Authors: Sin Yi Cheung and Stephen McKay (College of Social Service, University of Birmingham)
Summary
Who receives training?
This report concerns the training that people receive while in work, or in anticipation of working in the future, and the effects it has on people’s careers. This is training received after the end of education (in most cases). There is a great variety of activities that count as training, and in the statistical analysis we consider how far kinds of training are associated with different outcomes (hourly wages, in particular.
Data and Methods
We draw on three datasets for this study: the Labour Force Surveys (LFS) 1994 to 2008; the Families and Children Study (FACS); and the British Household Panel Survey (BHPS). A range of descriptive statistics were employed to chart the trends over time. Binary logic regression, ordinary least square (OLS) linear regression, and fixed effects models were used to estimate the effects of training over and above those accounted for by individual traits.
In 2008, training was most commonly received by:
- younger people;
- women;
- those working in the public sector (especially in local government, health or the armed forces), or working for non-profit organisations;
- those working in larger organisations;
- those with higher qualifications;
- higher earners (those in the top quintile of earners); and
- those relatively new to the job (training to aid induction.
Trends in training 1994-2008
The proportion of workers aged 16-69 [1] in training rose from about 20 per cent in 1994 to reach a high of around 28 per cent in 2003. This trend has been flat or on the decline since then, and particularly from 2005 onwards. Both the LFS and BHPS show this downward trend in the last few years. Training is also seasonal to some extent, with a lower proportion of the workforce in the third quarter (reflecting, perhaps, less training over the summer months).
This recent downward trend is found amoung virtually all groups. An important exception is older workers, aged 50 or older, who continue to enjoy increasing rates of training provision.
Changes in wages and training, longitudinal description
Hourly wages rates grew by 4.4 per cent between the 2006 and 2007 BHPS interviews, for those respondents working at both waves of interviews. They grew by five per cent where a respondent had received some training, and by four per cent otherwise. The rate of growth was higher where training was received, irrespective of the level of wages in 2006.
The highest increases in hourly earnings between 2006 and 2007 were achieved by young people, those aged between 16 and 34, and especially those at the younger half of this range.
For most age groups, except those under age 20, the rate of wage increase was raised if they had undergone a period of training.
Those who received training, compared to those who had not, showed greater variability in job satisfaction. That is, where a person had received training, they were both more likely to report an increase in job satisfaction, and more likely to report decreased job satisfaction. By contrast, there was greater stability in the reported levels of job satisfaction among those who did not receive training.
Changes in wages and training, longitudinal modelling 1998-2007
We look at the link between higher wages and having undertaken a spell of training in the recent past. This is based on data that tracks people over time. We first use models that control for a wide range of different background information. We then turn to look at statistical models that control for the unmeasured characteristics of people.
In standard linear regression models, the wage gain (measured by an increase in hourly earnings) to training (where received in the past year) were four per cent for men, and closer to two per cent for women. Modelling the median returns to training by quantile regression, rather than looking at the mean returns to training using the standard approach, produced quite similar results.
When we do not control for differences in individual traits (age, marital status, occupation) the increases in wages associated with past training appear to be much larger. This implies that what might appear to be the effect of training on wages is often largely due to differences in individual traits. Hence, it is important to control for these differences to isolate the specific effect of training on wage progression.
The current ‘state of the art’ within econometrics recommends the application of fixed-effect models to investigate the effect of training on wage returns. The purpose of these models is essentially to use individuals as their own control group in looking at changes in earnings and training. This provides a better estimate of the contribution of training to wage growth, as it controls for unobserved characteristics of individuals.
The estimated effect of training on wages is much reduced in these fixed-effects models. Training is then associated with an increase in wages of about 0.5 per cent, measured over the period from 1998-2007 [2]. However, where the training received was explicitly employer-funded or employer-provided, the size of gain was closer to two per cent.
If we adopt the recent suggestion in the econometric literature and restrict the analysis to only those who anticipated receiving training, the effects of training on wage progression can become statistically insignificant. This is a less tried and tested approach than the above statistical models.
Training does, however, seem to be strongly linked to labour market transitions – that is, undergoing a period of training seems to increase the rate of returning to work, and decreases the likelihood of job exit.
Limitations of the study
One of the difficulties in comparing findings across studies is the different measures and definitions used. They can be different even within the same study such as this one. We rely on existing data sets to provide detailed information on the duration, nature and type of training. The LFS essentially merges training with some aspects of education, while the BHPS does better at keeping them distinct. The degree of detail available in these data sets is also different.
1 We used ages 16-69 for most of this report. This takes advantage of the wider group of people to whom the questions were asked, not just those of pre-pension age, and permits some analysis of the increasing proportion of people who work after the age of 65. See Section 2.1.1 for further discussion of this selection.
2 In the BHPS, which we use for longitudinal analysis, training is measured by the question that mentions ‘training schemes or courses… or completed a course of training which led to a qualification’.
No comments:
Post a Comment