Weng, Ali & Leonardi-Bee (2012)
Smoking and absence from work: Systematic review and meta-analysis of occupational studies is a meta-study of 29 other studies that purports to show that smokers took 2.74 (2 days 5 hours, 55 minutes and 12 seconds,) more days off work than non-smokers, ex smokers had a 14% increase in risk of absenteeism but strangely no extra days off could be detected. Smokers had a 19% increase in risk over ex-smokers and 33% increase in risk on never-smokers.
Note that since no absolute figures are given in the abstract, the phenomenon of small numbers may be at play here. If in a workforce of 1000 people, 3 non-smokers are likely to have a day off, the results (if believed) would indicate that 4 smokers would also have a day off.
Using magic formulae not presented in the abstract, they calculated that this equates to a 'cost' of £1.4bn in 2011.
No obvious mention is made about the occupations of the people studied - if it is even known; for example it may be that the jobs where more people are likely to smoke (outdoor jobs for example) have higher absenteeism rates (because being outside in rain/snow is more likely to impair the immune system making the subject more susceptible to colds/flu.)
Additionally, and rather glaringly, no figures are actually given for non-smokers; specifically, how many absolute days are taken off by non-smokers (or indeed, smokers?) How much does do non-smokers 'cost' businesses?
As can be seen from the results, 29 studies were meta-analysed. No mention is made in the abstract of the total number of people the 29 studies covered, but if other research on other smoking matters is to be any guide, it won't be very many.
Note the selective use of the 29 studies - the most used for any particular result was 17 (58.621% for those that like percentages to ridiculously silly precision.) The least used was 8 (27.5862%)
Note the use of a weasel word, which indicates that perhaps they don't quite believe their own results, or that more primary research is required rather than cherry picking 29 studies from, what is presumably, the thousands available.