Manufacturing is a locus of innovation, yet standard industry data show manufacturing productivity stagnating in the US. This column argues that standard measurements understate true productivity growth in the sector, especially in industries where it is difficult to track changes in quality growth. Comparing producer and consumer price indexes, the authors show that ‘missing’ quality growth in standard industry output deflators is substantial in manufacturing and results in total factor productivity growth being understated. This is heavily driven by the Computer and Electronic Products industry, which is highly innovative and produces goods that have rapid turnover in product characteristics.
Despite its relatively small share of employment, manufacturing plays a central role in economic research and policymaking. The reasons are many. Manufacturing punches above its weight in innovation – roughly two-thirds of US patenting and corporate R&D expenditures emanate from the manufacturing sector (Autor et al. 2020). Each manufacturing plant supports many local jobs outside of manufacturing – local multipliers from manufacturing are large (Moretti 2010). The manufacturing sector has traditionally been a source of ‘good jobs’ for individuals without a college degree (Card et al. 2025). Furthermore, the manufacturing sector accounts for a large share of the aggregate productivity slowdown (Winkler et al. 2021). Partly in response to these factors, policymakers have traditionally and increasingly focused on boosting the scale and productivity of the manufacturing sector (Juhász et al. 2023).
Despite the high rates of measured innovation and the policy focus, US manufacturing productivity has stagnated over the last decade and a half. While total factor productivity (TFP) grew by 0.8% per year in the private economy between 2009 and 2023, for the manufacturing sector productivity growth has been negative, shrinking 0.1% annually (see Figure 1). This is in contrast to prior decades, where labour productivity and TFP growth were substantially faster in the manufacturing sector than elsewhere.
Figure 1 TFP growth in manufacturing and the private business sector
Data source: BLS Major Sector and Major Industry Total Factor Productivity dataset.
In a recent paper (Atalay et al. 2025), we study the recent slow productivity growth of the manufacturing sector. We argue that manufacturing productivity growth is understated in conventional statistics, primarily because deflators that are used to compute real output growth do not fully account for quality improvements that occur over time. This issue is pervasive throughout manufacturing, but is most salient in durable goods manufacturing, and particularly severe in the manufacturing of Computer and Electronic Products.
To begin, consider two main Bureau of Labor Statistics (BLS) price indexes: the Producer Price Index (PPI), which is a key input into the Bureau of Economic Analysis (BEA) gross output deflators, and the Consumer Price Index (CPI), which is a key input into the BEA’s Personal Consumption Expenditures (PCE) price index. For both price indexes, the BLS has the challenge of accounting for quality improvements. This is a tough problem that the BLS devotes considerable effort to. But there are reasons to believe the PPI may account for quality improvements less comprehensively than the CPI. For example, in product categories where the characteristics change a lot over time, the BLS’s preferred method to adjust for quality is hedonic adjustment, which essentially estimates the valuations of different product characteristics and then strips out the effect of changes in these characteristics on price growth. The BLS applies this technique in dozens of different product categories for the CPI, but only three for the PPI, with two of the three only incorporated in the 2010s. We hypothesise that when consumer and producer price indexes diverge for the same category, the gap often reflects under-adjustment for quality in the PPI.
Figure 2 provides intuition for our main comparisons. We consider a single consumption category, namely, televisions. For each year between 1997 and 2023, we plot PCE inflation in this category on the vertical axis. On the horizontal axis, we plot growth in the gross output deflator for the closest corresponding commodity, Audio and Video Equipment. In 2008, for example, quality-adjusted television prices fell by 19% according to the PCE price index, but by only 4% according to the gross output deflator. This large gap is typical throughout the 1997 to 2023 period. One thing missing from this discussion is that roughly half of domestic consumption is imported. Do price declines for imported Audio and Video Equipment make up for the difference? No. In the blue hollow circles, we plot inflation rates using the BLS import price index. These, too, are far from the 45-degree line. Just as with the gross output deflator, quality-adjusted price declines were much more modest in the import price index than in the PCE price index.
Figure 2 Television inflation
Notes: The vertical axis gives television inflation according to the PCE price index. The horizontal axis gives two measures of producer inflation. In orange filled circles, we plot changes in the gross output deflator for the Audio and Video Equipment Manufacturing industry (NAICS 3343). For this data series, we write out the corresponding year as well. The listed year, t, refers to the price growth between years t − 1 and t (e.g., the point for ’14 refers to price growth between 2013 and 2014). The import price index for Audio and Video Equipment Manufacturing is plotted using hollow blue circles without listing the year. This latter data series begins in 2005 only.
We construct a measure of ‘producer inflation’ that combines growth in domestic producer prices and import prices for the commodities in each consumption category. A data product from the BEA, the PCE Bridge Table, tells us the NAICS commodities associated with each detailed consumption category. Using the PCE Bridge Table, we weight NAICS commodities by their shares in each PCE category; using BEA Input-Output Tables, we compute the import share of each commodity to assign weights to imported versus domestically produced commodities.
In Figure 3, we plot PCE inflation now averaged over the 1997 to 2023 period versus our ‘producer inflation’ measure for a broad array of products. Producer inflation exceeds PCE inflation by 0.5 percentage points overall, with much larger gaps in nondurable goods (where the gap is 1.1 percentage points) and durable goods (a 2.6 percentage point gap) than in services (a gap that is close to zero). The largest gaps are in computers and other consumer electronics, including Other Video Equipment, Personal Computers, Telephones, and Televisions. These comparisons again indicate that quality improvements in goods, and especially high-tech products, are probably not fully reflected in the price indexes used to construct sector outputs.
Figure 3 Two measures of inflation across PCE categories, 1997–2023
Notes: Each point is a single PCE category. The vertical axis gives annual inflation according to the PCE price index between 1997 and 2023. The horizontal axis gives the corresponding-concept inflation measure using data from the gross output deflator and import price index. For the four data points in the bottom of the figure, the number preceding the colon is the NIPA line number.
In a final exercise, we consider the implications of these price growth gaps for productivity growth. Essentially, we ask what TFP growth would have looked like if official statistics used consumer rather than producer prices to deflate nominal variables. We use the BEA Input-Output Table to adjust measured productivity for potential mismeasurement in both output and intermediate input prices.
Figure 4 displays the results from this exercise. Manufacturing productivity growth is much greater throughout the sample: 1.4% annual TFP growth when corrected for mismeasurement versus 0.6% annual growth in the official data. TFP growth continues even past 2009 (0.6% annually), albeit at a slower rate than before (2.3% annually, between 1997 and 2009). The more than decade-long stagnation in manufacturing productivity therefore appears to be a price measurement issue. Importantly however, the similarly timed slowdown in manufacturing productivity growth remains, as the price measurement correction affects the sector’s productivity growth similarly both before and after the slowdown began. Our corrections have the largest impact on TFP growth rates for durable manufacturing (2.7% versus 1.0% in the official statistics), particularly so for the Computer and Electronic Products manufacturing industry (9.8% versus 4.0%), for which quality adjustments are likely to matter most. By contrast, productivity growth outside of manufacturing changes little when we account for mismeasurement.
Figure 4 Measured and corrected TFP growth
Notes: We plot annual TFP growth for five groups of industries: the entire manufacturing sector, the nondurable manufacturing sector, the durable manufacturing sector, Computer and Electronic Products Manufacturing, and the private nonmanufacturing sector. Official statistics come from the BLS Major Sector and Major Industry Total Factor Productivity dataset. The bars labelled “Corrected for Mismeasurement” add in our estimates of TFP mismeasurement in each group of industries.
Source : VOXeu
