Migrant Work Ethic

Table 2: A8 migrants and work absence – Regression results
	(1)	(2)	(3)	(4)
	Sickness Absence Probability	Sickness Absence Rate	Overall Absence Probability	Overall Absence Rate
A8	-0.0333***	-0.0211***	-0.0354***	-0.0202***
	[0.0058]	[0.0045]	[0.0069]	[0.0049]
A8*(Years in UK)	0.0080***	0.0063***	0.0072***	0.0059***
	[0.0021]	[0.0018]	[0.0022]	[0.0018]
Control variables	Yes	Yes	Yes	Yes
Source: UK QLFS 2005-2012. Notes: Sample size for all models is 113,804 observations; OLS estimates; Huber-White (robust) standard errors in brackets; all models include controls for gender, age, age squared, marital status, health, education, number of dependent children, age of youngest dependent child, usual basic hours of work, paid and unpaid overtime hours, public sector, permanent contract, holding a second job, working from home, managerial status, looking for new job, looking for extra job, number of fewer working hours desired, job tenure, establishment size, trade union status, flexible working arrangements, housing tenure, claiming any benefits and its interaction with A8 migrant status, industry, occupation, region of residence and survey year; full results are available from the authors upon request. *** Significant at 1%.

Figure 2: A8-UK Absence Gaps by Years of Migrant Residency in the UK

Source: UK QLFS 2005-2012.

Notes: The bold lines show the estimate of the A8-UK coefficient for each length of residency in the UK; the dashed lines indicate the 95% confidence intervals.

TABLE 3: Wage regression results
	(1)	(2)
A8 0-3 Years in UK	-0.5213***	-0.5633***
	[0.0294]	[0.0321]
A8 4-8 Years in UK	-0.4376***	-0.4812***
	[0.0242]	[0.0249]
Control variables	No	Yes
F-test for equality of A8 dummies (p-value)	0.0266	0.0400
Source: UK QLFS 2005-2012. Notes: Sample size for all models is 28,521 observations; dependent variable in both models is the log of the real hourly wage; OLS estimates; Huber-White (robust) standard errors are in brackets; controls include: female, age, age squared, education, marital status, health, survey year and region of residence dummies; full results are available from the authors upon request. *** Significant at 1%.

Table A1: Sample means for all variables by nationality

Variable	UK Nationals	A8 Migrants
Demographic characteristics
Female	0.535	0.451***
Age	42.526	31.112***
Education (years)	11.905	14.044***
Long-term Health Problem	0.262	0.092***
Long-term Health Problem Limits Work	0.059	0.029***
Married or Cohabiting	0.687	0.633***
Number of dependent children <16 y.o.	0.592	0.473***
Dummy Variables for Age of youngest dependent child:
1. 0-2 years	0.082	0.146***
2. 3-4 years	0.045	0.044
3. 5-9 years	0.101	0.083**
4. 10-15 years	0.133	0.061***
5. 16-18 years	0.046	0.014***
6. No child	0.592	0.652***
Job Characteristics
Dummy Variables for Usual Basic Hours
1. 1-15	0.058	0.015***
2. 16-29	0.178	0.099***
3. 30-35	0.159	0.082***
4. 36-40	0.489	0.620***
5. 41-48	0.072	0.134***
6. 48+	0.043	0.050
Public sector	0.326	0.055***
Paid Overtime Hours	1.143	2.384***
Unpaid Overtime Hours	1.792	0.214***
Holding Second Job	0.043	0.026***
Working from Home or Same Building	0.061	0.021***
Permanent Job	0.962	0.892***
Manager/Foreman/Supervisor	0.419	0.150***
Fewer Hours Desired (Number of hours)	1.049	0.261***
Looking for New Job	0.056	0.090***
Looking for Extra Job	0.008	0.015***
Dummy Variables for Establishment Size:
1. Size 1-24	0.324	0.255***
2. Size 25-49	0.137	0.135
3. Size 50-499	0.344	0.465***
4. Size 500+	0.195	0.145***
Dummy Variables for one-digit occupation:
1. Managers and S.O.	0.156	0.028***
2. Professionals	0.161	0.039***
3. Ass. Profess. And Technical	0.157	0.042***
4. Administrative and Secretarial	0.142	0.051***
5. Skilled Trades	0.070	0.117***
6. Personal Services	0.090	0.068***
7. Sales and Customer Services	0.068	0.030***
8. Plant and Machine Operatives	0.062	0.226***
9. Elementary	0.094	0.400***
Tenure Dummies:
1. 0-3 months	0.036	0.091***
2. 3-6 months	0.034	0.109***
3. 6-12 months	0.052	0.138***
4. 12-24 months	0.091	0.218***
5. 24-60 months	0.211	0.359***
6. 60+ months	0.575	0.085***
Dummy Variables for Industries:
1. Agriculture and Fishing	0.006	0.024***
2. Energy and Water	0.016	0.011
3. Manufacturing	0.127	0.297***
4. Construction	0.049	0.047
5. Distribution, Hotels and Restaurants	0.160	0.261***
6. Transport and Communications	0.076	0.112***
7. Banking, Finance and Insurance	0.154	0.112***
8. Public Admin., Education and Health	0.368	0.105***
9. Other Services	0.044	0.031**
Dummy variables for trade union status:
1. Covered member	0.242	0.076***
2. Covered non-member	0.123	0.100***
3. Not covered member	0.085	0.031***
4. Not covered non-member	0.549	0.794***
Dummy variables for flexible working arrangements:
1. Flexitime	0.140	0.045***
2. Annualized hours contract	0.049	0.029***
3. Term time working	0.054	0.001***
4. Other flexible arrangement	0.028	0.014***
5. No flexible arrangement	0.073	0.910***
Hourly wage
Real hourly wage	11.806	6.627***
Log of real hourly wage	2.303	1.816***
Housing Tenure & Benefits
1. Outright Owner	0.172	0.007***
2. Owned with Mortgage	0.621	0.080***
3.Private Renter	0.114	0.812***
4. Social Housing	0.093	0.101
Claims any benefits	0.306	0.251***
Observations	112,408	1,396
Source: UK QLFS 2005-2012. Notes: Numbers in table are sample means; total sample size is 113,804 observations; the sample size for average wage calculations is 28,521; *** Sample mean difference significant at 1%; ** at 5%.

1^ Throughout this article, it is argued that absence from work is a reasonable proxy of work ethic. This measure is the best available in our data source and can provide useful insights that are consistent with our theoretical framework set out below. Other relevant proxies, e.g. work intensity, are not available in the dataset.

2^ According to the Office for National Statistics, only around 70% of the Polish- and Latvian-speaking population in England and Wales can speak English “very well” or “well” (ONS, 2013).

3^ An alternative sample selection procedure would be to combine data from all quarters and only select individuals in their first wave in the survey to avoid repeated observations. However, certain variables used in the analysis (e.g. trade union status and flexible working arrangements) are only available in the October-December quarters of each year. Finally, all results presented below are based on unweighted data. Using weights to account for non-response and make the QLFS samples representative of the UK population, produced nearly identical results.

4^ Employees that did not work fewer hours than usual in their reference week do not answer this question. Hence, = 0 for these individuals. Other reasons for absence can include a variety of factors, ranging from dealing with a personal/family errand to pure shirking. These, of course, are closely related to the concept of effort we want to capture, and complement the more multifaceted phenomenon of sickness absence.

5^ Though the appropriate models would be binary choice ones in the case of the discrete dependent variables, we choose to present results from linear models estimated by OLS for ease of interpretation. The estimation of non-linear models for the two binary dependent variables gave qualitatively and quantitatively similar results. Moreover, the two fractional dependent variables (the absence rates) can also cause problems in standard statistical analysis. However, the estimation of fractional probit models also gave very similar results.

6^ For the model to be estimated, UK nationals are assigned a value of zero for the Years in UK variable.

7^ Education is captured as a continuous variable, computed from the age an individual left full-time education minus six. The QLFS does provide an alternative coding framework based on the UK education system. However, up to 2010, foreign qualifications were only recorded as “other qualifications” in the QLFS, irrespective of their level.

8^ Benefits include: income support (not as an unemployed person), sickness or disability benefits, family related benefits, child benefits, housing/council tax benefits or rent rebate, tax credits or other.

9^ Note that a wage variable is not included in the final models. Although earnings should be an important determinant of absence through an opportunity cost of absence or an “efficiency wage” argument (whereby employers pay workers above the “market” wage in order to increase their effort/productivity and reduce the costs associated with turnover), the inclusion of a wage variable is likely to lead to simultaneity bias (see Allen, 1984). Moreover, earnings questions are only asked to employees in their first and fifth wave in the LFS and the inclusion of the wage in the models would substantially decrease the final sample. We return to the issue of wages below.

10^ Due to the very small number of A8 migrants with eight years of UK residency in the sample (see Figure 1), a single dummy for seven or eight years of residency is used in these models.

11^ The use of cross-sectional data when analysing the assimilation of migrant wages has come under some scrutiny in the relevant literature (Borjas, 1985). Firstly, if there is a decrease in the quality of migrants belonging to different entry cohorts, migrant wage growth may be upward biased (Borjas, 1985). However, this typical shortcoming seems unlikely within the 9-year period examined here. Secondly, poorly performing migrants are typically the first to return home; consequently, the sample of A8 migrants with longer residency may be a selection of better than average migrants. This phenomenon would also lead to migrant wage growth being upward biased. An examination of such issues requires a longitudinal dataset and is out of the scope of this study.

12^ Future work can also extend the empirical analysis presented in this article to other migrant groups from different countries of origin. This may also shed light on the issue of English-language proficiency by examining the behavior of migrants from English-speaking countries.