Labor Composition and Measured Productivity

Why the Mix of Workers Matters

Aggregate output per hour can change for two distinct reasons. Workers in a given education, experience, or occupation group may produce more or less per hour than before. Alternatively, the shares of total hours supplied by those groups may change. If a larger share of hours is supplied by groups with higher output per hour, the aggregate measure can rise even if no group’s own rate changes. Conversely, a shift toward groups with lower measured output per hour can offset gains within groups.

This distinction matters when interpreting productivity trends. A rise in the aggregate does not by itself show that every group became more productive, and a fall does not prove that productivity declined within every group. The first task is to separate the two sources of change arithmetically.

Here, \(p_{it}\) is a group-specific rate and \(s_{it}\) is a weight based on hours, not the group’s share of the number of workers. A part-time worker and a full-time worker do not contribute equal hours merely because each counts as one person. The weighted-average formula follows because total output is \(\sum_i Q_{it}=\sum_i H_{it}p_{it}\); dividing by total hours gives \(\sum_i (H_{it}/\sum_j H_{jt})p_{it}\).

The groups must partition the hours being analyzed, and the output measure and period must be comparable across groups and dates. In aggregate productivity work, output should be measured consistently in real terms rather than mixing nominal values that reflect price changes. The formula is an accounting identity under these conditions; it does not say that every group’s output can be observed independently in every real-world production process.

Two Sources of Change

A within-group effect captures changes in the output-per-hour rates \(p_i\), holding the group weights at a specified average. A composition effect captures changes in the hours shares \(s_i\), holding group rates at a specified average. For a two-period comparison, define the midpoint of each variable as the arithmetic average of its period-0 and period-1 values:

The midpoint, or symmetric, decomposition divides the aggregate change exactly into these two effects:

The decomposition is an accounting attribution, not a causal model. For example, if the share of hours supplied by workers with more education increases, the composition term can be positive when that group has higher output per hour. That arithmetic alone does not establish that education caused the group’s higher rate, or that a change in education caused a particular productivity outcome. The earlier material on human capital from education, skills, and experience concerns characteristics that may matter for productivity; this decomposition addresses how group rates and group weights combine in an aggregate.

Why the Midpoint Formula Is Exact

The key is to separate the change in each group’s weighted contribution without leaving an interaction term behind. For any one group, the difference between its period contributions can be written as:

To verify the identity, expand its right-hand side. The first term is \((s_{i0}p_{i1}+s_{i1}p_{i1}-s_{i0}p_{i0}-s_{i1}p_{i0})/2\). The second term is \((p_{i0}s_{i1}+p_{i1}s_{i1}-p_{i0}s_{i0}-p_{i1}s_{i0})/2\). The cross-products \(s_{i1}p_{i0}\) and \(s_{i0}p_{i1}\) cancel across the two expressions, leaving \(s_{i1}p_{i1}-s_{i0}p_{i0}\). Summing over groups gives the aggregate identity because \(P_t=\sum_i s_{it}p_{it}\).

The midpoint method shares the interaction between changing rates and changing shares symmetrically. A base-period decomposition can also be constructed, but if both rates and shares change, omitting the interaction generally makes a two-part calculation inexact. The midpoint formula avoids that omitted residual.

Reading the Effects Correctly

A composition effect is positive when hours shift, on balance, toward groups with higher average output per hour, and negative when they shift toward groups with lower average output per hour. The relevant comparison is not simply whether a group’s share rose: the group’s average rate matters, as do the changes in the other groups’ shares. Because shares sum to one in each period, their changes sum to zero. A gain in one group’s share must be matched by a decline elsewhere.

The same method applies whether groups are defined by education, experience, or occupation, provided the categories are clearly specified and consistently measured. With more than two groups, calculate a contribution for each group and sum them. Small rounding differences may occur if the supplied shares or rates have already been rounded, but the unrounded decomposition should satisfy the identity.

Group output per hour can also reflect differences in tasks, capital, work organization, or the selection of workers into groups. It is not interchangeable with a wage rate. Wages may be affected by labor-market institutions, bargaining, and other factors, so substituting wage differences for measured output-per-hour differences requires additional assumptions. For this tutorial’s accounting exercise, use the specified group output rates and hours shares.

Practical Calculation Sequence

When a problem supplies group-specific rates and hours shares for two periods, keep the aggregate calculation and the decomposition distinct. First verify that the shares sum to one in each period; then compute both weighted aggregates. Next calculate the midpoint quantities and contributions. This order makes it easier to identify whether a discrepancy arises from a weighted-average arithmetic error or from a mistaken decomposition.

This accounting distinction is useful for assessing productivity changes without attributing every movement in the aggregate to a change in group-level performance. In applied work, it clarifies whether a measured change is associated with rates within worker groups, the distribution of hours across groups, or both. It should be presented as a decomposition of observed values, not as proof of the cause of those values. The treatment here is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.