Why a Third of Sick Leave Starts on a Monday
The short version: those numbers are what honest illness looks like. Monday and Friday are two of five working days, so 40% of sick days fall on them by default. And illnesses that start on Saturday or Sunday first show up at work on Monday, so in our model of one million illnesses that can start on any day of the week, 33% of absences began on a Monday, twice the share of any other day. Before calling a pattern suspicious, work out what chance alone would give. The suspicious sign would be more than that.
HR has analysed sick leave. The findings were presented with a gasp: a third of all absences start on a Monday. And 40% of all sick days are Mondays or Fridays. "People are extending their weekends," says the HR director. A new policy is coming: a doctor's note for any Monday absence.
Kevin had a temperature on Sunday. He is not sure which part of the weekend he was extending.
The 40%
Start with the easy one. A working week has five days, and two of them are Monday and Friday. If sick days fell on working days completely at random, Mondays and Fridays together would get 2 out of 5 of them: 40%. HR's shocking number is the number you'd expect if nothing was going on at all.
The question that matters is never "is the number big?" but "is it bigger than chance?" That chance level is the base rate.
The Monday third
The second number needs a small model. Illness doesn't check the calendar: you're as likely to fall ill on a Saturday as on a Wednesday. We simulated a million illnesses, each starting on a random day of the week and lasting one to three days, and recorded the first working day each person missed.
| First day of absence | Share of absences (honest illness) |
|---|---|
| Monday | 33% |
| Tuesday | 17% |
| Wednesday | 17% |
| Thursday | 17% |
| Friday | 17% |
Monday collects the illnesses that start on Monday, plus those that started on Saturday or Sunday and are still going. So a third of absences start on a Monday, twice as many as on any other day, with nobody faking anything. Some weekend illnesses are over by Monday and never appear at all: in our model, 14% of illnesses fitted entirely inside a weekend.
First day of absence, when illness can start on any day of the week
What a real test would look like
To find faked Mondays, you need a comparison with the right expectation. Some options: are absences after a weekend longer or shorter than absences starting midweek? Do Monday absences rise after big sporting events or bank holidays, compared with ordinary Mondays? Is the Monday share higher than the share you'd expect from honest illness for your own mix of absence lengths? Each of these compares the data with what chance predicts, instead of with zero.
Statistics for noobs: compare with chance, not with nothing
Many "shocking" patterns are just the calendar, the size of groups or the number of options. Before being shocked, write down what you'd see if nothing were going on. A pattern is only interesting by how far it differs from that.
Where it costs you money
"A third of big market falls start on a Monday" sounds like a pattern to trade. If Monday prices include the news of the whole weekend, some of that is the calendar again. Compare any day-of-the-week effect with what the extra hours of news alone would give before betting on it.
The moral
The doctor's-note policy came in. Monday absences fell slightly. Tuesday absences rose, since Kevin's colleagues now spend Monday at the doctor's surgery getting a note. HR has found that a suspicious number of absences now start on a Tuesday.
What this model can't tell you
- A simple model. Real illnesses vary in length, and some people work while ill. Different assumptions change the exact shares, not the Monday effect.
- Faking isn't ruled out. Some Monday absences may be weekend extensions. The model shows that the headline numbers can't prove it, not that it never happens.
- Weekends differ. Shift workers and part-timers have different calendars and different base rates.
HR, the policy and Kevin's temperature are an illustration. The figures come from our own simulation of one million illnesses starting on a random day of the week and lasting one to three days. Educational content, not HR, legal or financial advice.