with its own unique socio-political and cultural dynamics.

This analysis utilizes data from Northern Ireland to investigate the potential impact of sexual orientation on annual personal income. Northern Ireland presents a unique case study due to its distinctive historical and cultural landscape, particularly as it was the last region within the UK to decriminalize same-sex activity (McCormick & Stewart 2020). This historical context could have profound implications for the economic opportunities and societal attitudes towards individuals of varying sexual orientations in the region. Furthermore, the period between 2010 and 2012 saw a notable increase in hate incidents related to sexual orientation, rising from 175 to 201 cases as reported in 2014 (PSNI 2012). This rise underscores the ongoing relevance and urgency of addressing sexual orientation-related issues in Northern Ireland, making it a pertinent area for exploring the intersection of economic outcomes and sexual orientation. The research question central to this analysis is therefore: “Does sexual orientation affect annual personal income in Northern Ireland?” This question aims to uncover potential income disparities that could be attributed to sexual orientation, thereby contributing to a broader understanding of income inequality and its drivers. Furthermore I will use weekly hours worked as a second independent variable, in order to potentially expose pay disparity per hour of work.

Hypotheses-

Null Hypothesis (H₀): Sexual orientation does not have a significant impact on annual personal income in Northern Ireland.

This implies that there are no differences in the income levels of individuals based on their sexual orientation per hour of work.

Alternative Hypothesis (HA): Sexual orientation has a significant impact on annual personal income in Northern Ireland.

This would suggest that the income levels of individuals vary based on their sexual orientation, potentially indicating the presence of systemic biases or disparities in the job market, leading to inequalities in pay per hour of work.

The study will employ statistical analysis to test these hypotheses, with a particular emphasis on examining the differences in annual personal income, with hours worked per week accounted for, among different sexual orientation groups. The findings of this research could have profound implications for potential future policy development, workplace practices, and the promotion of diversity and inclusion in Northern Ireland’s economic and social landscape.

Data and method

The dataset used in this study is the Northern Ireland Life and Times (NILT) survey 2012, with credit to ARK. This social policy information hub in Northern Ireland is a collaborative initiative jointly managed by both Ulster University and Queen’s University Belfast. (ARK 2013). Data collection for this dataset was conducted from October 1, 2012, to January 10, 2013. During this period, interviews were completed with 1,204 adult participants aged 18 and over. The data was collected via interviews which were conducted using both ‘Computer Assisted Telephone Interviews (CATI)’ and ‘Computer Assisted Personal Interviews (CAPI) through Microsoft Teams video calls’. Additionally, interviewers were aso conducted through face-to-face visits to selected households to promote participation in the Survey (ARK 2013).

The dependent Variable-

In this study, we focus on personal income as the dependent variable, representing the sum of all income streams received by an individual over the course of a year. This includes but is not limited to earnings from employment such as wages and salaries, profits from business ventures, dividends from investments, and any other financial gains. The use of personal income as the dependent variable allows us to explore the effects of these various factors on an individual’s economic status. Our aim is to identify trends and income disparities for differing sexual orientations, shedding light on the broader picture of financial health and the impact of specific socio-economic variables on income levels.

install.packages("kableExtra", dependencies = TRUE)
devtools::install_github("kupietz/kableExtra")

## 
## ── R CMD build ─────────────────────────────────────────────────────────────────
##   
   checking for file ‘/tmp/RtmpaDD77y/remotes51535764130/kupietz-kableExtra-3bf9b21/DESCRIPTION’ ...
  
✔  checking for file ‘/tmp/RtmpaDD77y/remotes51535764130/kupietz-kableExtra-3bf9b21/DESCRIPTION’
## 
  
─  preparing ‘kableExtra’:
##    checking DESCRIPTION meta-information ...
  
✔  checking DESCRIPTION meta-information
## 
  
─  checking for LF line-endings in source and make files and shell scripts
## 
  
─  checking for empty or unneeded directories
## ─  building ‘kableExtra_1.3.4.9000.tar.gz’
## 
  
   Warning: invalid uid value replaced by that for user 'nobody'
## 
  
   
## 

nilt <- read_sav("data/nilt2012.sav") 
nilt <- nilt %>% mutate(orient = as_factor(orient)) 
nilt_subset <- select(nilt, orient, persinc2) 
install.packages('vtable') 
library(vtable) 
hist(nilt_subset$persinc2)

The distribution of the historgram is right-skewed, as the tail on the right side of the histogram is longer than the left side. This means there are a few very high values stretching out to the right and a small amount of potential outliers that report significantly higher incomes than most of the data. Furthermore, the bulk of the data is clustered on the left side of the histogram, suggesting that the median and mode are likely lower than the mean due to the rightword-skew, with there being an extremely large range in the dataset spanning from £0-60000.

nilt <- read_sav("data/nilt2012.sav") 
nilt <- nilt %>% mutate(orient = as_factor(orient)) 
nilt_subset <- select(nilt, orient, persinc2) 
install.packages('vtable') 
library(vtable) 
sumtable(nilt_subset, vars = c('persinc2'), group = 'orient')

Summary Statistics

orient	I am heterosexual or straight			I am gay or lesbian (homosexual)			I am bi-sexual			Other
Variable	N	Mean	SD	N	Mean	SD	N	Mean	SD	N	Mean	SD
persinc2	877	16384	13544	14	17921	10442	2	11960	7354	2	20670	2022

From the summary table above we can deduce there are 877 observations for heterosexual or straight individuals, 14 for gay or lesbian individuals, 2 for bi-sexual individuals, and 2 categorized as Other. The average value of personal income for heterosexual or straight individuals is 163, for gay or lesbian individuals it is 179, for bi-sexual individuals it is 1196 (which is notably higher but this is likely due to the small sample size of only 2 individuals), and for Other it is 206. This analysis shows that on average, bi-sexual individuals show an unusually high average, which could be accounted for due to the fact that the sample size is very small compared to the other categories. The standard deviations also suggest there is considerable variability in personal income within each category, though this may be skewed by outliers or the distribution of the data, however this again could be attributed to the largely varied sample sizes. However, despite this small sample size, these results do fall in line with UK averages, with the UK 2012 census showing that around only 1.5% of people in the UK identify as gay or bisexual (UK Census 2012)., with our observation showing that 1.78% of people identify as gay, lesbian or bisexual.

In my study, I intend to utilize regression analysis to examine the potential relationship between sexual orientation and personal income, incorporating ‘hours worked per week’ as an additional independent variable. This multifaceted approach will enable me to quantify the impact of both sexual orientation and hours worked on personal income, offering a more detailed understanding of the economic landscape.Specifically, a regression model will allow us to assess the magnitude of influence that various predictors exert on an outcome variable (Fogarty 2019).

The regression coefficients for sexual orientation will reveal how personal income varies across different sexual orientations in comparison to a reference group, while the coefficient for hours worked will indicate the average income change associated with additional hours worked per week. These coefficients are crucial for understanding the extent to which sexual orientation and hours worked independently correlate with income levels.

Statistical significance will be determined through p-values, with lower values indicating a stronger likelihood that the observed relationships are not due to chance. A positive coefficient for hours worked would suggest that longer working hours are associated with higher personal income, while the coefficients for sexual orientation will clarify whether income disparities exist between different sexual orientation groups.

The analysis will be supported by the adjusted R-squared value, which will reflect the percentage of the variability in personal income explained by the model after adjusting for the number of predictors. A high adjusted R-squared value would affirm the model’s explanatory power.

Furthermore, the residual standard error will offer insight into the accuracy of the income predictions made by the model, based on the average deviation of the observed incomes from the model’s predictions. The F statistic will be utilized to test the overall fit of the model, with a significant F statistic reinforcing the model’s validity.

By integrating hours worked per week into the regression model alongside sexual orientation, the analysis will provide a more nuanced view of the factors that contribute to personal income variations, ensuring a fair assessment of the influences on economic status within the sample size under study.

Results and discussion

library(tidyverse)
library(vtable) 
library(haven)
library(moderndive)
nilt <- read_sav("data/nilt2012.sav") 
nilt <- nilt %>% mutate(orient = as_factor(orient)) 
nilt_subset <- select(nilt, orient, persinc2) 
ggplot(nilt_subset, aes(y = orient, x= persinc2)) +
  geom_boxplot() +
  ggtitle('Annual Income vs Sexual Orientation')

The box plot presented above illustrates the distribution of annual personal income across different sexual orientation categories. Each box represents the interquartile range of income for each orientation, showing the middle 50% of the data. The median income indicated by the line within each box provides a central value around which the other incomes are spread. In the context of this box plot, the median is particularly relevant, as it is not affected by outliers that can skew averages. By comparing the medians across different categories, we can get a sense of which groups tend to earn more or less, relative to each other. A higher median line in one box compared to another suggests that the former group tends to have higher income levels.

Beginning with the Hetrosexual category, we notice a compact Inter quartile range, suggesting less variability in income among individuals within this group. Their median income, appears to be lower than that of the Other and bi-sexual categories. This group’s distribution has a few outliers, which are depicted as individual points beyond the ‘whiskers’, these lines extending from the top and bottom of the box. These outliers indicate that there are individuals with incomes significantly higher or lower than the typical range of the group, showing a lar. The homosexual category shows a wider Inter quartile range than the heterosexual category, hinting at greater diversity in the earnings of this group. The median of this category is higher than that of the heterosexual group, suggesting that the central tendency of income is greater for individuals identifying as gay or lesbian. For the bi-sexual grouping, the IQR is the widest, which points to a substantial variation in income within this group. The median income is higher compared to the heterosexual and homosexual categories, which may indicate that individuals in this group have a central tendency to earn more, essentially meaning that the median value within this group exceeds that of those within the hetrosexual and homosexual categories. The Other category has a relatively high median income, comparable to or slightly higher than the bisexual category, with a narrower IQR, indicating less variability in income. This group also displays outliers, which are only present on the higher end, suggesting that there are individuals with exceptionally high incomes in this category. However this may be attributed to the much lower sample size of respondents, much like our bi-sexual grouping, which will be commented on in my analysis. Overall, the box plot suggests that there are observable differences in the distribution of personal income across sexual orientation groups, with variation not only between the groups but also within them. The presence of outliers in several categories indicates potential income disparities among individuals within the same sexual orientation group. To creat more concrete conclusions about the differences in income across these groups, the introduction of hours worked per week may help to illustrate the pay disparities among the groups despite the small sample sizes.

library(tidyverse)
library(vtable)
library(haven)
nilt <- read_sav("data/nilt2012.sav")
nilt <- nilt %>% mutate(orient = as_factor(orient))
nilt_subset <- select(nilt, orient, persinc2)
ggplot(nilt,  aes(x = rhourswk, y = persinc2, color = orient)) +
  geom_point(position = "jitter") +
  labs(title = "Annual Income vs Hours worked vs Orientation",
       x = "Hours Worked per week", y = "Annual Income" )

With the addition of the amount of hours worked per week as an independent variable in a scatter plot form, we can now analyse the relationship between hours worked per week and annual income across different sexual orientations. The color-coded points represent various sexual orientation categories, showing the spread and concentration of annual incomes for the number of hours worked.

Firstly, it is important to note that a significant portion of the data points exists below the 50-hour workweek mark, aligning with typical full-time employment schedules. Across this range, there is a visible spread in income levels, with a notable cluster of points below the national average wage. This clustering below the average wage indicates that a considerable portion of the sample size, irrespective of sexual orientation, may earn less than the average national wage of £24011 (Northern Ireland Annual Survey of Hours and Earnings 2012).

However, When examining the plot through the sole lens of sexual orientation, it’s apparent that income distribution is varied within each category, with the main finding being that from our sample, only hetrosexual individuals appear to be the outliers making over the £40000 mark, significantly exceeding the national average wage of £24011. This finding that heterosexual individuals are the only group with a significant presence above the national average wage could imply several underlying dynamics. It may reflect the broader socio-economic advantages they possess in society, such as greater access to high-paying jobs and more opportunities for advancement based on their sexual orientation. It can then further be noted that the absence of a similar pattern of high-income earners among homosexual, bisexual, and individuals of other sexual orientations might indicate systemic barriers, such as discrimination or unequal opportunities in the workplace that can affect earning potential. However our data does not account for these potential externalities. Therefore, we can conlclude that there is a relationship between sexual orientation and hours worked relating to income, as those that identify differently to hetrosexual fail to break the potential glass cieling of earning over £40000, even when working similar amounts of hours to hetrosexual individuals.

Regression results

	Annual Personal Income (GBP)
Sexual Orientation: Homosexual (ref.: Heterosexual)	-2,199.928
	(4,687.426)
Sexual Orientation: bi-sexual	627.237
	(9,848.683)
Sexual Orientation: Other	465.931***
	(52.692)
Hours Worked	5,132.971**
	(1,973.586)
Observations	456
R2	.148
Adjusted R2	.142
Residual Std. Error	13,894.770 (df = 452)
F Statistic	26.069*** (df = 3; 452)
Notes:	*P < .05
	**P < .01
	***P < .001

The regression model further highlights the relationship between sexual orientation, hours worked and personal income. Firstly, individuals identifying as homosexual have an average annual income that is £2,199.928 less than those identifying as heterosexual, however this coefficient is not marked down as statistically signifcant, meaning there isn’t enough evidence to confidently assert that being homosexual, in itself, is associated with lower income in the population from which the sample is drawn. The model also highlights that the coefficient for bisexual individuals indicates a slightly higher income compared to heterosexuals. However, the absence of statistical significance here again similarly suggests that the data does not provide strong evidence of a real income difference due to sexual orientation. However, the Other category varies, showing a significantly higher average income with a strong level of statistical significance (p < 0.001). This significant result is crucial as it implies that the observed difference in income is highly unlikely to be due to random chance and may reflect a genuine disparity in earnings unlike our previous groups. As expected, the positive and statistically significant coefficient for hours worked indicates a clear relationship which shows as the number of hours worked increases, so does annual income. This significance therefore confirms that hours worked is a reliable predictor of income in this model. The slight decrease from R² to adjusted R² signifies that when the number of variables is accounted for, the model’s ability to explain income variance among different sexual orientations is moderately reduced from 14.8% to 14.2%. Furthermore the F statistic is highlighted as highly signifcant for my model, which shows infers that the overall regression model fits the data well compared to a model without independent variables.

Thereore, in relation to my original hyoptheses we reject the Null Hypothesis (H₀) in favor of the Alternative Hypothesis (Hₐ) for the ‘Other’ sexual orientation category, as the data indicates a significant effect of sexual orientation on income for this group. However, for the Homosexual’ and ‘Bi-sexual categories, we fail to reject the Null Hypothesis (H₀) due to the lack of statistically significant findings from the model. However these results do answer my original question of “Does sexual orientation affect annual personal income in Northern Ireland?”, as while no significant effect was found for homosexual and bisexual individuals, a significant effect was observed for the ’Other’ category. This mixed outcome implies that sexual orientation can affect annual personal income in Northern Ireland, but this effect is not similar across all sexual orientation groups. Brysons original study showing there continues to be disparity in pay between different sexual orientations can therefore be proved for certain groups, however there is not enough statistical signifcance in others to conclude that it is routed among all sexual orientations.

In terms of limitations for this model, it must be noted that there was an abnormally low sample size for sexual orientations outside of hetrosexuality, which could be attributed to factors the model cannot account for like societal pressures or perceived norms, potentially leading to underreporting. This aspect of data collection may have evolved considerably since 2012, as evidenced by the increase in self-identification among gay or bisexual individuals from 1.5% in 2012 to 3.3% in the 2023 UK census. This shift suggests a societal change towards greater openness and acceptance, which could impact the representativeness of data in studies like this one.

Conclusion

My analysis of the data suggests that sexual orientation may have an impact on personal income, but this effect is not uniformly observed across all sexual orientation groups. Key findings indicate significant income disparities for individuals categorized under ‘Other’, while no significant differences were found for those identifying as homosexual or bisexual when compared to heterosexuals. This study has revealed the importance of considering sexual orientation in economic research and the potential influence of societal changes on self-reporting in such studies.

What i have learned is that statistical significance plays a crucial role in understanding the relationship between variables like sexual orientation and income. The varying levels of significance across different groups underscore the complex nature of economic disparities and highlight the need for large, representative sample sizes. This becomes particularly important in light of societal shifts towards greater acceptance, as indicated by the change in self-reporting rates of sexual orientation from 2012 to 2023.

In the future, subsequent studies would benefit from incorporating a broader and more varied participant base to enhance the signifcance of the findings. Recognizing the profound impact societal dynamics can have on such research is crucial. A future investigation should delve into temporal patterns to grasp the evolving effects of societal attitudes on the income variation across sexual orientations. Incorporating a more expansive sample size is vital to lend greater validity to the results. Additionally, examining other influential elements like educational attainment, types of employment, and regional economic differences would offer a more complete picture of how sexual orientation intersects with financial outcomes.

References

ARK. Northern Ireland Life and Times Survey, 2012 [computer file]. ARK www.ark.ac.uk/nilt [distributor], June 2013.

Bryson, Alex. “Pay Equity after the Equality Act 2010: Does Sexual Orientation Still Matter?” Work, Employment & Society, vol. 31, no. 3, 2017, pp. 483–500. JSTOR, https://www.jstor.org/stable/26941754. Accessed 11 Jan. 2024.

Duggan, M. (2017). Lost in Transition? Sexuality and Justice in Post-conflict Northern Ireland.

Fogarty, B.J. (2019). Quantitative social science data with R an introduction. Los Angeles, Calif. Sage.

https://www.psni.police.uk/about-us/our-publications-and-reports/official-statistics/hate-motivation-statistics https://www.ons.gov.uk/peoplepopulationandcommunity/culturalidentity/sexuality/bulletins/sexualidentityuk/2019 https://www.ons.gov.uk/peoplepopulationandcommunity/culturalidentity/sexuality McCormick, C. (2020). The Legalisation of Same-sex Marriage in Northern Ireland.