Did the Rollout of Television Reduce Labor Supply by Increasing the Value of Leisure?

This study asks whether the introduction of television reduced labor supply by increasing the value of leisure. Using Social Security work histories linked to the 1978 CPS and leveraging natural variation from the staggered rollout of U.S. television stations between 1948 and 1960, the authors estimate labor supply effects. They find that each additional TV station reduced the probability of working by approximately 0.3 percentage points, primarily due to increased retirement among older workers. Younger workers showed no significant changes. The findings suggest that TV contributed to earlier retirements and modestly shifted labor-leisure decisions, especially at the retirement margin.

Full Citation and Link to Article

Fenton, George, and Felix Koenig. “Labor Supply and Entertainment Innovations: Evidence From the U.S. TV Rollout.” American Economic Journal: Applied Economics (Forthcoming, 2025). Available at: https://www.aeaweb.org/articles?id=10.1257/app.20230377 

Extended Summary

Central Research Question

This article investigates whether innovations in entertainment technology—in particular, the rollout of television—reduced labor supply in the mid-twentieth century United States. The authors ask: did the introduction of television, by increasing the attractiveness of leisure, lead Americans, especially older workers, to exit the labor force earlier? The paper situates this question within a broader inquiry into how changes in the value of non-work activities can influence labor-leisure trade-offs and overall employment behavior.

Previous Literature

Prior research has largely emphasized the role of wages in determining labor supply, with much less focus on how rising returns to leisure affect work decisions. Seminal work by Gary Becker (1965) introduced the idea that time allocation reflects trade-offs between market work and non-market activities. Later studies, such as Aguiar et al. (2021), demonstrated that video games increased the value of leisure and reduced labor supply among young men. Costa (1998) posited that a broader set of affordable leisure options, including television, helped transform retirement from a necessity into a lifestyle choice. Although these studies suggest a possible connection between entertainment and work behavior, they rely on limited causal evidence. Abraham and Kearney (2020), for example, noted a lack of credible estimates of the impact of leisure technology on labor supply. The present study contributes clean identification by exploiting the staggered and partially randomized introduction of TV across the U.S.

Data

The study combines two main datasets. First, the authors use the Current Population Survey–Social Security Earnings Records Exact Match file (SSA-CPS), which links respondents from the 1978 March CPS to their full Social Security work histories from the 1930s to 1960. This panel data allows precise tracking of labor market participation over time for nearly 300,000 individuals, with particular attention to workers aged 50 and older.

Second, the authors construct a novel dataset on television access during the initial rollout (1948–1960). They digitize technical data from historical Television Factbook volumes and apply the Irregular Terrain Model (ITM) to compute TV signal strength for each U.S. county. This enables precise measurement of both the presence and strength of broadcast signals over time, allowing for intensity-based treatment variables (i.e., number of stations received). They also incorporate data on Community Antenna Television (CATV) and signal translators to capture indirect access.

A key component of their identification strategy is based on a regulatory freeze by the Federal Communications Commission (FCC) from 1948 to 1952, which delayed the launch of numerous stations. These “ghost stations,” approved but unable to broadcast, provide a quasi-experimental control group for causal inference.

Methods

The core empirical strategy is a difference-in-differences (DiD) design that compares employment outcomes before and after the introduction of television across geographic areas with different rollout timings. The authors estimate the effect of each additional station on the probability of working, controlling for individual fixed effects, gender-specific year effects, and demographic variables.

The main regression takes the form:

ΔEaigt = γgt + δi + βk·TVat + π·Xaigt + εaigt

Where:

• Eaigt is an employment indicator (1 if working, 0 otherwise)

• TVat is the number of TV stations available in area a at time t

• γgt are gender-specific year fixed effects

• δi are individual fixed effects

• Xaigt are time-varying demographic controls

They supplement the main DiD analysis with an event-study framework to visualize dynamic effects, placebo tests using frozen “ghost” stations, and robustness checks including region-specific trends and alternative functional forms (e.g., log TV counts, quadratic terms). To address potential bias from migration (because CPS location is from 1978), they test for correlations between migration and rollout timing and apply bounding exercises to assess measurement error. They also use a smaller dataset from the Current Employment Statistics (CES) to examine effects on hours worked.

Findings/Size Effects

The paper finds that television rollout led to modest but statistically significant reductions in labor supply, concentrated among older individuals. Specifically:

• Each additional television station reduced the probability of working by approximately 0.3 percentage points for individuals aged 50 and older.

• Younger workers (under 50) showed no significant employment changes.

• Retirement rates for older workers rose by about 0.25 percentage points per station, equivalent to retiring about two months earlier.

• The most pronounced effects came from the first few station launches, with diminishing marginal impacts from subsequent stations.

• Hours worked (intensive margin) did not change significantly, consistent with historical rigidity in work schedules; instead, the response occurred entirely at the extensive margin (whether to work or not).

• Women responded similarly to men, suggesting that retirement decisions may have been jointly made at the household level.

Placebo tests using blocked “ghost stations” showed no effect on labor supply, lending credibility to the causal interpretation. Moreover, rollout timing was uncorrelated with most demographic trends, supporting the parallel trends assumption of DiD models. Alternative identification strategies using regulatory freeze-affected locations yielded consistent estimates.

In a final step, the authors develop a simple representative agent model to quantify how improvements in leisure value shift labor supply. They find that most of the adjustment to TV came from substitution among leisure activities (e.g., radio or reading), rather than from reductions in work. Based on time-use data, they estimate that TV increased leisure time among older adults by 2.4% to 3.8%, with entertainment activities’ share of leisure rising from 50% to 89%. Using these values, they estimate a high elasticity of substitution among leisure activities (εcii ≈ 11–19), much higher than values assumed in prior literature (e.g., 1.6 in Aguiar et al. 2021). This suggests that past models overstated the labor-supply impact of entertainment technologies.

Conclusion

This paper offers compelling evidence that television—perhaps the most significant entertainment innovation of the 20th century—modestly reduced labor supply by increasing the value of leisure. The effects were concentrated among older individuals on the margin of retirement, reinforcing the narrative that post-war retirement became a lifestyle choice enabled by compelling, low-cost leisure options. While the magnitude of TV’s effects on employment was limited, the study provides a well-identified estimate and helps refine theoretical models of labor supply by grounding them in credible empirical evidence.

Importantly, the authors find that substitution among leisure activities plays a dominant role in how people respond to new technologies. Thus, while new forms of entertainment may capture time from other non-work activities, their impact on total labor supply is likely to be modest unless the innovation is both large and uniquely compelling. This insight carries implications for future research and policymaking, particularly as society continues to grapple with rapid technological change and its effects on work behavior.