How do I build a frequency table of two categories in pandas?

The short version

Use pd.crosstab(rows, columns) to count how often each combination of two categorical columns occurs, arranged as a grid with one category down the side and the other across the top. Add normalize=True for proportions instead of counts, and margins=True to include row and column totals.

A worked example

import pandas as pd

df = pd.DataFrame({
    "city": ["Leeds", "Leeds", "Bath"],
    "size": ["S", "L", "S"],
})

print(pd.crosstab(df["city"], df["size"]))

size   L  S
city
Bath   0  1
Leeds  1  1

Each cell counts the rows with that city and size combination. Bath has one small and no large; Leeds has one of each.

Proportions instead of counts

normalize turns counts into shares. normalize="index" makes each row sum to 1, "columns" makes each column sum to 1, and True divides by the grand total.

pd.crosstab(df["city"], df["size"], normalize="index")

This answers questions like "what proportion of Leeds rows are large", which raw counts do not show directly.

Adding totals

margins=True adds an "All" row and column with the totals, handy for a quick sanity check that the parts sum correctly.

pd.crosstab(df["city"], df["size"], margins=True)

crosstab versus value_counts versus pivot_table

value_counts counts one column. crosstab is the natural tool for counting the interaction of two columns. pivot_table is the more general tool that can do the same and more, aggregating a third value column with any function; crosstab is really a convenient special case of it for counting. Reach for crosstab when you specifically want a count grid of two categories, and pivot_table when you need to summarise a numeric column across them.

Aggregating a value instead of counting

If you want to summarise a numeric column rather than count rows, pass values and aggfunc, which pushes crosstab towards pivot_table territory.

pd.crosstab(df["city"], df["size"], values=df["amount"], aggfunc="mean")

A note on missing combinations

Combinations that never occur show as 0, not as absent rows, which is usually what you want for a complete table. If a category value is itself missing (NaN), it is dropped by default; pass dropna=False to keep it as its own row or column.