Productivity: time to ask employees?

There’s been a lot of discussion about productivity during the last few years: both trying to explain why labour productivity has hardly increased since 2008 and working out what can be done to increase the UK’s productivity.

CIPD asked employers questions about productivity in its summer 2014 and summer 2015 Labour Market Outlook surveys, analysed in reports available here and here. In these reports, we looked at the extent to which workplace-related factors could explain variation between private sector firms in their (self-reported) productivity. According to this research, holding other things constant, investment in capital equipment, training, learning and development and (some) people management practices are all associated with higher relative productivity.

 In the autumn 2015 Employee Outlook survey we asked employees a similar question, whether they thought they were more or less productive than their colleagues. The survey report provided some initial analysis of this question, together with data on what employees thought made them more or less productive.

The distribution of responses to the productivity question was highly skewed: 43% of employees think they are more (or much more) productive than their colleagues but just 5% think they are less productive. This bias is widespread. One example of this is over-confidence in our own abilities which means, for example, that most people think they’re safer than the average driver. It can also apply even if they are making this judgement on behalf of someone or something else (such as the HR representatives surveyed in the Labour Market Outlook, as shown in the table below).

This doesn’t mean the data are useless. If we’re prepared to assume the bias is widespread, data like these still have value in helping us understand why some people are more productive than others.

As shown in the chart below, the distribution of responses is very similar for men and women. There is a (statistically significant) age-related dimension to relative productivity, however. The youngest and oldest employee groups (those aged 18-24 and 55+) have lower proportions rating themselves more productive than their colleagues. Employees under 25 will be aware they are starting on their careers. And over-confidence is thought to diminish with age, even if it isn’t eliminated. So older employees may simply be more realistic when rating their individual productivity.

The chart below shows a similar analysis for various job- and organisation-related characteristics. There is no significant difference in the pattern of responses between private, public and voluntary sector employees. Remember: this does not tell us whether employees in the private sector are more or less productive than employees in the public sector. Nor does it tell us whether employees in one sector think they are more productive than employees in another sector. Private sector employees are presumably comparing their productivity with their private sector colleagues and public sector employees are presumably comparing their productivity with their public sector colleagues. It would be surprising if we did see systematic variation in these responses by organisation-specific characteristics such as sector or industry.

There is one possible exception in this chart. Employees in the smallest organisations (with 2-9 employees) are more likely to rate their productivity as average. When every employee is likely to know everyone else personally, employees may find it a little harder to convince themselves that they stand out from the crowd.

Junior, middle and senior managers are more likely to rate their productivity above average. Presumably their reference group will include the people they are managing.

The Employee Outlook survey collects a lot more data on how employees feel about their job, their manager and their employer. As many of these variables are correlated with each other, looking at each in isolation could lead to many spurious results. I therefore carried out a similar analysis to that done for the employer-provided data, estimating a multivariate model that aims to ‘explain’ the variability in responses to the productivity question through variation in other factors expected to influence productivity. The model used 38 different independent variables that, between them, covered

  • Demographic characteristics (age, gender etc.)
  • Organisation-specific characteristics (sector, industry etc.)
  • Job-specific characteristics (job tenure, managerial status etc.)
  • Job satisfaction
  • Views about line managers and senior managers
  • Work-life balance
  • Engagement and commitment
  • Challenge and opportunities to progress

From this model, eight variables turned out to be significant in explaining variation in individual productivity ratings, listed in the table below. [Technical details: An ordered logit model was fitted to the data and these are the eight variables that passed a joint significance test at the 5% confidence level].

The multivariate analysis confirms the results for age and managerial status visible in the charts. There are also some results which have common sense interpretations. The minority of employees who say they are dissatisfied with their job are less likely to rate their relative productivity highly – a failure to be as productive as they would like to be (for whatever reason) may be the source of their dissatisfaction; alternatively, dissatisfaction may mean they recognise they are not working as hard or effectively as their colleagues. Sensing that senior managers have a clear vision for the organisation and understanding very clearly the organisation’s core purpose are associated with higher productivity ratings – perhaps these individuals think they are especially focused on what matters?

 There are also a couple of results that are more challenging to rationalise. The 30% of employees who see themselves as over-qualified for the job are more likely to regard themselves as more productive than their colleagues. This could be a genuine performance effect arising from higher level skills but sour grapes can’t be ruled out either (“I’m the only graduate here and wasted in this job, so I must be more productive than everyone else”).

 Even more intriguing is a result suggesting that employees who disagree or strongly disagree that senior managers treat them with respect – a widespread perception, held by 29% of employees – are more likely to rate their productivity highly than colleagues. Is a failure to recognise or reward the employee’s talents – either real or imagined – taken as evidence of a lack of respect? This is a good point to remind ourselves that any statistical analysis will occasionally generate false positives.

 The other noteworthy point about this analysis is how poorly the regression model performed in accounting for the variation in individual responses. Of the 38 variables tested, only eight were (jointly) significant. The Pseudo R² – a goodness of fit measure applicable to models of this kind – was 0.10. By comparison, the final model explaining firm-level relative productivity estimated using the employer-provided data in the summer 2015 Labour Market Outlook survey had a pseudo R² of 0.33.

 There are possible explanations for these (unexpectedly) weak results. There may be a lot of random error in these answers. In other words, employees don’t have a very strong sense of their relative productivity at all – apart from a general sense of over-confidence in their abilities – so employees tick a box (almost) at random. This doesn’t seem especially credible. There was a ‘don’t know’ category available and only 3% of employees chose this option.

 A second possibility is that factors not measured in the survey may have an important bearing on how employees answer these questions (omitted variable bias is the technical term). For example, individuals with strong narcissistic tendencies are likely to rate themselves highly regardless of their ability and the survey did not collect any data on personality traits.

 Another potential explanation is that some of the variables thought likely to affect an individual’s productivity – such as organisational commitment – might vary much more across work groups and organisations than within work groups and organisations. Getting round this problem requires data that links the performance of individual employees with that of their employers.

CIPD has argued that people are at heart of productivity. We’ve put our toe in the water by collecting data from employees on how productive they think they are. It’s fair to say the results have raised as many questions as answers. We need a better understanding of what employees think ‘productivity’ means and how they form judgements about the productivity of themselves and those around them.

Thank you for your comments. There may be a short delay in this going live on the blog page as we moderate the comments added to our blogs.