Industry growth patterns: a closer look at output, productivity, and hours worked from 1990 to 2024

Where does economic growth come from? Major areas of economic research are devoted to studying how to increase it. Economic growth makes it easier to balance the national budget and gives prestige to a nation. It can also improve the average person’s standard of living.¹

The U.S. Bureau of Labor Statistics measures labor productivity as the ratio of output per hour of work.² Thus, one way to look at output growth is to break it down into two components:

• Hours of work
• Output per hour worked, known as productivity

Without growth in either hours worked or productivity, there is economic stagnation.³ As chart 1 demonstrates, the growth rate of output can be approximated as the sum of the growth rates of these two components. (For more information, see Appendix: Derivation of the output growth formula.)

$\left(1\right)\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}\mathrm{y}=\frac{\mathrm{O}\mathrm{u}\mathrm{t}\mathrm{p}\mathrm{u}\mathrm{t}}{\mathrm{H}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{s}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{k}\mathrm{e}\mathrm{d}}$

The end dates for the periods on the chart align with business cycles.⁴ (Data are reported for the nonfarm business sector.⁵) Overall, the chart displays a decelerating trend in the annual percent changes for output from 1990 to 2019, with a moderate resurgence, so far, after 2019.⁶

Productivity grew faster than hours worked in all periods. Hours growth flattened during the 2000–07 “jobless recovery,” after the “dot-com bust” recession. Productivity growth was responsible for nearly all output growth during this period. Finally, a weak but stable growth trend for hours worked appeared in the two periods from 2007–24. In the 2007–19 period, productivity grew roughly two and a half times as fast as hours worked. In the 2019–24 period, productivity then grew more than four times as fast as hours worked. However, productivity growth in the last two business cycles—so far—has remained a little slower than in the 1990–2007 periods.

The big picture tells only part of the story. Businesses throughout the U.S. economy produce goods and services through a variety of processes, using different combinations of labor, equipment, materials, and technologies. As a result, they will react differently to economic shocks. For example, consider how various workplaces reacted to the disruptions of the COVID-19 pandemic. Some types of workplaces remained fully or mostly staffed; some shut down. Some businesses were able to produce goods and services despite reduced workforces; some struggled.

Similarly, during economic booms, not all types of businesses grow in the same way—or at all. To understand the statistics for the U.S. economy, it is useful to analyze trends in the underlying industries. Different industries have different stories to tell.

In most industries, productivity was the main source of growth

The BLS Office of Productivity and Technology (OPT) publishes labor productivity measures for hundreds of U.S. industries.⁷ The North American Industrial Classification System (NAICS) is used throughout U.S. statistics to classify businesses into industries. NAICS further groups these industries into sectors. Figure 1 conveys the sector-industry hierarchy.

Figure. 1. Hierarchy of industries and sectors in the economy

$\left(2\right)\mathrm{O}\mathrm{u}\mathrm{t}\mathrm{p}\mathrm{u}\mathrm{t}=\mathrm{H}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{s}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{k}\mathrm{e}\mathrm{d}\times \mathrm{P}\mathrm{r}\mathrm{o}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}\mathrm{y}$

Chart 2 presents annual growth in output, hours worked, and productivity for seven sectors, across four business cycles, from 1990–2024.⁸ Output growth in all of these sectors, except accommodation and food services, resulted mainly, or entirely, from gains in productivity rather than hours worked.⁹ Only in accommodation and food services did hours worked grow faster than productivity (although data for the comparison only go through 2022).¹⁰ Throughout the article, this case will be called “hours-driven growth,” in contrast to the “productivity-driven growth” of the other sectors.

$\left(3\right)\dot{\frac{\mathrm{Y}}{\mathrm{Y}}}=\frac{\dot{\mathrm{H}\mathrm{W}}}{\mathrm{H}\mathrm{W}}+\frac{\dot{\mathrm{L}\mathrm{P}}}{\mathrm{L}\mathrm{P}}$

In industries, the source of growth varied. Since there are so many industries for which BLS publishes data, it is helpful to limit the pool of the analysis. Chart 3 displays the industries from each of the sectors in chart 2, with the largest increases in annual real sectoral output quantities between 1990 and 2024.¹¹ This narrows the list to industries that have contributed large amounts to national output growth over the last several decades. This means a lower bar of entry for some industries than others, consistent with the principle that sectoral output measures may understate the economic importance of goods-producing industries relative to some service-providing industries.¹²

$\left(4\right)\%∆Output\approx \%∆Hoursworked+\%∆Productivity$

Where did these expanding industries’ output growth come from?¹³ As shown in chart 2, the most frequent growth pattern shown in chart 3 was productivity-driven. Commercial equipment wholesalers was a good example of this: trivial gains in annual hours worked but very fast growth in productivity. For software publishers (unlike its parent sector, information), there is a pattern of roughly equal rates of growth in productivity and hours worked. Throughout this article, this will be referred to as “balanced growth.” Finally, the restaurants and other eating places industry (like the accommodation and food services sector), demonstrates a largely hours-driven growth pattern.

To go further into each of these growth patterns, this article will highlight these three industries. Given the unique stories of these industries, this article will apply different lenses to look at the data for each industry.

Productivity-driven growth: commercial equipment wholesalers

The commercial equipment industry (NAICS 4234) comprises merchant wholesalers of equipment and supplies for various commercial and professional purposes. Among others, these include photographic, office, information technology, medical, dental, and ophthalmic products.

Chart 4 shows that output grew at double-digit percentage rates during the 1990–2000 and 2000–07 business cycles, before slowing down to a more moderate pace of growth after 2007. Productivity, which grew 9.7 percent per year over the full 1990–2024 reference period, was the engine of almost all gains in output (10.0 percent annually). Even from 2007 to 2019, after a much sharper slowdown in annual productivity growth (to 4.3 percent), the industry’s output growth came entirely from productivity growth. Since 2019, productivity growth has been negligible, meaning that output has risen almost in line with the moderate increase in annual hours worked.

Some of the technological changes driving an acceleration in productivity growth from 1990–2000, especially after 1995, included the ongoing implementation of Electronic Data Interchanges and Universal Product Codes.¹⁴ The increasing use of these innovations helped enable the consolidation of wholesalers into larger firms, with increasing scale economies.¹⁵ These changes were all likely facilitated by an increasingly skilled workforce.¹⁶ The split by business cycles tells us that wholesale’s productivity growth from 2000–07 (3.3 percent annually) was not all that much lower than from 1990 to 2000 (4.1 percent). It seems reasonable to infer that technological developments underway in the 1990s had some lasting effects on the productivity of the sector into the 2000s. By 2007, perhaps, the productivity-enhancing benefits of various late 20th century wholesale industry transformations had already been largely absorbed. Productivity in the wholesale sector increased steadily, at 1.0 percent per year from 2007 to 2019 and 0.7 percent from 2019 to 2024.

A point to consider, in an industry with such diverse product lines as commercial equipment wholesaling, is whether shifts in the product mix might be reflected in the productivity statistics. In the years leading up to the 2007–09 Great Recession, most of the industry’s revenues came from the wholesaling of computers and peripheral equipment and software. (See chart 5.) However, after 2007, the share of sales in medical, dental, and hospital equipment and supplies has rapidly increased, while computers’ share of sales has fallen.¹⁷

BLS source data are not quite detailed enough to enable us to compare the productivity of product lines within this industry. However, it may be instructive to look at related industries. Table 1 indicates impressive productivity growth from 1990–2007 in computer and peripheral equipment manufacturing, which supplied many of the products sold by commercial equipment wholesalers. The rapid productivity growth in the production of computer-related products may be related to the same pattern of growth in commercial equipment wholesalers. Likewise, the sharp slowdown in productivity growth after 2007 appears in both industries.

The story of computer and peripheral equipment manufacturing seems especially relevant in the broader context of productivity trends during and after the “Information Revolution.” Trends in industries such as this one had ripple effects across the economy. Productivity at the national level boomed during the late 1990s and early 2000s period of “computing and internet-related innovations.”¹⁸ Productivity then decelerated, starting around 2005. Many researchers accept that industries which “produce information technology (IT) or that use IT intensively” were central to both the productivity speedup and the slowdown.¹⁹ These industries included computer and electronics manufacturing, the retail sector, and—most pertinent to this article—wholesale.²⁰

One final takeaway from table 1 is that the medical equipment and supplies manufacturing industry, which produced an increasingly large share of the products supplying commercial wholesalers, also had a productivity slowdown after 2007. In fact, it saw declining productivity from 2007 to 2024. As with the computer manufacturers, it is difficult to disentangle the effects of a supplier industry from the productivity trends of a wholesale industry “down the value chain.” Nonetheless, this relationship can be seen. The productivity trends of the commercial equipment wholesalers industry, both before and after 2007, are broadly consistent with both of its most related manufacturing industries and with the wholesale sector overall.

Balanced growth: software publishers

The software publishers industry (NAICS 511210) designs, develops, and publishes software. The two main categories of software products are “systems” and “applications.” Systems software helps your device run other programs; the operating system of your computer or smartphone is an example. Applications, or “apps,” are programs that a typical human user would access.²¹

Chart 6 reveals an astonishing growth rate of output from 1990 to 2000 (27.9 percent per year). While software publishers did not reach those heights of output growth in later periods, the annual rate over this article’s 1990–2024 reference period was still 12.8 percent.

The growth pattern of software publishers looks balanced between hours worked and productivity at first glance. Both productivity and hours worked grew fast between 1990 and 2024. However, the split by business cycles shows that productivity outpaced hours worked in the 1990–2000 and 2000–07 periods; hours worked even declined in the second period. Then, annual hours worked rose faster than productivity in the 2007–19 period, accelerating further from 2019 to 2024, while productivity growth continued to slow down. Between these two periods, the pattern shifted: output growth came mainly from productivity growth up to 2007, but more from growth in hours worked after 2007. This occurred even though the growth rate of hours worked from 1990 to 2000 exceeded all of the subsequent business cycles.

A few notable changes to the software industry’s products have occurred during this article’s 1990–2024 period: One is the transition from purchased software to software-as-a-service (SAAS). This transition occurred largely between the 1990s and 2000s.²² Another development to note is the transition to smartphones after 2007. The BLS source data do not break down software products by desktop versus mobile devices. Nonetheless, data from Statista suggest that smartphone sales increased at an accelerating rate after 2007, as new devices broadened the appeal of the mobile format.²³ Despite the proliferation of SAAS subscription models and smartphones, the industry has yet to see the growth rates of the 1990s again.

The software publishers industry also faced some structural disruptions. First, the dot-com bust of 2000, during which stocks of internet companies crashed, seems to have temporarily reversed the growing demand for software products. The relationship between the declining NASDAQ index and the “stifled demand for semiconductors and other IT-related goods” is well-known.²⁴ However, the crash can also be seen in the services side of IT. Chart 7 indicates an abrupt deceleration in annual domestic capital investment in software after 2000. Growth recovered somewhat starting in 2003, but not to the towering rates of the late 1990s.

Finally, there is one more example of the enduring slowdowns from the mid-to-late 2000s involving users and producers of IT. Table 2 reveals a trend of decelerating output growth in most information sector industries after 2007 (radio and television broadcasting was the exception).²⁵

Yet, among these industries with decelerating output growth, software publishers stands out for the fact that productivity growth sharply decelerated, while robust growth in hours worked continued after 2007. The other information industries saw their quantities of work hours contract after 2007, while most maintained, or even increased, their productivity growth rates. The only exception was wired telecommunications, which saw its productivity decelerate, though not as much as in software publishers.

Chart 8 presents two different paths of industry-labor utilization at recent period-endpoint years. In 2007, workers at software publishers contributed 9.0 percent of the 5.6 billion hours worked in the information sector. By 2024, this climbed to 23.0 percent of the sector’s 5.4 billion hours worked.²⁶ For comparison, the sectoral share of hours worked in wireless telecommunication carriers (except satellite) fell from 6.7 percent to 2.8 percent. This divergence in labor-hours trends stands in contrast to the output-growth trends, which were more similar between the industries. (See table 2.)

Hours-driven growth: restaurants and other eating places

The restaurants and other eating places industry (NAICS 72251) contains establishments serving predominantly food and/or nonalcoholic beverages, in all their varieties. Chart 9 shows that hours worked in the restaurants industry grew faster than productivity in all three completed business cycles shown (1990–2019), thereby accounting for most of output growth. However, between 2019 and 2024, productivity rose by 3.3 percent per year. Here, even as work hours dropped (-0.4 percent), output increased by 2.9 percent. What happened to flip the growth pattern after three decades of consistent hours-driven growth?

Fortunately, more industry detail is available in the data. The BLS productivity program breaks out the restaurants and other eating places industry into two more detailed industries: full-service restaurants (722511) and limited-service eating places (a combination of NAICS 722513, 4, 5).²⁷ As the industry titles suggest, the key difference is the type of service offered. If the customers are seated, look over a menu, and wait for a server to take their food or beverage order, then the establishment is probably full-service. Meanwhile, limited-service establishments may include fast-food, fast-casual, cafeterias, buffets, snack bars, and all other varieties of counter-service eateries and predominantly nonalcoholic drinking places.

Chart 10 breaks out the data for limited-service and full-service restaurants. The three completed business cycles from 1990 to 2019 look similar. Productivity grew slowly in both industries—a little more so in limited-service industries—as most of the growth in output came from increasing work hours.

The 2019–24 trends, for both industries, look different from the preceding three decades: a productivity speed-up paired with a deceleration in hours worked. While the pace of output growth seems to have fully recovered from the pandemic crash in both industries, the growth accounting patterns changed. The fluctuations in the trends were more extreme in full-service restaurants. That industry saw hours worked decline in 2019–24, rather than merely slow down, and a somewhat stronger rate of productivity growth, compared to the limited-service industry.

Charts 11 and 12 dissect the latest business cycle, by year, for each of the restaurant industries.

The disruptions of 2020 reduced hours worked and output in both limited-service eating places (chart 11) and full-service restaurants (chart 12), but more severely in full-service restaurants. The recovery in 2021 in full-service restaurants was also stronger, aided by rapid growth in productivity. After 2021, annual productivity growth was moderate in both industries. Hours worked recovered more quickly in full-service restaurants, though from a lower base. The increases in output were correspondingly larger in this industry than in limited-service restaurants.

The reasons for the discrepancies between the industries in the current business cycle are unclear. The pandemic seems to have hastened the adoption of newer business practices and technologies in both restaurant industries, which could have boosted measured productivity in either industry. A survey from a trade association indicated that the shift to off-premises revenues accelerated during 2020 in all restaurant types.²⁸ Later surveys confirmed that off-premises consumption in restaurants was still higher in 2021 than before the pandemic, concurrent with the drop in on-site dining.²⁹ Even well after most pandemic restrictions loosened, off-premises dining remained higher in 2023 than in 2019 throughout the restaurant trade.³⁰

Complementary to the increase in off-site consumption was the expanded use of digital technologies like ordering, payments, and delivery-management online or with mobile apps.³¹ Again, this does not seem to be merely a short-term reaction to pandemic conditions. Market survey data from 2023 suggest that many full-service restaurant operators introduced delivery service during the pandemic—including many fine-dining establishments—and that most of those who introduced it planned to retain it.³²

One study demonstrated how an increase in off-premises dining can show up in the productivity statistics—specifically, in the limited-service restaurants industry.³³ The researchers found evidence of a strong recovery in productivity based on analysis of proprietary microdata. They determined that productivity in limited-service restaurants rose “to a level some 15 percent higher than the pre-COVID steady state.”³⁴ The strong recovery in the industry was attributed to the fact that customers spent less time indoors, especially due to an increase in takeout orders. The authors observed that the restaurants that increased takeout orders the most were the ones that increased productivity the most.³⁵ The mechanism of increased productivity was a substitution of restaurant labor for home production. For example, takeout and delivery customers cleaned up their own tables and washed their own dishes.³⁶

If these changes in customer behavior led to a substantial increase in the productivity of limited-service restaurants, it may be interesting to learn what a similar study would show for full-service restaurants. The post-2019 increase in off-premises dining was arguably a more radical change for full-service restaurants, where table service is a core attribute of the industry.

Conclusion

Since 1990, output growth at the macroeconomic level has come mostly from increases in productivity, rather than hours worked. Output growth decelerated from 2000 to 2007, and again from 2007 to 2019, but accelerated somewhat after 2019. Hours worked showed the most significant deceleration from 2000–07 and modest growth over the past two business cycles.

At the more detailed level, the picture varies. Among large industries with large output gains over the last 30 years, some industries have grown due to rapid gains in productivity, despite flat or declining hours worked. While this pattern is especially common, the commercial equipment wholesalers industry is a good example. Other industries have seen balanced growth from increases in both productivity and hours worked. Software publishers fits this description. Finally, some industries have grown from expansions in hours worked despite slow-growing or declining productivity. The restaurants and other eating places industry typifies this pattern.

Yet, for each of these three industries, the 2019–24 growth pattern was contrary to what came before. Change does not always happen in the same direction across industries. Low-productivity-growth industries become high-productivity-growth industries, and vice versa, even within the same national business cycle.

In the meantime, workers may shift from low-productivity industries to high-productivity industries, and vice versa, wherever the supply and demand for their skills converge. But, it is important to remember that any aggregate frame for measuring the economy is just the sum of more detailed sectors, industries, and workplaces—and their various patterns of growth.

Appendix: Derivation of the output growth formula

Productivity, by definition, is the rate at which one hour of labor converts to a price-neutral unit of output, measured over time as an index. Stated in the way BLS measures it:

Implied in the definition is an equation of output as a multiple of hours worked and productivity:

.

Take the natural logarithms of each side and then differentiate with respect to time. With some substitution of terms and rearranging, equation (3) is created (shown here in dot notation), where Y is output, HW is hours worked, and LP is labor productivity:

.$\left(3\right)\dot{\frac{Y}{Y}}=\frac{\dot{\mathrm{H}\mathrm{W}}}{\mathrm{H}\mathrm{W}}+\frac{\dot{\mathrm{L}\mathrm{P}}}{\mathrm{L}\mathrm{P}}$

This calculus equation can be rewritten in more general terms as equation (4), an approximation of the growth rate of output as the sum of contributions of the growth rates of productivity and hours worked:

.

The additive equation (4) is more accurate for smaller percentage changes. However, the multiplicative equation (2), as an identity, is always true. If the quantity of hours worked decreases without an increase in productivity, then output will also decrease.