“The Death of a Technical Skill”, John J. Horton, Prasanna Tambe2020-10-17 (, ; backlinks; similar)⁠:

In 2010, Steve Jobs announced that [“Thoughts On Flash”, TOF] Apple would no longer support Adobe Flash—a popular set of tools for creating Internet applications. After the announcement, the use of Flash declined precipitously.

We show there was no reduction in Flash hourly wages because of a rapid supply response: Flash specialists, especially those who were younger, were less specialized, or had good “fall back” skills quickly transitioned away from Flash; new market entrants also avoided Flash, leaving the effective supply per Flash job opening unchanged.

As such, there was no factor market reason for firms to stay with Flash longer.

…Perhaps the best contemporary indicators of software developer interest in a given technology are the questions being asked on Stack Overflow, an enormously popular programming Q&A site. The left facet of Figure 1 shows the volume of questions per month for Flash and for comparison, a basket of popular IT skills, all normalized to 1 in the TOF month.3 The y-axis is on a log scale. We can see from this comparison that the numbers of questions about Flash and our chosen basket of skills are growing more or less in lockstep pre-TOF, reflecting growth in the Q&A platform and the wider IT industry, but that after TOF, Flash shows a clear absolute decline. There is some delay in this drop, likely reflecting the diffusion of the news of Apple’s plans as well as the completion of already-planned projects.

To study how the decline in Flash affected workers specializing in Flash, we use data from a large online labor market (Horton2010). The decline in Flash is also readily apparent in the longitudinal data from this market: the right facet of Figure 1 plots the number of job openings posted per month for jobs requiring Flash skills and for those requiring PHP (one of the “basket skills” from the left facet of Figure 1). Both Flash and PHP are normalized to 1 for the TOF-month. For comparison, we truncate the data to the first year of the Stack Overflow data (in 2008), even though the online labor market is considerably older. As we saw with the Stack Overflow data, both Flash and PHP move closely together pre-TOF and then diverge. Following TOF, the number of Flash job openings began to decline relative to PHP, falling by more than 80% 201052015.

As we will show, despite a large decline in the number of Flash openings posted, very little else about the market for Flash changed. There is no evidence employers were inundated with applications from out-of-work Flash programmers—the number of applicants per opening remained roughly constant.4 There was no increase in the likelihood that Flash openings were filled, nor was there a reduction in the wages paid to hired Flash programmers. In short, despite a roughly 80% reduction in posted Flash jobs, we observe a reduction in the quantity of hours-worked, but no reduction in the price.

…There is heterogeneity in the choices made by individual workers and their outcomes. Although there was no decline in wages on average, workers who were older seemed to have experienced wage declines, whereas younger workers experienced modest wage increases. Older workers also experienced declines in hours-worked that younger workers did not. We also observe that younger workers increasingly demanded a premium to work with Flash post-TOF, whereas older workers did not.