Methodology

Authors

Jolyon Miles-Wilson

Celestin Okoroji

Published

May 29, 2025

1 Participants and Design

[Insert Name of Report] contains data from two studies. The first of these studies was a nationally representative survey of 10155 [Workers?] conducted by Opinium Research between 25th November 2023 and the 21st December 2023.

To achieve a robust estimate of outsourced workers, the sample was weighted by age, gender, and education, region, and ethnicity. The ethnic minority sub-sample (1,435 respondents) was also weighted separately by age, gender, and region to ensure that findings related to ethnic minority adults were fully representative. Targets were estimated using data from the Labour Force Survey, the 2021 Census for England and Wales, and the Northern Ireland Census.

This sample had median age 42 (SD = 13.02). 51% of respondents identified as female, 48% as male, 0.14% as other, and 0.65% preferred not to identify a gender. 76% of respondents identified as ‘English / Welsh / Scottish / Northern Irish / British’ (see Section 7.3 for a detailed breakdown of ethnicity).

A follow-up survey of Outsourced workers (as defined in Section 2.1) was conducted by Opinium Research between 19th April to the 16th of May 2024 with a total sample of 1814. The purpose of this study was to further probe the experiences of outsourced workers and to understand the impact of outsourcing on their work and lives (see Section 2.2).

Soft quotas on age, gender, and region were implemented to ensure broad representativeness, and the final data was weighted to targets based on age, gender, education, region, and ethnicity. The targets were based on the weighted data from study 1. The survey population had a mean Age of 38.9 (SD = 13.0). 42.4% Female and 65.5% White British. A small proportion of respondents had previously participated in study 1 and met the outsourced criteria (5%).

Both surveys were administered online.

An initial pilot study aimed to refine the diagnostic questions used to identify outsourced workers, ensuring they aligned with JRF’s initial definition and could be accurately answered by survey respondents. The diagnostic questions and feedback follow-ups were run on Opinium’s political omnibus, a nationally and politically representative sample of 2,055 UK adults between 30 August and 1 September 2023. The questions were filtered to those in work, resulting in a total of 1,200 respondents. Data from this pilot study is not reported here.

[POTENTIALLY ADD A TABLE HERE WITH CROSSTABS FOR THE TWO SAMPLES OR TABBED VISUALISATIONS]

2 Measures

2.1 Study 1- Nationally representative survey

The survey covered personal demographics, employment demographics (e.g. occupation, hours worked, pay), and the outsourced diagnostic questions. The main objectives were to ensure an accurate estimate of the size and demographic makeup of the outsourced population, and to analyse the data alongside the Labour Force Survey (LFS). [MORGAN - WHAT EXACTLY WAS INTENDED TO BE COMPARABLE? SPECIFICALLY WHICH QUESTIONS HAVE BEEN REPLICATED FROM TLFS]

Comparability to the LFS posed challenges, primarily because the LFS is conducted face-to-face, with interviewers playing a significant role in ensuring the accuracy of data and respondents’ understanding of questions. However, as the Transformed Labour Force Survey (TLFS)– an online first version of the survey set to replace the LFS— was underway, where possible we used the TLFS versions. While question wording is still under review, this was deemed the best approach, as some TLFS waves had already taken place and findings on comparability to LFS [MORGAN - CITATION].

2.1.1 Income calculations ([perhaps more detailed than necessary in this section])

Respondents could choose how they provided information about their income. Firstly, they could choose the payment period for which to express their income from the following options:

  • Annually / per year
  • Monthly
  • Weekly
  • Hourly

Secondly, they could choose either an ‘open’ form of reporting or a ‘closed’ form. The open form required respondents to type in their pay for the payment period they chose. The closed form required respondents to select which income bracket their pay belonged to from a list of options.

The annual options were:

  • Less than £5,600 a year
  • £5,600 up to £11,200
  • £11,201 up to £16,800
  • £16,801 up to £22,400
  • £22,401 up to £28,000
  • £28,001 up to £33,600
  • £33,601 up to £39,200
  • £39,201 up to £44,800
  • £44,801 up to £50,400
  • £50,401 up to £56,000
  • Over £56,000 a year
  • Prefer not to say

The monthly options were:

  • Less than £470 a month
  • £470 up to £940
  • £941 up to £1,410
  • £1,411 up to £1,880
  • £1,881 up to £2,350
  • £2,351 up to £2,820
  • £2,821 up to £3,290
  • £3,291 up to £3,760
  • £3,761 up to £4,230
  • £4,231 up to £4,700
  • Over £4,700 a month
  • Prefer not to say

The weekly options were:

  • Less than £110 a week
  • £110 up to £220
  • £221 up to £330
  • £331 up to £440
  • £441 up to £550
  • £551 up to £660
  • £661 up to £770
  • £771 up to £880
  • £881 up to £990
  • £991 up to £1,100
  • Over £1,100 a week
  • Prefer not to say

The hourly options were:

  • Less than £8.91 an hour
  • £8.91 up to £10.00
  • £10.01 up to £12.00
  • £12.01 up to £14.00
  • £14.01 up to £16.00
  • £16.01 up to £18.00
  • £18.01 up to £20.00
  • £20.01 up to £22.00
  • £22.01 up to £24.00
  • £24.01 up to £27.00
  • £27.01 up to £30.00
  • Over £30.00 an hour
  • Prefer not to say

7499 respondents answered using the open method. 1445 respondents answered using the closed method. 1211 did not answer either.

We equivalised respondents’ income across the reporting options in two steps. Firstly, we converted closed income responses to continuous numeric values by taking the midpoint of the income brackets, or the value of the “less than” and “over” values. For example, a closed response of “£5,600 up to £11,200” would be converted to £8400; and a closed response of “Less than £5,600 a year” would be converted to £5600. These converted closed responses were combined with the open responses to produce a single continous income variable across payment periods.

Next, we expressed all respondents’ income in annual, weekly, and hourly periods. To do this we made an assumption about the number of working weeks in a year based on the minimum holiday entitlement of 28 days. We calculated the total number of weeks in a year as 365 / 7 = 52.14, the total number of non-working weeks as 28 / 5 = 5.6, and thus the total number of working weeks as 52.14 - 5.6 = 46.54.

With this figure and the number of hours worked per week, we could convert incomes provided in one payment period to another. The table below shows how this was achieved.

[ADD OUTLIER EXCLUSION CRITERIA]

Income provided... Formula to convert to annual Formula to convert to weekly Formula to convert to hourly
... annually = income = income / working weeks = weekly income / hours worked per week
... monthly = income x 12 = (income x 12) / working weeks = weekly income / hours worked per week
... weekly = income x working weeks = income = weekly income / hours worked per week
... hourly = income x hours per week x working weeks = income x hours per week = weekly income / hours worked per week

2.2 Study 2 - Outsourced workers survey

In the follow up survey of outsourced workers the data focuses on workers experiences and perceptions of outsourced work. The dataset is large containing 214 variables. Analysis of all the variables was beyond the scope of the report thus we focus on a subset of the data pertaining to outsourced workers experiences of rights violations, discrimination, job clarity, benefits and drawbacks of outsourced work and potential improvements to their work arrangements.

In the process of data cleaning we set hours per week to NA for participants who gave an impossible number of work hours per week (e.g. \(\ge\) 168, N=11). Relatedly we construct variables to determine hourly, weekly, monthly and annual pay as in study 1 and flag outlier responses. Through this method 11.22% (183) participants were dropped from all subsequent analysis leaving a final sample of 1631 participants. We also determine whether the participant is low paid using the method from study 1.

A data dictionary is available from the Github Repository associated with this project along with all code used to produce the analyses.

3 Analysis - Study 1

3.1 Defining outsourcing

Workers were defined as outsourced based on responses to a set of diagnostic questions. Three questions asked respondents directly about whether they considered themselves outsourced and/or agency workers.

The first of these questions asked respondents to indicate directly whether they considered themselves outsourced by selecting one of the following options:

  1. I am sure I’m an outsourced worker
  2. I think I might be an outsourced worker
  3. I am not an outsourced worker

The second question asked respondents to indicate whether they considered themselves an agency worker by selecting from three options. For respondents who responded 1 or 2 to question 1, the options were:

  1. I am sure that I’m also an agency worker
  2. I think I might also be an agency worker
  3. I am not an agency worker

For respondents who responded 3 to question 1, the options were:

  1. I am sure that I’m an agency worker
  2. I think I might be an agency worker
  3. I am not an agency worker

Respondents were also asked whether the work they do was long- or short-term by selecting one of:

  1. I’m hired to do work which an organisation needs doing on a long-term or ongoing basis.
  2. I’m hired to do work which an organisation needs doing on a short-term or temporary basis.
  3. Other (please specify)

Finally, respondents were asked about aspects of their work that might indicate that the work they do is outsourced work. Respondents were asked: “Please read each of the following statements and tell us whether or not they are true for you and your work.” The statements were:

  1. I am paid by one organisation but I do work for a different organisation.
  2. The organisation I’m paid by is a ‘third party’ organisation which other organisations hire to do work for them, rather than doing that w [FIND QUESTION IN DATA DICT]
  3. My employer / agency provides people to do work for other organisations (i.e. they might provide people to do cleaning, security, administratio [FIND QUESTION IN DATA DICT]
  4. On a day-to-day basis, I’m paid by one organisation but I get given tasks or instructions by people who are paid by a different organisation.
  5. I am paid by one organisation, but I work in a space which has the logo or branding of a different organisation.
  6. I wear a uniform which has the logo or branding of my employer / agency, and which marks me out as being paid by a different organisation to so [FIND QUESTION IN DATA DICT]

Workers were categorised into three mutually exclusive sub groups based on their responses to the above questions.

  1. A respondent was categorised as ‘clearly outsourced’ if they responded ‘I am sure I’m an outsourced worker’ or ‘I think I might be an outsourced worker’ and ‘I’m hired to do work which an organisation needs doing on a long-term or ongoing basis.’.

  2. A respondent was categorised as ‘likely agency’ if they responded ‘I am sure that I’m an agency worker’ and ‘I’m hired to do work which an organisation needs doing on a long-term or ongoing basis’, excluding those people who are already defined as being ‘clearly outsourced’.

  3. A respondent was categorised as belonging to the ‘high indicators’ group if they responded TRUE to five or six [CAN THIS BE EXPRESSED AS \(\ge\) 5?] of the outsourcing indicators, as well as responding ‘I’m hired to do work which an organisation needs doing on a long-term or ongoing basis’, excluding those people who were already defined as ‘clearly outsourced’ or ‘likely agency’.

Together, these three sub groups form the classification of ‘outsourced workers’ considered in this report. Throughout the report, the term ‘outsourced’ refers to workers across the three sub groups. In places, analysis considers the three sub groups separately, in which case the groups will be referred to by name as ‘clearly outsourced’, ‘likely agency’, or ‘high indicators’.

3.2 Defining low pay

A ‘low pay’ binary variable was created by implementing an income threshold below which respondents were considered to be on a relatively low income. In line with with the Organisation for Economic Co-operation and Development, we set the threshold at two-thirds median weekly income. The two-thirds threshold was based on the weekly median income for respondents’ region to account for regional variations in earnings.

Regional weekly median income values were drawn from the Annual Survey of Hours and Earnings (2023 provisional edition). Respondents whose reported weekly income was less than or equal to two-thirds of the median weekly income in their region were assigned to the ‘low pay’ group, while those whose reported weekly income was greater than two-thirds of the median weekly income in their region were assigned to the ‘not low pay’ group.

3.3 Aggregating ethnicity

For reference, the table below provides a disambiguation of how ethnicities have been grouped in this analysis.

For analyses using the disaggregated (survey) categories with 21 levels, the reference category is “English / Welsh / Scottish / Northern Irish / British”.

For analyses using the aggregated categories with 9 levels, the reference category is “White British”

For analyses using teh aggregated categories with 4 levels, the reference category is “White”.

Ethnicity: Survey (21 levels) Ethnicity: Aggregated (9 levels) Ethnicity: Binary (4 levels)
English / Welsh / Scottish / Northern Irish / British White British White
Irish White other White
Gypsy or Irish Traveller White other White
Roma White other White
Any other White background White other White
White and Black Caribbean Mixed/Multiple ethnic group Non-White
White and Black African Mixed/Multiple ethnic group Non-White
White and Asian Mixed/Multiple ethnic group Non-White
Any other Mixed / Multiple ethnic background Mixed/Multiple ethnic group Non-White
Indian Asian/Asian British Non-White
Pakistani Asian/Asian British Non-White
Bangladeshi Asian/Asian British Non-White
Chinese Asian/Asian British Non-White
Any other Asian background Asian/Asian British Non-White
African Black/African/Caribbean/Black British Non-White
Caribbean Black/African/Caribbean/Black British Non-White
Any other Black, Black British, or Caribbean background Black/African/Caribbean/Black British Non-White
Arab Arab/British Arab Non-White
Any other ethnic group Other ethnic group Non-White
Don’t think of myself as any of these Don't think of myself as any of these Don't think of myself as any of these
Prefer not to say Prefer not to say Prefer not to say
NA NA NA

3.4 Models

In this section we describe the statistical models used in the report. In all models we applied survey weights so that the estimates can be considered representative of employees nationally.

3.4.1 Outsourced pay gap

To investigate the pay gap been outsourced and non-outsourced workers we constructed a linear regression model predicting annual and weekly income (in separate models) from outsourcing membership. We included other variables in the model to account for their potential influence on income. The full regression model can be expressed as:

\[ Income = Age + Gender + Education + Ethnicity + Migration + Region + Outsourcing \]

where

  • Income is a continuous numeric variable indicating a the respondent’s income (weekly or annual, in different models)
  • Age is a continuous numeric variable indicating the respondent’s age
  • Gender is a categorical variable with three levels:
    • Male (reference category)
    • Female
    • Other
  • Education is a categorical variable indicating whether the respondent has a degree, with three levels:
    • Yes (reference category)
    • No
    • Don’t know
  • Ethnicity is a categorical variable with eight levels:
    • White British (reference category)
    • Arab/British Arab
    • Asian/Asian British
    • Black/African/Caribbean/Black British
    • Mixed/Multiple ethnic group
    • Other ethnic group
    • Prefer not to say
    • White other
    • Don’t think of myself as any of these
  • Migration is a categorical variable indicating when the respondent arrived in the UK, with 10 levels:
    • I was born in the UK (reference category)
    • Within the last year
    • Within the last 3 years
    • Within the last 5 years
    • Within the last 10 years
    • Within the last 15 years
    • Within the last 20 years
    • Within the last 30 years
    • More than 30 years ago
    • Prefer not to say
  • Region is a categorical variable indicating the respodent’s region of residence, with 12 levels:
    • London (reference category)
    • East Midlands
    • East of England
    • North East
    • North West
    • Northern Ireland
    • Scotland
    • South East
    • South West
    • Wales
    • West Midlands
    • Yorkshire and the Humber
  • Outsourcing is a categorical variable indicating whether the respondent is outsourced, with two levels:
    • Not outsourced (reference category)
    • Outsourced

The annual income model was statistically significant (R2 = 0.18, F(35, 8071) = 51.29, p < .001). The table below shows the coefficients for the annual income model.

  Annual income
Predictors Estimates CI p
Intercept 39068.13 37794.38 – 40341.88 <0.001
Age 14.39 -6.41 – 35.19 0.175
Gender: Female -7002.82 -7535.16 – -6470.48 <0.001
Gender: Other -6032.87 -12748.83 – 683.09 0.078
Gender: Prefer not to say -2828.76 -9792.72 – 4135.20 0.426
Education: Don't have degree -8170.64 -8723.33 – -7617.95 <0.001
Education: Don't know -9849.13 -12104.71 – -7593.55 <0.001
Ethnicity: Arab/British Arab -177.61 -4873.46 – 4518.23 0.941
Ethnicity: Asian/Asian British -471.78 -1573.79 – 630.22 0.401
Ethnicity: Black/African/Caribbean/Black British -1203.90 -2816.77 – 408.97 0.143
Ethnicity: Don't think of myself as any of these -3198.11 -12756.84 – 6360.62 0.512
Ethnicity: Mixed/Multiple ethnic group -1507.68 -3488.56 – 473.20 0.136
Ethnicity: Other ethnic group 3596.90 -998.30 – 8192.10 0.125
Ethnicity: Prefer not to say -82.72 -5289.34 – 5123.89 0.975
Ethnicity: White other -637.07 -2018.88 – 744.74 0.366
Region: East Midlands -5854.69 -7085.16 – -4624.23 <0.001
Region: East of England -4103.34 -5262.01 – -2944.67 <0.001
Region: North East -4834.89 -6372.61 – -3297.16 <0.001
Region: North West -4472.28 -5597.32 – -3347.24 <0.001
Region: Northern Ireland -6336.40 -8132.24 – -4540.55 <0.001
Region: Scotland -5448.95 -6649.58 – -4248.32 <0.001
Region: South East -3460.88 -4512.27 – -2409.49 <0.001
Region: South West -5748.69 -6947.04 – -4550.34 <0.001
Region: Wales -5215.03 -6681.40 – -3748.66 <0.001
Region: West Midlands -4759.33 -5932.19 – -3586.48 <0.001
Region: Yorkshire and the Humber -5451.06 -6649.03 – -4253.09 <0.001
Outsourcing: Outsourced -2995.19 -3715.49 – -2274.89 <0.001
Migration: Arrived within the last year -6032.95 -8309.80 – -3756.10 <0.001
Migration: Arrived within the last 3 years -2375.85 -4406.64 – -345.06 0.022
Migration: Arrived within the last 5 years -1830.71 -4132.93 – 471.51 0.119
Migration: Arrived within the last 10 years -691.71 -2485.65 – 1102.24 0.450
Migration: Arrived within the last 15 years 747.68 -1267.29 – 2762.65 0.467
Migration: Arrived within the last 20 years 1625.36 -508.65 – 3759.37 0.135
Migration: Arrived within the last 30 years 2911.95 401.12 – 5422.79 0.023
Migration: Arrived more than 30 years ago -46.10 -2002.94 – 1910.74 0.963
Migration: Prefer not to say -1667.72 -5284.42 – 1948.98 0.366
Observations 8107
R2 / R2 adjusted 0.182 / 0.178

As expected, the model statistics for weekly income model were identical to the those of the annual income model. The model was statistically significant (R2 = 0.18, F(35, 8071) = 51.29, p < .001). The table below shows the coefficients for the weekly income model.

  Weekly income
Predictors Estimates CI p
Intercept 839.40 812.03 – 866.77 <0.001
Age 0.31 -0.14 – 0.76 0.175
Gender: Female -150.46 -161.90 – -139.02 <0.001
Gender: Other -129.62 -273.92 – 14.68 0.078
Gender: Prefer not to say -60.78 -210.40 – 88.85 0.426
Education: Don't have degree -175.55 -187.43 – -163.68 <0.001
Education: Don't know -211.61 -260.08 – -163.15 <0.001
Ethnicity: Arab/British Arab -3.82 -104.71 – 97.08 0.941
Ethnicity: Asian/Asian British -10.14 -33.81 – 13.54 0.401
Ethnicity: Black/African/Caribbean/Black British -25.87 -60.52 – 8.79 0.143
Ethnicity: Don't think of myself as any of these -68.71 -274.09 – 136.66 0.512
Ethnicity: Mixed/Multiple ethnic group -32.39 -74.95 – 10.17 0.136
Ethnicity: Other ethnic group 77.28 -21.45 – 176.01 0.125
Ethnicity: Prefer not to say -1.78 -113.64 – 110.09 0.975
Ethnicity: White other -13.69 -43.38 – 16.00 0.366
Region: East Midlands -125.79 -152.23 – -99.35 <0.001
Region: East of England -88.16 -113.06 – -63.27 <0.001
Region: North East -103.88 -136.92 – -70.84 <0.001
Region: North West -96.09 -120.26 – -71.92 <0.001
Region: Northern Ireland -136.14 -174.73 – -97.56 <0.001
Region: Scotland -117.07 -142.87 – -91.28 <0.001
Region: South East -74.36 -96.95 – -51.77 <0.001
Region: South West -123.51 -149.26 – -97.77 <0.001
Region: Wales -112.05 -143.55 – -80.54 <0.001
Region: West Midlands -102.26 -127.46 – -77.06 <0.001
Region: Yorkshire and the Humber -117.12 -142.86 – -91.38 <0.001
Outsourcing: Outsourced -64.35 -79.83 – -48.88 <0.001
Migration: Arrived within the last year -129.62 -178.54 – -80.70 <0.001
Migration: Arrived within the last 3 years -51.05 -94.68 – -7.41 0.022
Migration: Arrived within the last 5 years -39.33 -88.80 – 10.13 0.119
Migration: Arrived within the last 10 years -14.86 -53.41 – 23.68 0.450
Migration: Arrived within the last 15 years 16.06 -27.23 – 59.36 0.467
Migration: Arrived within the last 20 years 34.92 -10.93 – 80.77 0.135
Migration: Arrived within the last 30 years 62.56 8.62 – 116.51 0.023
Migration: Arrived more than 30 years ago -0.99 -43.03 – 41.05 0.963
Migration: Prefer not to say -35.83 -113.54 – 41.87 0.366
Observations 8107
R2 / R2 adjusted 0.182 / 0.178

3.4.2 Gender pay gap

The above model was also used to assess a possible gender pay gap. As shown in the preceding two tables, there is a significant difference in pay between men and women. Annually, women earn £7002.82 less than men. Per week, women earn £150.46 less than men.

We next explored whether outsourcing compounds this gender pay gap by adding an interaction term into the previous models so that

\[ Income = Age + Gender + Education + Ethnicity + Migration + Region + Outsourcing + Gender:Outsourcing \]

For both models, adding the interaction effect did not improve model fit (R2 = 0.18, F(3, 8068) = 0.74, p = 0.531). The tables below show the coefficients for each model.

  Annual income
Predictors Estimates CI p
Intercept 39092.47 37809.50 – 40375.44 <0.001
Age 14.30 -6.51 – 35.10 0.178
Gender: Female -7004.18 -7586.05 – -6422.30 <0.001
Gender: Other -3445.67 -10995.26 – 4103.92 0.371
Gender: Prefer not to say -2634.20 -9886.89 – 4618.49 0.477
Education: Has degree -8169.16 -8722.09 – -7616.22 <0.001
Education: Don't know -9849.18 -12104.91 – -7593.45 <0.001
Ethnicity: Arab/British Arab -170.96 -4867.74 – 4525.81 0.943
Ethnicity: Asian/Asian British -472.69 -1574.78 – 629.39 0.401
Ethnicity: Black/African/Caribbean/Black British -1203.91 -2816.96 – 409.13 0.143
Ethnicity: Don't think of myself as any of these -3193.67 -12753.23 – 6365.88 0.513
Ethnicity: Mixed/Multiple ethnic group -1511.45 -3492.47 – 469.58 0.135
Ethnicity: Other ethnic group 3593.79 -1002.26 – 8189.84 0.125
Ethnicity: Prefer not to say -81.55 -5288.77 – 5125.67 0.976
Ethnicity: White other -601.41 -1984.70 – 781.89 0.394
Region: East Midlands -5879.63 -7110.82 – -4648.44 <0.001
Region: East of England -4135.62 -5295.46 – -2975.79 <0.001
Region: North East -4865.76 -6404.23 – -3327.29 <0.001
Region: North West -4502.26 -5628.38 – -3376.14 <0.001
Region: Northern Ireland -6358.52 -8158.35 – -4558.69 <0.001
Region: Scotland -5476.68 -6678.10 – -4275.26 <0.001
Region: South East -3488.28 -4540.37 – -2436.18 <0.001
Region: South West -5772.53 -6971.45 – -4573.60 <0.001
Region: Wales -5238.94 -6705.78 – -3772.10 <0.001
Region: West Midlands -4783.07 -5956.82 – -3609.31 <0.001
Region: Yorkshire and the Humber -5477.80 -6676.53 – -4279.07 <0.001
Outsourcing: Outsourced -2979.46 -3945.17 – -2013.76 <0.001
Migration: Arrived within the last year -6043.23 -8320.40 – -3766.06 <0.001
Migration: Arrived within the last 3 years -2386.09 -4418.13 – -354.05 0.021
Migration: Arrived within the last 5 years -1849.05 -4152.16 – 454.06 0.116
Migration: Arrived within the last 10 years -719.66 -2514.13 – 1074.81 0.432
Migration: Arrived within the last 15 years 718.31 -1297.19 – 2733.81 0.485
Migration: Arrived within the last 20 years 1602.21 -532.38 – 3736.80 0.141
Migration: Arrived within the last 30 years 2893.95 382.76 – 5405.14 0.024
Migration: Arrived more than 30 years ago -58.23 -2016.55 – 1900.09 0.954
Migration: Prefer not to say -1683.14 -5300.38 – 1934.09 0.362
Interaction: Outsourcing x Gender Female 18.18 -1412.01 – 1448.37 0.980
Interaction: Outsourcing x Gender Other -12395.16 -28915.23 – 4124.91 0.141
Interaction: Outsourcing x Gender Prefer not to say -2506.28 -28505.03 – 23492.47 0.850
Observations 8107
R2 / R2 adjusted 0.182 / 0.178
  Weekly income
Predictors Estimates CI p
Intercept 839.92 812.36 – 867.49 <0.001
Age 0.31 -0.14 – 0.75 0.178
Gender: Female -150.49 -162.99 – -137.99 <0.001
Gender: Other -74.03 -236.24 – 88.18 0.371
Gender: Prefer not to say -56.60 -212.43 – 99.23 0.477
Education: Has degree -175.52 -187.40 – -163.64 <0.001
Education: Don't know -211.62 -260.08 – -163.15 <0.001
Ethnicity: Arab/British Arab -3.67 -104.59 – 97.24 0.943
Ethnicity: Asian/Asian British -10.16 -33.84 – 13.52 0.401
Ethnicity: Black/African/Caribbean/Black British -25.87 -60.52 – 8.79 0.143
Ethnicity: Don't think of myself as any of these -68.62 -274.01 – 136.77 0.513
Ethnicity: Mixed/Multiple ethnic group -32.47 -75.04 – 10.09 0.135
Ethnicity: Other ethnic group 77.21 -21.53 – 175.96 0.125
Ethnicity: Prefer not to say -1.75 -113.63 – 110.13 0.976
Ethnicity: White other -12.92 -42.64 – 16.80 0.394
Region: East Midlands -126.33 -152.78 – -99.87 <0.001
Region: East of England -88.86 -113.78 – -63.94 <0.001
Region: North East -104.54 -137.60 – -71.49 <0.001
Region: North West -96.73 -120.93 – -72.54 <0.001
Region: Northern Ireland -136.62 -175.29 – -97.95 <0.001
Region: Scotland -117.67 -143.48 – -91.86 <0.001
Region: South East -74.95 -97.55 – -52.34 <0.001
Region: South West -124.03 -149.79 – -98.27 <0.001
Region: Wales -112.56 -144.08 – -81.05 <0.001
Region: West Midlands -102.77 -127.99 – -77.55 <0.001
Region: Yorkshire and the Humber -117.69 -143.45 – -91.94 <0.001
Outsourcing: Outsourced -64.02 -84.76 – -43.27 <0.001
Migration: Arrived within the last year -129.84 -178.77 – -80.92 <0.001
Migration: Arrived within the last 3 years -51.27 -94.93 – -7.61 0.021
Migration: Arrived within the last 5 years -39.73 -89.21 – 9.76 0.116
Migration: Arrived within the last 10 years -15.46 -54.02 – 23.09 0.432
Migration: Arrived within the last 15 years 15.43 -27.87 – 58.74 0.485
Migration: Arrived within the last 20 years 34.42 -11.44 – 80.29 0.141
Migration: Arrived within the last 30 years 62.18 8.22 – 116.13 0.024
Migration: Arrived more than 30 years ago -1.25 -43.33 – 40.82 0.954
Migration: Prefer not to say -36.16 -113.88 – 41.56 0.362
Interaction: Outsourcing x Gender Female 0.39 -30.34 – 31.12 0.980
Interaction: Outsourcing x Gender Other -266.32 -621.26 – 88.63 0.141
Interaction: Outsourcing x Gender Prefer not to say -53.85 -612.45 – 504.75 0.850
Observations 8107
R2 / R2 adjusted 0.182 / 0.178

The interaction term is non-significant. Estimated marginal means show that:

  • Among not outsourced workers, men are paid £7004.18 more than women
  • Among outsourced workers, men are paid £6986 more than women
  • Among men, not outsourced workers are paid £2979.46 more than outsourced workers.
  • Among women, not outsourced workers are paid £2961.28 more than outsourced workers.

The plot below illustrates the main effects that men are paid more than women and that outsourced men and women are paid less than non-outsourced men and women. The lack of interaction indicates that the difference in pay between men and women does not significantly differ between outsourced and non-outsourced people.

3.4.3 Demographic models

3.4.3.1 Ethnicity

Several regressions were run to assess the likelihood of being outsourced from demographics. These models underlie the claims in the report in relation to ethnicity, migration, and gender.

The overall model was defined as:

\[ Outsourcing = Ethnicity + Age + Gender + Education + Region + Migration \]

where

  • Outsourcing is a categorical variable indicating whether the respondent is outsourced, with two levels:
    • Not outsourced (reference category)
    • Outsourced
  • Age is a continuous numeric variable indicating the respondent’s age
  • Gender is a categorical variable with three levels:
    • Male (reference category)
    • Female
    • Other
  • Education is a categorical variable indicating whether the respondent has a degree, with three levels:
    • Yes (reference category)
    • No
    • Don’t know
  • Migration is a categorical variable indicating when the respondent arrived in the UK, with 10 levels:
    • I was born in the UK (reference category)
    • Within the last year
    • Within the last 3 years
    • Within the last 5 years
    • Within the last 10 years
    • Within the last 15 years
    • Within the last 20 years
    • Within the last 30 years
    • More than 30 years ago
    • Prefer not to say
  • Region is a categorical variable indicating the respodent’s region of residence, with 12 levels:
    • London (reference category)
    • East Midlands
    • East of England
    • North East
    • North West
    • Northern Ireland
    • Scotland
    • South East
    • South West
    • Wales
    • West Midlands
    • Yorkshire and the Humber

For this exploration we modelled ethnicity in three ways.

  1. As a categorical variable with four levels:
    • White (reference category)
    • Not White
    • Don’t think of myself as any of these
    • Prefer not say
  2. As a categorical variable with eight levels:
    • White British (reference category)
    • Arab/British Arab
    • Asian/Asian British
    • Black/African/Caribbean/Black British
    • Don’t think of myself as any of these
    • Mixed/Multiple ethnic group
    • Other ethnic group
    • Prefer not to say
    • White other
  3. As a categorical variable with 21 levels:
    • English/Welsh/Scottish/Northern Irish/British (reference category)
    • Irish
    • Gypsy or Irish Traveller
    • Roma
    • Any other White background
    • White and Black Caribbean
    • White and Black African
    • White and Asian
    • Any other Mixed/Multiple ethnic background
    • Indian
    • Pakistani
    • Bangladeshi
    • Chinese
    • Any other Asian background
    • African
    • Caribbean
    • Any other Black, Black British, or Caribbean background
    • Arab
    • Any other ethnic group
    • Don’t think of myself as any of these
    • Prefer not to say

We used svyglm() from the survey package to construct survey-weighted generalised linear models. This approach allows us to take into account survey weights to produce design-based standard errors by assuming a ‘quasibinomial’ distribution to the data. Specifically, the survey-weighted data contains overdispersion; the variance is greater than expected by a binomial distribution (which assumes variance = mean(1 - mean)). The quasibinomial distribution estimates a dispersion parameter that allows the variance to be greater than expected by the true binomial distribution. For more information see Lumley, Thomas, and Alastair Scott. ‘Fitting Regression Models to Survey Data’. Statistical Science 32, no. 2 (2017): 265–780.

We used Rao–Scott adjusted Wald tests to compare nested survey-weighted models fit using a quasibinomial family. This method accounts for the survey design and is appropriate given that quasi-likelihood models do not support likelihood-ratio testing. For more information see Rao, J. N. K., and A. J. Scott. ‘On Chi-Squared Tests for Multiway Contingency Tables with Cell Proportions Estimated from Survey Data’. The Annals of Statistics 12, no. 1 (March 1984): 46–60. https://doi.org/10.1214/aos/1176346391.

For model 1, a saturated model including all variables was a significantly better fit to the data than an intercept-only model, F(29, 9782) = 8.72, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.65 0.49 – 0.86 0.002
Ethnicity: Not White 1.38 1.15 – 1.66 0.001
Ethnicity: Don't think of myself as any of these 2.50 0.76 – 8.18 0.131
Ethnicity: Prefer not to say 1.41 0.39 – 5.14 0.600
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.69 0.61 – 0.79 <0.001
Gender: Other 0.87 0.22 – 3.50 0.842
Gender: Prefer not to say 0.82 0.26 – 2.53 0.724
Education: Don't have degree 1.06 0.93 – 1.21 0.372
Education: Don't know 1.14 0.64 – 2.04 0.650
Region: East Midlands 0.94 0.71 – 1.25 0.676
Region: East of England 0.59 0.44 – 0.79 0.001
Region: North East 0.63 0.44 – 0.91 0.012
Region: North West 0.85 0.66 – 1.08 0.185
Region: Northern Ireland 0.70 0.46 – 1.07 0.098
Region: Scotland 0.68 0.50 – 0.93 0.015
Region: South East 0.62 0.49 – 0.79 <0.001
Region: South West 0.68 0.52 – 0.90 0.006
Region: Wales 0.91 0.66 – 1.25 0.551
Region: West Midlands 0.83 0.65 – 1.08 0.166
Region: Yorkshire and the Humber 0.68 0.52 – 0.89 0.005
Migration: Arrived within the last year 1.69 1.13 – 2.52 0.010
Migration: Arrived within the last 3 years 1.03 0.68 – 1.56 0.885
Migration: Arrived within the last 5 years 1.22 0.78 – 1.89 0.386
Migration: Arrived within the last 10 years 1.61 1.14 – 2.26 0.006
Migration: Arrived within the last 15 years 1.58 1.05 – 2.36 0.026
Migration: Arrived within the last 20 years 1.54 1.00 – 2.35 0.048
Migration: Arrived within the last 30 years 0.45 0.22 – 0.92 0.029
Migration: Arrived more than 30 years ago 2.01 1.31 – 3.08 0.001
Migration: Prefer not to say 1.22 0.71 – 2.10 0.472
Observations 9812

For model 2, a saturated model including all variables was a significantly better fit to the data than an intercept-only model, F(34, 9777) = 8.07, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.67 0.50 – 0.88 0.004
Ethnicity: Arab/British Arab 1.98 0.77 – 5.11 0.159
Ethnicity: Asian/Asian British 1.24 0.96 – 1.59 0.094
Ethnicity: Black/African/Caribbean/Black British 1.47 1.10 – 1.96 0.010
Ethnicity: Don't think of myself as any of these 2.35 0.72 – 7.73 0.159
Ethnicity: Mixed/Multiple ethnic group 1.37 1.00 – 1.89 0.053
Ethnicity: Other ethnic group 1.03 0.27 – 4.01 0.963
Ethnicity: Prefer not to say 1.37 0.37 – 5.01 0.639
Ethnicity: White other 0.81 0.58 – 1.11 0.191
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.69 0.61 – 0.78 <0.001
Gender: Other 0.88 0.21 – 3.62 0.860
Gender: Prefer not to say 0.81 0.26 – 2.50 0.711
Education: Don't have degree 1.06 0.93 – 1.21 0.357
Education: Don't know 1.16 0.65 – 2.06 0.618
Region: East Midlands 0.93 0.71 – 1.23 0.628
Region: East of England 0.58 0.43 – 0.78 <0.001
Region: North East 0.62 0.43 – 0.89 0.009
Region: North West 0.83 0.65 – 1.07 0.143
Region: Northern Ireland 0.72 0.47 – 1.10 0.129
Region: Scotland 0.67 0.50 – 0.91 0.011
Region: South East 0.61 0.48 – 0.78 <0.001
Region: South West 0.67 0.50 – 0.88 0.004
Region: Wales 0.89 0.65 – 1.22 0.476
Region: West Midlands 0.83 0.64 – 1.07 0.147
Region: Yorkshire and the Humber 0.67 0.51 – 0.88 0.004
Migration: Arrived within the last year 1.71 1.13 – 2.60 0.011
Migration: Arrived within the last 3 years 1.08 0.71 – 1.65 0.713
Migration: Arrived within the last 5 years 1.31 0.83 – 2.07 0.249
Migration: Arrived within the last 10 years 1.80 1.23 – 2.63 0.002
Migration: Arrived within the last 15 years 1.74 1.13 – 2.68 0.013
Migration: Arrived within the last 20 years 1.68 1.07 – 2.63 0.024
Migration: Arrived within the last 30 years 0.49 0.23 – 1.02 0.056
Migration: Arrived more than 30 years ago 2.09 1.36 – 3.22 0.001
Migration: Prefer not to say 1.24 0.72 – 2.14 0.430
Observations 9812

For model 3, a saturated model including all variables was a significantly better fit to the data than an intercept-only model, F(46, 9765) = 6.88, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.67 0.50 – 0.88 0.004
Ethnicity: Irish 0.76 0.40 – 1.45 0.410
Ethnicity: Gypsy or Irish Traveller 1.00 0.18 – 5.50 0.999
Ethnicity: Roma 1.25 0.32 – 4.87 0.751
Ethnicity: Any other White background 0.80 0.55 – 1.15 0.225
Ethnicity: White and Black Caribbean 0.49 0.25 – 0.95 0.035
Ethnicity: White and Black African 2.66 1.51 – 4.66 0.001
Ethnicity: White and Asian 1.19 0.58 – 2.46 0.634
Ethnicity: Any other Mixed/Multiple ethnic background 2.00 1.08 – 3.67 0.026
Ethnicity: Indian 1.18 0.81 – 1.71 0.396
Ethnicity: Pakistani 2.17 1.45 – 3.25 <0.001
Ethnicity: Bangladeshi 1.32 0.71 – 2.45 0.375
Ethnicity: Chinese 0.62 0.33 – 1.19 0.150
Ethnicity: Any other Asian background 1.19 0.70 – 2.04 0.521
Ethnicity: African 1.44 1.04 – 2.00 0.027
Ethnicity: Caribbean 1.27 0.63 – 2.54 0.499
Ethnicity: Any other Black, Black British, or Caribbean background 1.84 0.88 – 3.86 0.106
Ethnicity: Arab 1.97 0.76 – 5.10 0.164
Ethnicity: Any other ethnic group 1.03 0.26 – 4.03 0.963
Ethnicity: Don't think of myself as any of these 2.31 0.71 – 7.57 0.166
Ethnicity: Prefer not to say 1.37 0.37 – 5.04 0.632
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.69 0.61 – 0.79 <0.001
Gender: Other 0.88 0.22 – 3.63 0.865
Gender: Prefer not to say 0.83 0.27 – 2.56 0.747
Education: Don't have degree 1.06 0.93 – 1.20 0.402
Education: Don't know 1.16 0.65 – 2.06 0.623
Region: East Midlands 0.92 0.70 – 1.22 0.583
Region: East of England 0.57 0.42 – 0.77 <0.001
Region: North East 0.61 0.43 – 0.88 0.008
Region: North West 0.81 0.63 – 1.05 0.111
Region: Northern Ireland 0.73 0.46 – 1.14 0.165
Region: Scotland 0.67 0.49 – 0.91 0.009
Region: South East 0.60 0.48 – 0.77 <0.001
Region: South West 0.66 0.50 – 0.87 0.003
Region: Wales 0.88 0.64 – 1.21 0.423
Region: West Midlands 0.81 0.62 – 1.05 0.104
Region: Yorkshire and the Humber 0.65 0.50 – 0.86 0.002
Migration: Arrived within the last year 1.76 1.14 – 2.72 0.011
Migration: Arrived within the last 3 years 1.10 0.72 – 1.69 0.652
Migration: Arrived within the last 5 years 1.23 0.77 – 1.97 0.389
Migration: Arrived within the last 10 years 1.76 1.20 – 2.57 0.004
Migration: Arrived within the last 15 years 1.84 1.18 – 2.87 0.007
Migration: Arrived within the last 20 years 1.69 1.07 – 2.66 0.024
Migration: Arrived within the last 30 years 0.49 0.24 – 1.01 0.054
Migration: Arrived more than 30 years ago 2.11 1.37 – 3.26 0.001
Migration: Prefer not to say 1.22 0.70 – 2.12 0.478
Observations 9812

3.4.3.2 Migration

We next focus on predicting whether a person was outsourced based on wehther the person was born in the UK. This binary variable was constructed by collapsing the 10-level migration variable down into two levels, so that “I was born in the UK” becomes “Born in UK”, and all levels apart from “I was born in the UK” and “Prefer not to say” become “Not born in UK”.

A saturated model including all variables was a significantly better fit to the data than an intercept-only model, F(39, 9772) = 7.52, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.66 0.50 – 0.88 0.004
Migration: Not born in the UK 1.52 1.22 – 1.89 <0.001
Migration: Prefer not to say 1.23 0.71 – 2.13 0.465
Ethnicity: Irish 0.76 0.40 – 1.44 0.396
Ethnicity: Gypsy or Irish Traveller 1.04 0.20 – 5.55 0.962
Ethnicity: Roma 1.11 0.26 – 4.73 0.883
Ethnicity: Any other White background 0.82 0.57 – 1.16 0.253
Ethnicity: White and Black Caribbean 0.48 0.25 – 0.94 0.033
Ethnicity: White and Black African 2.56 1.45 – 4.54 0.001
Ethnicity: White and Asian 1.22 0.58 – 2.57 0.595
Ethnicity: Any other Mixed/Multiple ethnic background 1.81 1.04 – 3.16 0.037
Ethnicity: Indian 1.13 0.77 – 1.65 0.530
Ethnicity: Pakistani 2.13 1.40 – 3.22 <0.001
Ethnicity: Bangladeshi 1.27 0.68 – 2.37 0.457
Ethnicity: Chinese 0.60 0.31 – 1.15 0.125
Ethnicity: Any other Asian background 1.21 0.72 – 2.04 0.480
Ethnicity: African 1.46 1.08 – 1.98 0.014
Ethnicity: Caribbean 1.24 0.61 – 2.51 0.545
Ethnicity: Any other Black, Black British, or Caribbean background 1.73 0.83 – 3.64 0.146
Ethnicity: Arab 2.04 0.80 – 5.22 0.136
Ethnicity: Any other ethnic group 1.04 0.27 – 4.01 0.951
Ethnicity: Don't think of myself as any of these 2.28 0.71 – 7.27 0.164
Ethnicity: Prefer not to say 1.29 0.35 – 4.77 0.704
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.70 0.61 – 0.79 <0.001
Gender: Other 0.88 0.22 – 3.62 0.865
Gender: Prefer not to say 0.82 0.27 – 2.54 0.737
Education: Don't have degree 1.05 0.93 – 1.20 0.420
Education: Don't know 1.19 0.67 – 2.11 0.561
Region: East Midlands 0.93 0.70 – 1.23 0.598
Region: East of England 0.56 0.42 – 0.76 <0.001
Region: North East 0.61 0.42 – 0.88 0.007
Region: North West 0.81 0.63 – 1.04 0.100
Region: Northern Ireland 0.72 0.46 – 1.13 0.157
Region: Scotland 0.66 0.49 – 0.90 0.008
Region: South East 0.60 0.47 – 0.76 <0.001
Region: South West 0.66 0.50 – 0.87 0.003
Region: Wales 0.87 0.64 – 1.20 0.403
Region: West Midlands 0.80 0.62 – 1.04 0.098
Region: Yorkshire and the Humber 0.65 0.49 – 0.85 0.002
Observations 9812

3.4.3.3 Gender

We used the same generalised linear model as in the previous section to estimate the effect of Gender on outsourcing, where Gender is a categorical variable with four levels:

  • Male (reference category)
  • Female
  • Prefer not to say
  • Other

The model indicates that women are 0.7 times as likely (i.e. 30% less likely) to be outsourced than men.

3.4.3.4 Age

Again using the same model, we found that age was a significant predictor of the likelihood of being outsourced. The model indicates that for each year older a worker is, they are 0.98 times as likely (i.e. 2% less likely) to be outsourced.

We also explored how age predicted whether a person was on low pay. The model formula is:

\[ Income Group = Age + Outsourcing + Ethnicity + Gender + Education + Region + Migration \]

A saturated model including all variables was a significantly better fit to the data than an intercept-only model, F(40, 8066) = 34.7, p < .001. The table below shows the model coefficients.

  Low income group membership
Predictors Odds Ratios CI p
Intercept 0.09 0.06 – 0.12 <0.001
Age 1.00 1.00 – 1.01 0.079
Outsourcing: Outsourced 1.48 1.25 – 1.75 <0.001
Ethnicity: Irish 1.52 0.87 – 2.68 0.144
Ethnicity: Gypsy or Irish Traveller 0.26 0.03 – 2.00 0.195
Ethnicity: Roma 0.00 0.00 – 0.00 <0.001
Ethnicity: Any other White background 1.00 0.69 – 1.45 0.985
Ethnicity: White and Black Caribbean 1.27 0.68 – 2.39 0.454
Ethnicity: White and Black African 1.03 0.43 – 2.49 0.949
Ethnicity: White and Asian 1.61 0.90 – 2.91 0.111
Ethnicity: Any other Mixed/Multiple ethnic background 1.93 1.04 – 3.61 0.038
Ethnicity: Indian 0.82 0.53 – 1.25 0.349
Ethnicity: Pakistani 1.50 0.92 – 2.45 0.102
Ethnicity: Bangladeshi 1.66 0.80 – 3.43 0.172
Ethnicity: Chinese 0.70 0.36 – 1.39 0.307
Ethnicity: Any other Asian background 1.39 0.69 – 2.80 0.358
Ethnicity: African 1.38 0.94 – 2.02 0.096
Ethnicity: Caribbean 0.88 0.42 – 1.85 0.736
Ethnicity: Any other Black, Black British, or Caribbean background 1.24 0.49 – 3.10 0.651
Ethnicity: Arab 1.36 0.43 – 4.34 0.603
Ethnicity: Any other ethnic group 0.53 0.06 – 4.62 0.566
Ethnicity: Don't think of myself as any of these 0.44 0.07 – 2.90 0.395
Ethnicity: Prefer not to say 1.66 0.50 – 5.49 0.406
Gender: Female 3.26 2.85 – 3.72 <0.001
Gender: Other 3.68 1.14 – 11.86 0.029
Gender: Prefer not to say 1.99 0.42 – 9.33 0.385
Education: Don't have degree 2.59 2.28 – 2.93 <0.001
Education: Don't know 3.74 2.14 – 6.55 <0.001
Region: East Midlands 0.99 0.74 – 1.33 0.960
Region: East of England 0.97 0.72 – 1.30 0.831
Region: North East 0.76 0.53 – 1.10 0.143
Region: North West 0.69 0.52 – 0.92 0.010
Region: Northern Ireland 0.99 0.67 – 1.49 0.978
Region: Scotland 1.14 0.84 – 1.54 0.397
Region: South East 0.91 0.71 – 1.17 0.467
Region: South West 0.87 0.65 – 1.15 0.324
Region: Wales 0.63 0.44 – 0.92 0.015
Region: West Midlands 0.86 0.65 – 1.14 0.290
Region: Yorkshire and the Humber 0.87 0.66 – 1.16 0.351
Migration: Not born in the UK 0.96 0.74 – 1.24 0.749
Migration: Prefer not to say 1.73 0.76 – 3.93 0.190
Observations 8107

3.4.3.5 Ethnicity-migration interaction

We next explored whether there was an interaction between ethnicity and migration in predicting outsourcing using generalised linear models by adding an interaction effect to the model predicting outsourcing above so that the model formula is:

\[ Outsourcing = Ethnicity + Age + Gender + Educaton + Region + Migration + Ethnicity:Migration \]

where Ethnicity:Migration represents the interaction term.

We did this twice: first for the aggregated eight-level ethnicity variable, and then for the disaggregated 21-level variable.

3.4.3.5.1 Ethnicity 9

A model including the ethnicity:migration interaction term had significantly improved fit compared to a model without the interaction term, F(13, 9771) = 33.88, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.62 0.47 – 0.82 0.001
Migration: Not born in the UK 2.16 1.60 – 2.91 <0.001
Migration: Prefer not to say 1.88 0.94 – 3.79 0.076
Ethnicity: Arab/British Arab 1.86 0.39 – 8.76 0.435
Ethnicity: Asian/Asian British 1.48 1.09 – 2.00 0.011
Ethnicity: Black/African/Caribbean/Black British 1.53 1.00 – 2.35 0.049
Ethnicity: Don't think of myself as any of these 5.49 1.00 – 30.12 0.050
Ethnicity: Mixed/Multiple ethnic group 1.16 0.81 – 1.67 0.406
Ethnicity: Other ethnic group 3.35 0.65 – 17.23 0.147
Ethnicity: Prefer not to say 1.99 0.44 – 8.97 0.369
Ethnicity: White other 1.06 0.64 – 1.74 0.825
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.69 0.61 – 0.79 <0.001
Gender: Other 0.88 0.22 – 3.52 0.857
Gender: Prefer not to say 0.75 0.23 – 2.45 0.636
Education: Don't have degree 1.06 0.94 – 1.21 0.341
Education: Don't know 1.24 0.70 – 2.20 0.467
Region: East Midlands 0.95 0.72 – 1.26 0.744
Region: East of England 0.59 0.44 – 0.80 0.001
Region: North East 0.64 0.44 – 0.91 0.015
Region: North West 0.85 0.66 – 1.10 0.209
Region: Northern Ireland 0.69 0.44 – 1.08 0.106
Region: Scotland 0.69 0.51 – 0.94 0.020
Region: South East 0.62 0.49 – 0.79 <0.001
Region: South West 0.69 0.52 – 0.91 0.009
Region: Wales 0.92 0.67 – 1.26 0.597
Region: West Midlands 0.83 0.64 – 1.08 0.159
Region: Yorkshire and the Humber 0.68 0.52 – 0.90 0.007
Interaction: Not born in UK x Arab/Arab British 0.86 0.12 – 6.12 0.881
Interaction: Not born in UK x Asian/Asian British 0.52 0.32 – 0.87 0.012
Interaction: Prefer not to say x Asian/Asian British 0.26 0.05 – 1.27 0.096
Interaction: Not born in UK x Black/African/Caribbean/Black British 0.66 0.37 – 1.18 0.161
Interaction: Prefer not to say x Black/African/Caribbean/Black British 0.97 0.23 – 3.99 0.964
Interaction: Not born in UK x Don't think of myself as any of these 0.13 0.01 – 1.80 0.128
Interaction: Not born in UK x Mixed/Multiple ethnic group 1.27 0.60 – 2.70 0.530
Interaction: Prefer not to say x Mixed/Multiple ethnic group 0.39 0.03 – 4.64 0.453
Interaction: Not born in UK x Other ethnic group 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x Prefer not to say 0.49 0.04 – 5.99 0.577
Interaction: Prefer not to say x Prefer not to say 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x White other 0.53 0.28 – 1.01 0.053
Interaction: Prefer not to say x White other 0.59 0.10 – 3.61 0.564
Observations 9812
3.4.3.5.1.1 Post-hoc

We explored the interaction effect using targeted contrasts comparing

  1. The effect of each ethnicity versus “White British” within each level of migration.
  2. The effect of “Not born in UK” versus “Born in UK” within each level of ethnicity

Contrasts were calculated using survey::svycontrast() and p values were adjusted for multiple comparisons using the FDR method (Benjamini and Hochberg, 1995). Results were only considered where the sample for the contrast was greater than 10.

Exploring the effect of each ethnicity versus “White British” within each level of migration, we found that, among people not born in the UK, White other workers were 0.56 times as likely (i.e., 44% less likely) to be outsourced than “White British” people.

No differences by ethnicity were observed among people born in the UK.

Examining the effect of “Not born in UK” versus “Born in UK” within each ethnicity, we found

  • among “White British”, workers not born in the UK are 2.16 times more likely to be outsourced than workers born in the UK.
  • among people of Mixed/multiple ethnic groups, workers not born in UK are 2.74 times more likely to be outsourced than workers born in the UK.

No significant differences between people born and not born in the UK were observed for any other ethnicities. The figure below shows these effects.

3.4.3.5.2 Ethnicity 21

A model including the ethnicity:migration interaction term had significantly improved fit compared to a model without the interaction term, F(32, 9740) = 64.66, p < .001. The table below shows the model coefficients.

Note: Migration x Roma not estimable as model matrix rank deficient
  Outsourcing
Predictors Odds Ratios CI p
Intercept 0.63 0.47 – 0.83 0.001
Migration: Not born in the UK 2.15 1.59 – 2.90 <0.001
Migration: Prefer not to say 1.89 0.94 – 3.79 0.075
Ethnicity: Irish 0.93 0.44 – 1.96 0.846
Ethnicity: Gypsy or Irish Traveller 1.67 0.31 – 8.85 0.547
Ethnicity: Roma 0.81 0.19 – 3.49 0.775
Ethnicity: Any other White background 1.11 0.55 – 2.26 0.769
Ethnicity: White and Black Caribbean 0.53 0.27 – 1.04 0.066
Ethnicity: White and Black African 3.38 1.68 – 6.82 0.001
Ethnicity: White and Asian 0.90 0.37 – 2.19 0.817
Ethnicity: Any other Mixed/Multiple ethnic background 1.87 0.96 – 3.65 0.067
Ethnicity: Indian 1.32 0.80 – 2.19 0.280
Ethnicity: Pakistani 2.67 1.68 – 4.25 <0.001
Ethnicity: Bangladeshi 1.81 0.84 – 3.86 0.128
Ethnicity: Chinese 0.53 0.16 – 1.74 0.299
Ethnicity: Any other Asian background 1.06 0.36 – 3.11 0.916
Ethnicity: African 1.52 0.87 – 2.64 0.140
Ethnicity: Caribbean 1.12 0.49 – 2.53 0.787
Ethnicity: Any other Black, Black British, or Caribbean background 2.60 1.05 – 6.41 0.038
Ethnicity: Arab 1.85 0.39 – 8.69 0.435
Ethnicity: Any other ethnic group 3.34 0.65 – 17.32 0.150
Ethnicity: Don't think of myself as any of these 5.43 1.00 – 29.50 0.050
Ethnicity: Prefer not to say 1.99 0.44 – 8.97 0.370
Age 0.98 0.97 – 0.98 <0.001
Gender: Female 0.69 0.61 – 0.79 <0.001
Gender: Other 0.87 0.22 – 3.44 0.846
Gender: Prefer not to say 0.77 0.24 – 2.50 0.662
Education: Don't have degree 1.06 0.93 – 1.20 0.405
Education: Don't know 1.21 0.67 – 2.18 0.519
Region: East Midlands 0.94 0.71 – 1.25 0.692
Region: East of England 0.59 0.44 – 0.80 0.001
Region: North East 0.63 0.44 – 0.91 0.014
Region: North West 0.84 0.65 – 1.08 0.179
Region: Northern Ireland 0.71 0.44 – 1.14 0.153
Region: Scotland 0.69 0.50 – 0.93 0.017
Region: South East 0.62 0.49 – 0.79 <0.001
Region: South West 0.68 0.51 – 0.90 0.007
Region: Wales 0.90 0.65 – 1.24 0.524
Region: West Midlands 0.80 0.61 – 1.04 0.093
Region: Yorkshire and the Humber 0.67 0.51 – 0.88 0.004
Interaction: Not born in UK x Irish 0.28 0.07 – 1.19 0.085
Interaction: Prefer not to say x Irish 1.31 0.09 – 18.48 0.842
Interaction: Not born in UK x Gypsy or Irish Traveller 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x Any other White background 0.52 0.23 – 1.18 0.118
Interaction: Prefer not to say x Any other White background 0.38 0.04 – 3.92 0.413
Interaction: Not born in UK x White and Black Caribbean 0.00 0.00 – 0.00 <0.001
Interaction: Prefer not to say x White and Black Caribbean 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x White and Black African 0.46 0.14 – 1.52 0.203
Interaction: Prefer not to say x White and Black African 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x White and Asian 2.76 0.48 – 15.70 0.253
Interaction: Not born in UK x Any other Mixed/Multiple ethnic background 0.69 0.21 – 2.26 0.541
Interaction: Prefer not to say x Any other Mixed/Multiple ethnic background 1.10 0.03 – 41.99 0.960
Interaction: Not born in UK x Indian 0.55 0.26 – 1.17 0.120
Interaction: Prefer not to say x Indian 0.38 0.04 – 3.31 0.380
Interaction: Not born in UK x Pakistani 0.43 0.17 – 1.09 0.074
Interaction: Prefer not to say x Pakistani 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x Bangladeshi 0.38 0.10 – 1.48 0.162
Interaction: Prefer not to say x Bangladeshi 0.30 0.03 – 2.99 0.305
Interaction: Not born in UK x Chinese 0.88 0.21 – 3.70 0.864
Interaction: Not born in UK x Any other Asian background 0.90 0.26 – 3.13 0.864
Interaction: Prefer not to say x Any other Asian background 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x African 0.69 0.35 – 1.36 0.278
Interaction: Prefer not to say x African 0.89 0.19 – 4.18 0.886
Interaction: Not born in UK x Caribbean 0.84 0.14 – 5.19 0.851
Interaction: Prefer not to say x Caribbean 8523821.18 897250.55 – 80975740.17 <0.001
Interaction: Not born in UK x Any other Black, Black British, or Caribbean background 0.25 0.05 – 1.15 0.075
Interaction: Prefer not to say x Any other Black, Black British, or Caribbean background 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x Arab 0.86 0.12 – 6.11 0.880
Interaction: Not born in UK x Any other ethnic group 0.00 0.00 – 0.00 <0.001
Interaction: Not born in UK x Don't think of myself as any of these 0.13 0.01 – 1.82 0.129
Interaction: Not born in UK x Prefer not to say 0.49 0.04 – 5.98 0.579
Interaction: Prefer not to say x Prefer not to say 0.00 0.00 – 0.00 <0.001
Observations 9812
3.4.3.5.2.1 Post-hoc

We explored the interaction effect using targeted contrasts comparing

  1. The effect of each ethnicity versus “English / Welsh / Scottish / Northern Irish / British” within each level of migration.
  2. The effect of “Not born in UK” versus “Born in UK” within each level of ethnicity

Contrasts were calculated using survey::svycontrast() and p values were adjusted for multiple comparisons using the FDR method (Benjamini and Hochberg, 1995) and results were only considered where the sample for the contrast was greater than 10.

Exploring the effect of each ethnicity versus “English / Welsh / Scottish / Northern Irish / British” within each level of migration, we found that, among people born in the UK

  • White and Black African people were 3.38 times more likely to be outsourced than “English / Welsh / Scottish / Northern Irish / British” people.
  • Pakistani people were 2.67 times more likely to be outsourced than “English / Welsh / Scottish / Northern Irish / British” people.

Among people not born in the UK, no significant differences between ethnicities were observed. The figure below shows the effects for “English / Welsh / Scottish / Northern Irish / British”, “White and Black African”, and Pakistani respondents.

Examining the effect of “Not born in UK” versus “Born in UK” within each level of ethnicity, we found that among “English / Welsh / Scottish / Northern Irish / British”, workers not born in the UK are 2.15 times more likely to be outsourced than workers born in the UK.

No significant differences between people born and not born in the UK were observed for any other ethnicities. The figure below shows these effects.

3.4.3.6 Ethnicity-outsourced interaction

A generalised linear model was constructed to test whether the interaction between ethnicity and outsourcing predicted whether a person had a low income.

\[ Income Group = Outsourcing + Ethnicity + Age + Gender + Education + Region + Migration + Outsourcing:Ethnicity \]

As in preceding sections, we constructed three models; one for the binary ethnicity variable, one for the eight-level ethnicity variable, and one for the 21-level ethnicity variable.

3.4.3.6.1 Ethnicity binary

A model including the ethnicity:outsourcing interaction term was not a significant improvement on a model without the interaction term, F(3, 8073) = 0.53, p = 0.66. The table below shows the model coefficients.

  Low income group
Predictors Odds Ratios CI p
Intercept 0.09 0.06 – 0.12 <0.001
Outsourcing: Outsourced 1.53 1.27 – 1.84 <0.001
Ethnicity: Not White 1.13 0.89 – 1.44 0.320
Ethnicity: Don't think of myself as any of these 0.92 0.11 – 7.99 0.943
Ethnicity: Prefer not to say 1.31 0.38 – 4.46 0.666
Age 1.01 1.00 – 1.01 0.052
Gender: Female 3.26 2.85 – 3.73 <0.001
Gender: Other 3.68 1.14 – 11.88 0.029
Gender: Prefer not to say 2.04 0.43 – 9.62 0.365
Education: Don't have degree 2.60 2.29 – 2.94 <0.001
Education: Don't know 3.84 2.18 – 6.76 <0.001
Region: East Midlands 0.97 0.73 – 1.29 0.835
Region: East of England 0.95 0.71 – 1.28 0.746
Region: North East 0.75 0.52 – 1.08 0.122
Region: North West 0.68 0.52 – 0.90 0.006
Region: Northern Ireland 1.06 0.71 – 1.57 0.776
Region: Scotland 1.11 0.82 – 1.50 0.501
Region: South East 0.90 0.70 – 1.16 0.428
Region: South West 0.85 0.64 – 1.12 0.247
Region: Wales 0.62 0.43 – 0.90 0.011
Region: West Midlands 0.85 0.65 – 1.12 0.239
Region: Yorkshire and the Humber 0.86 0.65 – 1.14 0.292
Migration: Arrived within the last year 1.75 1.08 – 2.86 0.024
Migration: Arrived within the last 3 years 1.05 0.65 – 1.68 0.848
Migration: Arrived within the last 5 years 1.06 0.64 – 1.77 0.809
Migration: Arrived within the last 10 years 0.96 0.62 – 1.47 0.842
Migration: Arrived within the last 15 years 0.90 0.56 – 1.46 0.676
Migration: Arrived within the last 20 years 0.83 0.52 – 1.32 0.433
Migration: Arrived within the last 30 years 0.65 0.28 – 1.47 0.297
Migration: Arrived more than 30 years ago 0.89 0.55 – 1.45 0.641
Migration: Prefer not to say 1.86 0.82 – 4.25 0.139
Interaction: Outsourced x Non-White 0.87 0.57 – 1.33 0.530
Interaction: Outsourced x Don't think of myself as any of these 0.22 0.01 – 5.13 0.349
Interaction: Outsourced x Prefer not to say 2.49 0.11 – 54.84 0.563
Observations 8107
3.4.3.6.2 Ethnicity 9

A model including the ethnicity:outsourcing interaction term significantly improved model fit compared to a model without the interaction term, F(8, 8063) = 18.02, p < .001. The table below shows the model coefficients.

  Low income group
Predictors Odds Ratios CI p
Intercept 0.08 0.06 – 0.12 <0.001
Outsourcing: Outsourced 1.58 1.30 – 1.92 <0.001
Ethnicity: Arab/British Arab 1.29 0.30 – 5.51 0.734
Ethnicity: Asian/Asian British 1.11 0.79 – 1.55 0.537
Ethnicity: Black/African/Caribbean/Black British 1.44 0.99 – 2.08 0.057
Ethnicity: Don't think of myself as any of these 0.97 0.11 – 8.30 0.975
Ethnicity: Mixed/Multiple ethnic group 1.11 0.74 – 1.67 0.601
Ethnicity: Other ethnic group 0.00 0.00 – 0.00 <0.001
Ethnicity: Prefer not to say 1.33 0.39 – 4.55 0.644
Ethnicity: White other 1.18 0.82 – 1.70 0.384
Age 1.01 1.00 – 1.01 0.042
Gender: Female 3.30 2.89 – 3.78 <0.001
Gender: Other 3.81 1.17 – 12.39 0.026
Gender: Prefer not to say 2.05 0.44 – 9.67 0.364
Education: Don't have degree 2.61 2.30 – 2.95 <0.001
Education: Don't know 3.81 2.15 – 6.75 <0.001
Region: East Midlands 0.97 0.73 – 1.29 0.833
Region: East of England 0.95 0.71 – 1.28 0.736
Region: North East 0.76 0.53 – 1.09 0.134
Region: North West 0.68 0.51 – 0.89 0.006
Region: Northern Ireland 1.03 0.70 – 1.52 0.891
Region: Scotland 1.12 0.83 – 1.52 0.461
Region: South East 0.90 0.70 – 1.17 0.433
Region: South West 0.84 0.63 – 1.12 0.237
Region: Wales 0.62 0.43 – 0.90 0.011
Region: West Midlands 0.84 0.64 – 1.11 0.225
Region: Yorkshire and the Humber 0.87 0.65 – 1.15 0.321
Migration: Arrived within the last year 1.84 1.09 – 3.11 0.023
Migration: Arrived within the last 3 years 1.03 0.65 – 1.66 0.887
Migration: Arrived within the last 5 years 1.00 0.59 – 1.70 0.996
Migration: Arrived within the last 10 years 0.93 0.57 – 1.51 0.764
Migration: Arrived within the last 15 years 0.87 0.52 – 1.46 0.600
Migration: Arrived within the last 20 years 0.81 0.50 – 1.33 0.410
Migration: Arrived within the last 30 years 0.59 0.26 – 1.32 0.199
Migration: Arrived more than 30 years ago 0.89 0.55 – 1.45 0.639
Migration: Prefer not to say 1.88 0.80 – 4.41 0.149
Interaction: Outsourced x Arab/British Arab 1.04 0.08 – 13.13 0.978
Interaction: Outsourced x Asian/Asian British 0.81 0.46 – 1.43 0.474
Interaction: Outsourced x Black/African/Caribbean/Black British 0.35 0.17 – 0.72 0.004
Interaction: Outsourced x Don't think of myself as any of these 0.21 0.01 – 4.86 0.334
Interaction: Outsourced x Mixed/Multiple ethnic group 2.46 1.14 – 5.29 0.022
Interaction: Outsourced x Other ethnic group 1172194.21 101948.54 – 13477773.39 <0.001
Interaction: Outsourced x Prefer not to say 2.44 0.11 – 53.01 0.571
Interaction: Outsourced x White other 0.70 0.36 – 1.33 0.274
Observations 8107
3.4.3.6.2.1 Post-hoc

We explored the interaction effect using targeted contrasts comparing

  1. The effect of each ethnicity versus “White British” within each level of outsourcing
  2. The effect of outsourcing within each level of ethnicity

Contrasts were calculated using survey::svycontrast() and p values were adjusted for multiple comparisons using the FDR method (Benjamini and Hochberg, 1995). Results were only considered where the sample for the contrast was greater than 10.

Examining the effect of ethnicity within each level of outsourcing, we find that:

  • Among people who are outsourced, people of “Mixed/Multiple ethnic group” are 2.74 times more likely to be in the low income group than White British people.
  • Among people who are not outsourced, people of “Other ethnic group” are 0.0000034 times as likely (i.e. 99.9996584% less likely) than White British people to be in the low income group (NB sample size for this cell is 11).

No other significant differences were observed. The plot below visualises this.

Examining the effect of outsourcing within each ethnicity, we find that:

  1. Among White British people, outsourced workers are 1.58 times more likely to be in the low income group than non-outsourced workers.
  2. Among people of Mixed/Multiple ethnic groups, outsourced workers are 3.88 times more likely to be in the low income group than non-outsourced workers.

For all other ethnicities, there is no significant difference between outsourced and non-outsourced people in the odds of being in the low income group.

The plot below visualises these effects.

3.4.3.6.3 Ethnicity 21

A model including the ethnicity:migration interaction term had significantly improved fit compared to a model without the interaction term, F(20, 8039) = 13.1, p < .001. The table below shows the model coefficients.

  Outsourcing
Predictors Odds Ratios p
Intercept 0.08 <0.001
Outsourcing: Outsourced 1.58 <0.001
Ethnicity: Irish 1.85 0.033
Ethnicity: Gypsy or Irish Traveller 0.40 0.381
Ethnicity: Roma 0.00 <0.001
Ethnicity: Any other White background 1.09 0.704
Ethnicity: White and Black Caribbean 1.12 0.751
Ethnicity: White and Black African 0.75 0.626
Ethnicity: White and Asian 1.09 0.815
Ethnicity: Any other Mixed/Multiple ethnic background 1.36 0.424
Ethnicity: Indian 0.86 0.539
Ethnicity: Pakistani 1.43 0.253
Ethnicity: Bangladeshi 1.72 0.225
Ethnicity: Chinese 0.64 0.258
Ethnicity: Any other Asian background 1.72 0.168
Ethnicity: African 1.43 0.130
Ethnicity: Caribbean 1.00 0.993
Ethnicity: Any other Black, Black British, or Caribbean background 2.67 0.029
Ethnicity: Arab 1.26 0.757
Ethnicity: Any other ethnic group 0.00 <0.001
Ethnicity: Don't think of myself as any of these 0.97 0.976
Ethnicity: Prefer not to say 1.34 0.635
Age 1.01 0.032
Gender: Female 3.30 <0.001
Gender: Other 3.78 0.027
Gender: Prefer not to say 1.98 0.394
Education: Don't have degree 2.62 <0.001
Education: Don't know 3.80 <0.001
Region: East Midlands 0.98 0.899
Region: East of England 0.94 0.662
Region: North East 0.74 0.109
Region: North West 0.67 0.004
Region: Northern Ireland 0.95 0.785
Region: Scotland 1.12 0.466
Region: South East 0.89 0.377
Region: South West 0.83 0.211
Region: Wales 0.61 0.010
Region: West Midlands 0.83 0.183
Region: Yorkshire and the Humber 0.85 0.270
Migration: Arrived within the last year 1.90 0.022
Migration: Arrived within the last 3 years 1.14 0.615
Migration: Arrived within the last 5 years 1.06 0.845
Migration: Arrived within the last 10 years 0.91 0.723
Migration: Arrived within the last 15 years 0.89 0.671
Migration: Arrived within the last 20 years 0.84 0.506
Migration: Arrived within the last 30 years 0.58 0.221
Migration: Arrived more than 30 years ago 0.88 0.621
Migration: Prefer not to say 1.78 0.177
Interaction: Outsourced x Irish 0.27 0.133
Interaction: Outsourced x Gypsy or Irish Traveller 0.00 <0.001
Interaction: Outsourced x Roma 2.50 0.275
Interaction: Outsourced x Any other White background 0.83 0.591
Interaction: Outsourced x White and Black Caribbean 2.65 0.254
Interaction: Outsourced x White and Black African 1.55 0.626
Interaction: Outsourced x White and Asian 5.76 0.009
Interaction: Outsourced x Any other Mixed/Multiple ethnic background 2.64 0.178
Interaction: Outsourced x Indian 0.79 0.617
Interaction: Outsourced x Pakistani 1.06 0.902
Interaction: Outsourced x Bangladeshi 0.86 0.842
Interaction: Outsourced x Chinese 1.19 0.842
Interaction: Outsourced x Any other Asian background 0.28 0.068
Interaction: Outsourced x African 0.39 0.017
Interaction: Outsourced x Caribbean 0.59 0.622
Interaction: Outsourced x Any other Black, Black British, or Caribbean background 0.04 0.004
Interaction: Outsourced x Arab 1.04 0.974
Interaction: Outsourced x Any other ethnic group 3223238.39 <0.001
Interaction: Outsourced x Don't think of myself as any of these 0.22 0.340
Interaction: Outsourced x Prefer not to say 2.43 0.571
Observations 8107
3.4.3.6.3.1 Post-hoc

We explored the interaction effect using targeted contrasts comparing

  1. The effect of each ethnicity versus “White British” within each level of outsourcing
  2. The effect of outsourcing within each level of ethnicity

Contrasts were calculated using survey::svycontrast() and p values were adjusted for multiple comparisons using the FDR method (Benjamini and Hochberg, 1995). Results were only considered where the sample for the contrast was greater than 10.

Examining the effect of ethnicity within each level of outsourcing, we find that:

  • Among people who are outsourced, “White and Asian” people are 6.28 times more likely to be in the low income group than White British people (NB sample is 12 for this cell).
  • Among people who are not outsourced, people of “Any other ethnic group” are 0.0000012 times as likely (i.e. 99.9998755% less likely) than White British people to be in the low income group (NB sample size for this cell is 11).

No other significant differences were observed. The plot below visualises this.

Examining the effect of outsourcing within each ethnicity, we find that:

  1. Among “English / Welsh / Scottish / Northern Irish / British”, outsourced workers are 1.58 times more likely to be in the low income group than non-outsourced workers.
  2. Among people of “White and Asian” ethnicity, outsourced workers are 9.08 times more likely to be in the low income group than non-outsourced workers (NB sample for this cell is 12).

For all other ethnicities, there is no significant difference between outsourced and non-outsourced people in the odds of being in the low income group.

The plot below visualises these effects.

4 Analysis - Study 2

Analysis from Study 2 appearing in [NAME OF REPORT] primarily employs a descriptive approach to understand the data. We conducted several cross-tabulations focusing on key demographic variables including Migration Status, Low Pay, and Ethnicity.

Due to the extensive number of variables examined, these cross-tabulations are not reproduced in this document. However, researchers can easily recreate all analyses by running the “Crosstabulations.qmd” script available in the GitHub repository associated with this project (see @reproducibility). The repository contains all necessary data files and code to replicate our findings.

[TO CHECK WITH MORGAN WHETHER ANY OF THE MODELLING IS USED IN THE REPORT]

5 Limitations and Future Research?

6 Reproducibility

All analyses presented in this report can be fully reproduced using the code and data provided in the Just Knowlegde GitHub repository.

7 Appendices

7.1 Study 1 - Age

The table below shows weighted descriptive statistics of the sample, and the figure below shows the frequency of respondents at each single year of age.

Mean Median Min Max Standard dev.
42.1 42 16 80 13.17

7.2 Study 1 - Gender

The table below shows the weighted gender breakdown of the sample

Gender Weighted frequency Weighted percentage
Male 4957.18 48.82
Female 5117.61 50.39
Other 15.37 0.15
Prefer not to say 64.84 0.64

7.3 Study 1 - Ethnicity

The table below shows the weighted ethnicity breakdown using the full range of Census 2021 categories. Note that ‘NA’ indicates non-responses.

Ethnicity Weighted frequency Weighted percentage
English / Welsh / Scottish / Northern Irish / British 7732.24 76.14
Irish 113.61 1.12
Gypsy or Irish Traveller 10.79 0.11
Roma 7.49 0.07
Any other White background 479.38 4.72
White and Black Caribbean 58.75 0.58
White and Black African 35.04 0.35
White and Asian 41.52 0.41
Any other Mixed / Multiple ethnic background 49.49 0.49
Indian 311.73 3.07
Pakistani 149.94 1.48
Bangladeshi 76.50 0.75
Chinese 145.53 1.43
Any other Asian background 163.15 1.61
African 227.05 2.24
Caribbean 71.67 0.71
Any other Black, Black British, or Caribbean background 37.39 0.37
Arab 32.50 0.32
Any other ethnic group 30.46 0.30
Don’t think of myself as any of these 8.81 0.09
Prefer not to say 30.45 0.30
NA 341.51 3.36

We also make use of an aggregated ethnicity variable that groups ethnicities into fewer categories. The table below shows how the Census categories map onto the aggregated categories.

Census categories Aggregated categories
English / Welsh / Scottish / Northern Irish / British White British
Irish White other
Gypsy or Irish Traveller White other
Roma White other
Any other White background White other
White and Black Caribbean Mixed/Multiple ethnic group
White and Black African Mixed/Multiple ethnic group
White and Asian Mixed/Multiple ethnic group
Any other Mixed / Multiple ethnic background Mixed/Multiple ethnic group
Indian Asian/Asian British
Pakistani Asian/Asian British
Bangladeshi Asian/Asian British
Chinese Asian/Asian British
Any other Asian background Asian/Asian British
African Black/African/Caribbean/Black British
Caribbean Black/African/Caribbean/Black British
Any other Black, Black British, or Caribbean background Black/African/Caribbean/Black British
Arab Arab/British Arab
Any other ethnic group Other ethnic group
Don’t think of myself as any of these Don't think of myself as any of these
Prefer not to say Prefer not to say
NA NA

The table below shows the weighted ethnicity breakdown using the aggregated set of categories

Ethnicity Weighted frequency Weighted percentage
White British 7732.24 76.14
Arab/British Arab 32.50 0.32
Asian/Asian British 846.86 8.34
Black/African/Caribbean/Black British 336.10 3.31
Don't think of myself as any of these 8.81 0.09
Mixed/Multiple ethnic group 184.80 1.82
Other ethnic group 30.46 0.30
Prefer not to say 30.45 0.30
White other 611.27 6.02
NA 341.51 3.36