Conditional Probability

What Is Conditional Probability?

Conditional probability measures how likely an outcome is once you know something about related conditions. It answers a narrower question than a simple probability. Instead of asking how likely a stock is to fall on any given day, it asks how likely a fall is on days when the broader market has already dropped.

The notation P(A|B) means the probability of event A, given that event B has occurred. The vertical bar is read as "given."

How It Is Calculated

The standard formula is:

P(A|B) = P(A and B) / P(B)

In words, divide the joint probability that both events occur by the probability of the condition B. The formula only works when P(B) is greater than zero.

An Example

Suppose a hypothetical stock and a market index are tracked over 250 trading days. The index fell on 100 days, so P(index down) is 0.40. Both the stock and the index fell on 60 days, so P(both down) is 0.24. The stock fell on 110 days overall, so its unconditional probability of falling is 0.44.

Applying the formula, P(stock down | index down) = 0.24 / 0.40 = 0.60. Knowing the index fell raises the estimated chance of a stock decline from 44% to 60% in this sample.

If the two events were unrelated, the conditional probability would equal the unconditional one. Events with this property are called independent events. Comparing the two numbers is a simple way to check whether knowing one event changes the odds of another.

Why It Matters to Investors

Many market questions are conditional. Historical return patterns, correlations between assets, and the effect of economic data releases all depend on what else is happening at the time.

Common mistakes include:

• Confusing P(A|B) with P(B|A). A test that rarely gives false alarms does not mean a positive result is rarely false, because the answer also depends on the base rate.
• Treating a pattern in past data as a reliable condition for the future without checking the sample size.

Working with conditional probabilities is also the basis of Bayes' theorem, which updates estimates as new information arrives.