Two-Period DiD From a 2-by-2 Table

From a 2-by-2 Table to a Difference-in-Differences Contrast

A two-period difference-in-differences analysis begins with four group-by-time outcome means: one for each combination of group and period. The calculation then compares two changes. First, measure how the outcome mean changes from before to after for the treated group. Then measure the same change for the control group. The difference-in-differences (DiD) contrast is the treated-group change minus the control-group change.

This subtraction is easy to reverse or misread when the four means are presented in a table. The most reliable approach is to name the four cells, calculate each group’s change separately, and only then subtract. At this stage, the result is a descriptive contrast in the outcome means. Calculating it does not, by itself, establish that the contrast is a causal effect.

The notation describes observed group-time means, not individual-level outcomes. For example, \(\bar Y_{T,0}\) is the mean outcome among the treated-group observations in the pre period. The four cell means can be laid out as follows:

The period order matters: a change is defined here as post minus pre for both groups. The group order matters too: the control change is subtracted from the treated change. Applying the same order in both comparisons makes the calculation transparent and keeps the sign interpretable.

How to Calculate the Contrast

For a numerical table, calculate in three steps: find the treated-group change, find the control-group change, then subtract the second change from the first. Keep the units throughout. If the outcome is measured in minutes, the two changes and the DiD contrast are also measured in minutes. If the outcome is a proportion, the contrast is in proportion units unless it is converted to percentage points.

Computing Cell Means Before Computing DiD

Sometimes the table gives cell totals and sample sizes rather than the four means. In that case, first calculate each cell mean as its outcome total divided by its number of observations. Only after obtaining all four means should you calculate the two changes and their difference. This separates the arithmetic of constructing the table from the arithmetic of the DiD contrast.

Understanding the Sign and the Components

The sign of the contrast comes from the direction of the two changes and from the order of subtraction. A positive DiD contrast means the treated-group change is numerically greater than the control-group change. A negative contrast means it is numerically smaller. For outcomes where lower values are desirable, a negative contrast might describe a larger decrease in the treated group, but whether that is a beneficial or causal effect depends on the context and assumptions beyond the calculation.

A useful algebraic rearrangement makes the four-cell structure explicit:

The post-period control mean has a minus sign, while the pre-period control mean has a plus sign. These signs follow from subtracting the entire control change, including its parentheses. For instance, if both groups increase by the same amount, their changes subtract to zero. This is an arithmetic property of the contrast; it does not require the groups to have equal outcome means at baseline.

What the Table Does and Does Not Establish

A 2-by-2 table supplies the four descriptive means needed for the contrast. It does not, by itself, explain the source of the differences or justify a causal interpretation. In a causal analysis, interpreting the DiD contrast as an effect requires additional design and identifying assumptions. Those assumptions are separate from the arithmetic and must be assessed for the study at hand. Thus, it is precise to report that the treated-group mean changed by a certain amount more or less than the control-group mean; a causal claim requires more than that calculation.

The comparison is also specific to the chosen outcome scale. A contrast computed from means in minutes is expressed in minutes; one computed from proportions is expressed in proportion units or percentage points. Changing the scale can change the numerical contrast and its interpretation. The four entries must refer to the same outcome definition and units for the subtraction to be meaningful.

Common Mistakes

• Subtracting post-period means only. The post-period treated-control difference is \(\bar Y_{T,1}-\bar Y_{C,1}\), not the DiD contrast. The DiD also accounts arithmetically for the pre-period difference by comparing changes.
• Reversing the order between groups. Calculate post minus pre for both groups, then treated change minus control change. Mixing pre minus post for one group reverses or corrupts the contrast.
• Dropping parentheses around the control change. Subtracting \((\bar Y_{C,1}-\bar Y_{C,0})\) changes the sign of both control terms when expanded.
• Calling the descriptive contrast a causal effect automatically. The four means determine the contrast, but a causal interpretation also requires identifying assumptions.
• Confusing percentage points with percent. A change from 12% to 18% is an increase of 6 percentage points. A difference-in-differences of 3 percentage points is not automatically a 3% relative change.