Interpreting an odds ratio from a logistic for a continuous variable instead of a dummy

[Alex Jeffries]

I've got a logit for the probability somebody takes out insurance based on some personal characteristics, with most of the variables being dummies.
I interpret an odds ratio of say 50 for one of my dummy variables as the probability of insurance increases 50 to 1 if D=1.

However I also have a variable which is just income in pounds, which gives an odds ratio of 1.005.
So £1 more of income raises the odds of insurance by 1.005, which means very little.
What do I do to the coefficient if I want to say the probabiliy rises X% for every £10 or every £100, I'm not entirely sure what to do with the coefficient?

Regards

[Rich Goldstein]

there are two ways to handle this; I prefer the following: decide in advance what size "unit" you want and make a new data with that unit; for example, if you want every 100 pounds, then divide income by 100 (I suggest that 100 pounds is probably too small; when using US incomes, I usually divide by $10,000)

[Jorrit Gosens]

Another way to have your reporting be more in touch with reality is to look at the distribution of income. Example: if the average salary in the top quintile is 10 times as high as in the lowest quintile, you expect a 5% higher chance that these most well of are insured, compare to the the least well of. You might also consider log transforming income data, as this is usually quite skewed. This is not necessarily a big issue for finding statistical significance, but could be an issue when you look at the distribution as I suggest here.

Edit:
That's not exactly correct. You should look at differences in real values in GBP, not nr of times higher in group X vs Y