When Similarity Becomes a Demand Signal

The U.S. evidence shows that founder and employee composition remain related as firms age, but composition alone cannot say why. This chapter examines the next question: do workers sort toward founders who resemble them, do founders demand those workers more strongly, or do both sides respond to an organization-specific production process? The wage decomposition and model in Who Hires Whom? provide a disciplined way to move from a persistent pattern to a possible mechanism—while keeping a possible mechanism separate from a proven motive or causal effect.

Wages as a test of the mechanism

A simple hiring count cannot tell us whether a group is concentrated in a firm because the firm wants those workers, because those workers prefer the firm, or because other opportunities are less available. Wages provide another signal. If workers from a founder’s background receive a higher firm-specific wage premium than comparable workers receive elsewhere, that is more consistent with relative demand from the firm than with a story in which those workers merely accept lower compensation to work for a similar founder.

The U.S. authors use a two-step decomposition related to the AKM approach in labor economics. The first step separates observed wages into an individual component and a firm component. The individual component captures persistent differences associated with the worker; the firm component captures the wage level associated with the workplace after accounting for age, age squared, and tenure. The researchers estimate these firm effects separately for a demographic group and for its complement. The second step compares the two firm effects within a workplace and relates that difference to the share of founders from the group.

In plain language, the procedure asks whether a firm associated with more female founders, for example, pays female workers more relative to their comparison group than it pays at other firms, after the decomposition has accounted for worker and workplace differences. The same exercise is performed for Black, Hispanic, Asian, and college-educated workers. It is not a direct survey of what a founder values. It is a comparison of the wage component associated with the firm.

The estimated slopes are positive across the groups: about 0.129 for female founders, 0.068 for Black founders, 0.057 for Hispanic founders, 0.029 for Asian founders, and 0.049 for college-educated founders. The authors interpret these positive differentials as evidence that relative labor demand dominates relative labor supply in explaining the employee-composition patterns. If workers who resemble founders were present mainly because they accepted a relative wage penalty in exchange for a preferred social environment, the corresponding firm-effect differential would be expected to point in the opposite direction. The positive estimates instead indicate that similar-background workers receive higher firm-specific pay relative to their outside pattern.

That is a meaningful strengthening of the earlier composition result. A firm can have more workers from a group for many reasons. Evidence that the firm-specific wage component is also relatively higher for that group makes a pure worker-supply explanation less plausible. It is still a conditional interpretation of a decomposition, not a complete causal proof.

The sign of the comparison is doing the conceptual work. Imagine two workers with similar estimated individual effects, one from the focal group and one from the comparison group.