Review of Job loss, job insecurity and health
By: Ronán M Conroy
 
 

This is an important paper on an important topic which provides a useful addition to existing knowledge in two ways: the first is that the baseline assessments were carried out before the employees had any inkling of the privatisation of their employment, with the consequent loss of job security and, indeed, jobs. The second important contribution is has to make is that it is data from the UK, taken in the context of a programme of privatisation which is still continuing. As such, it can inform the debate on the desirability of privatisation, and in particular it can draw attention to the existence of consequences other than cost savings. For this reason I believe that it merits a general rather than specialist audience.

The underlying research question is well defined and the analysis is a well-conducted examination of it. Because of the potential confounders, as well as the number of possible explanatory variables, the analysis is a rather tough read. I have made some suggestions, but I feel that there's no easy way of presenting the results.

The paper on the whole is well-written and the presentation and discussion of the findings is well balanced.

Specific comments

1. Introduction

1. The history of privatisation in the UK (page 4, paragraph 2) is given in
excessive detail.

2. Methods

1. The range of salaries does not provide a good picture of the extent of
variation. The 90th and 10th (or 95th and 5th) percentiles would be more
useful.

2. I feel that the similarity in health outcome between those in secure
employment and those no longer seeking work is worth including in the
paper as a result rather than explaning the omission of the latter group
from the analysis. I think that it strengthens the authors' case that the
effects are due to job insecurity that two different secure groups—one in
work and the other out of work—have similar health profiles.

3. Adjusted means and confidence intervals would be preferable to the
standard errors reported.

4. Some of the variables such as symptom score, GHQ, and life event count
would be expected to have binary error structures. The authors should
verify that the models they constructed took this into account.

3. Results

1. All tables needs confidence intervals throughout, and one decimal place.

2. The caption of table 2 should make it clear that the insecure and
unemployed are being compared with the securely employed; this
information is too important to relagate to one of many footnotes.

3. Table 4 is an unenviable task of formatting, but I think that the present
form could be improved by bringing the information from the footnotes up
into the table itself, and prefacing each row with the factors being
adjusted for.

RCSI BIOSTATISTICS SERVICE

CSI B