Bayesian insight

For simplicity, we are rounding the numbers.

Let’s consider 1010 companies: 1000 are clean, 10 are to avoid.

Of the 1000 cleans, 100 will be identified as culprit since the test is 90% valid.

Of the 10 culprits, 9 will be identified.

 

In total, the test therefore identifies 109 culprits out of which 100 are “false positives”.

Spotted stocks have a 100 out of 109 chance of being clean, or 92%, and 9 out of 109, or 8%, of being guilty.

So the answer is: less than 10 %