The Impact of Information Technology
on Metro Manufacturing Productivity
by
James A. Kurre, Ph.D.
8070 Gulf Road
North East, PA 16428
(814) 725-6888
June 30, 2007
A study for:
The Center for eBusiness and
Advanced Information Technology
Special thanks to research assistants Ben Schlosser and Alex Kazmierczak
for help in compiling the data for this and the underlying productivity project.
The Impact of Information Technology
on Metro Manufacturing Productivity
by
James A. Kurre, Ph. D.
EXECUTIVE SUMMARY
The purpose of this study is to determine the impact of information technology (IT) on manufacturing productivity. To do this, we calculated output per hour worked in manufacturing in 359 metro areas in 2002, and estimated the impact of several likely determinants, including education, scale of the place, capital investment, patents granted, and three skill levels of IT occupations.
The key conclusions are:
-Productivity in manufacturing varied significantly across metro areas in the U.S. in 2002. While the U.S. average for manufacturing productivity was $92.30 of value added per hour worked, the most productive metro area had productivity of $385.38 per hour, while the least productive area produced only $20.26 per hour. Erie’s productivity was $71.20, about 23% below the national average. The Appendices to the study give productivity values for all 359 metro areas. The key question of this study is what causes productivity to vary so much across metro areas.
-This study had full data on 272 American metro areas for 2002, and that was the data set used for the statistical analysis. This included 91% of the metro population of the country, and 75% of all American population. The 272 areas in our study included 90% of all metro manufacturing employees, and more than 71% of all American manufacturing employment.
-IT varied from 0.6% of all jobs in Victoria Texas to 14.1% in San Jose-Sunnyvale-Santa Clara CA, averaging 3.2% in the 272 metro areas of the study. Erie had 2.6% of its jobs in IT occupations in 2002. High-skill IT workers accounted for 1.7% of all jobs in the metro areas, and 0.9% in Erie.
-The presence of IT occupations clearly had a positive effect on manufacturing productivity. Metro areas that had a greater percentage of their workforce in IT occupations also tended to have higher productivity. This effect accounted for about 6% of the variation in productivity levels across metro areas.
-The presence of IT occupations in total had a positive impact, but the high-skill IT occupations are the ones that had most of the effect. Moderate skill IT occupations had some impact, but low skill IT occupations had virtually no impact on productivity.
-Capital investment per worker was the single most important determinant of productivity found in this study. Places with a greater amount of investment in capital equipment per worker tended to have higher manufacturing productivity. Alone, this variable accounted for nearly 9% of productivity variation across metro areas.
-Higher education also exerted a significant impact on productivity. A higher percentage of the population with bachelor’s or higher degrees resulted in higher productivity. This variable accounted for approximately 5% of the productivity variation.
-Places with higher amounts of innovation, measured by more patents granted per 100,000 residents, had greater productivity.
-The absolute scale of the place had little or no effect on manufacturing productivity. Bigger places, in terms of population, did not tend to have higher or lower levels of productivity. Manufacturing scale, on the other hand, had a slight impact, with metro areas that had a larger manufacturing sector in absolute size (i.e., amount of manufacturing employment) tending to have slightly higher levels of productivity. But this effect only accounted for 1%-2% of productivity variation across metro areas.
-IT occupations, education levels, and patents granted tend to be positively correlated across metro areas, meaning that where one of these variables is higher, the others tend to be, also. That makes it difficult to disentangle their effects and determine the impact of a single variable.
-The best models (combinations of variables) in this study explained about 15% of variation in manufacturing productivity across metro areas in the U.S. in 2002. Clearly, productivity is a complicated phenomenon that involves many factors. Nevertheless, this study found that IT occupations, capital investment, higher education, and patents granted all help increase manufacturing productivity in a metro area.
-Empirical work done by other researchers informs us that increased productivity need not mean decreased employment—jobs lost. If higher productivity results in lower costs of production and, through competition, lower prices for consumers, it is possible for the resulting increase in quantity demanded to more than offset the labor-demand reducing effect of productivity increases, and result in higher employment. And the lower costs that result from increased productivity means a higher standard of living for consumers generally.
1
The Impact of Information Technology
on Metro Manufacturing Productivity
I. INTRODUCTION
"The world economy is not a zero-sum game. Many nations can improve their prosperity if they can improve productivity. The central challenge in economic development, then, is how to create the conditions for rapid and sustained productivity growth."[1]
-Michael E. Porter, BishopWilliamLawrenceUniversity
Professor, HarvardBusinessSchool
You don’t have to be professor at the HarvardBusinessSchool to know that productivity (the amount of output produced for one hour’s work) is crucial. Even a little thought about how to be competitive, or how to increase living standards typically leads to the idea of increasing efficiency, of producing more output with less input.
The Organisation for Economic Co-Operation and Development (OECD) publishes rankings of countries based on labor productivity, and it is clear that productivity varies dramatically across countries of the world. Table 1 shows the results for 2004 in terms of GDP per hour worked.
In a similar vein, the World Economic Forum annually ranks about 125 nations of the world in terms of global competitiveness. The good news: the U.S. placed 6th out of 125 nations in 2006. The bad news: it was down from first place in 2005. Their rankings are available online.
We know that productivity varies within the U.S., too, although this is much less discussed, perhaps because the data are a bit harder to come by. Previous work by this author (Kurre, 2004) found that in 1997 labor productivity in manufacturing ranged from $14 per hour in BismarkND to $323 in Albuquerque, NM. The most productive metro area produced more than twenty times as much output from an hour’s labor as the least productive. That’s quite a range! Our own hometown, EriePA, ranked 241 out of 327 metro areas, with $58.62 of output produced per hour worked in 1997.
But what causes productivity to vary from place to place? What factors help explain the patterns we see? Previous work has identified some important determinants—but not all of them. This project will update the previous work, making use of Economic Census data that have become available since the earlier paper. But more importantly, it will explore the role that information technology plays in fostering productivity.
Table 1
OECD Rankings of Labor Productivity, 2004
Rank / Country / GDP per hour worked, US $ / Index relative to U.S.1 / Norway / 56.6 / 122
2 / Luxembourg / 55.9 / 121
3 / Belgium / 50.8 / 110
4 / France / 47.7 / 103
5 / Ireland / 47.1 / 102
6 / United States / 46.3 / 100
7 / Netherlands / 44.2 / 95
8 / Germany / 42.1 / 91
9 / Denmark / 40.9 / 88
10 / Sweden / 39.9 / 86
11 / United Kingdom / 39.6 / 86
12 / Finland / 39.2 / 85
13 / Austria / 38.4 / 83
14 / Switzerland / 36.7 / 79
15 / Spain / 36.5 / 79
16 / Italy / 36.3 / 78
17 / Canada / 35.2 / 76
18 / Australia / 34.7 / 75
19 / Iceland / 33.7 / 73
20 / Japan / 32.5 / 70
21 / Greece / 28.6 / 62
22 / New Zealand / 26.4 / 57
23 / Portugal / 23.9 / 52
24 / SlovakRepublic / 21.6 / 47
25 / Hungary / 21.5 / 46
26 / CzechRepublic / 20.7 / 45
27 / Korea / 18.6 / 40
28 / Poland / 17.7 / 38
29 / Mexico / 13.5 / 29
30 / Turkey / 12.7 / 28
Source: OECD, INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY LEVELS -
ESTIMATES FOR 2004, SEPTEMBER 2005. Paris: OECD. Available online at:
II. PRODUCTIVITY BASICS
First, we need to start with a definition of productivity. In the United States, the Bureau of Labor Statisticsis one key government agency that provides official measures of productivity, and their definition is:
“Productivity is a measure of economic efficiency which shows how effectively economic inputs are converted into output. Productivity is measured by comparing the amount of goods and services produced with the inputs which were used in production.”[2]
For this project, we elect to focus on a single input: labor. We measure productivity as value added per hour worked by production workers in manufacturing industries. We opt not to use a measure of the value of output, such as value of shipments, since that would involve double-counting of inputs within the manufacturing industry. A metro area that produces steel sheets, steel fabrication (turning the sheet into fenders), and automobiles would have a total value of shipments that double-counts the steel fabrication and triple-counts the steel itself. This would artificially inflate the value of shipments in that area compared to the true value produced—the final product, the car. Using value added at each step in the production process will avoid this problem.
And we focus on the productivity of labor rather than other factors of production. This is appropriate since labor costs account for the lion’s share of costs for most businesses. And it’s the source of most income for most Americans. Nationally, employee compensation accounted for 64% of national income in 2006, compared with 13.8% for corporate profits, 8.7% for proprietors’ income, 4.3% for net interest income, and 0.7% for rental income.[3] And employment is a key focus of government policy, both at the national and the local level.
But if increased employment is a goal of economic development, is greater productivity an ally or an enemy? Clearly, increased productivity can be seen as a two-edged sword. On the one hand higher productivity means getting more output from our resources, which in turn means lower costs of production so people can afford to buy more goods and services with a given amount of income. And higher labor productivity means that a worker is able to produce more output per hour, and is therefore worth more to his or her employer, which in turn can lead to higher pay. Both of these lead to a higher standard of living for the average worker/consumer.
On the other hand, if each worker can produce more output, an employer doesn’t need as much labor to produce a given level of output, so higher productivity can lead to lower employment levels. Jobs can be lost. This has been the story of the manufacturing industry in the U.S. over the last three decades.
However, Schweitzer and Zaman (2006) show that at the national level productivity growth in various industries typically tended to be more closely associated with output increases than with employment decreases over the period from 1990 to 2003. In fact, the service industries typically experienced productivity growth and employment increases, not decreases. And even in manufacturing, in the 2000-2003 period there was a slight positive correlation between productivity growth in a manufacturing industry and employment change; industries with higher productivity growth had increases in employment, not decreases. Some examples can illustrate the point. From 1990-2000, employment in the software publishing industry rose by 10% at the same time that productivity increased by 15.3%. And in wireless telecommunications, employment rose by 18% as productivity rose by 7.6%. (Schweitzer and Zaman, 2006, p.2 ) Clearly, productivity growth need not mean employment decline.
But how can this be? If employers need fewer workers to produce a given amount of output, won’t the demand for labor fall? Not necessarily. Since increased productivity will lead to lower labor costs, firms can produce their products more inexpensively. Competition among firms will ensure that this results in lower prices for goods and services, and as prices fall consumers will buy more of those products. The increase in quantity demanded from lower prices may be more than enough to offset the employment effects from productivity increases. Schweitzer and Zaman conclude that productivity growth does not necessarily mean decreased employment in an industry.[4] This means that a policy of encouraging productivity growth will not necessarily result in lost jobs.
Nordhaus (2005) at the National Bureau of Economic Research concurs with this view. In his study of labor productivity in the U.S. from 1948 to 2004, he concludes: “theresults here suggest that productivity is not to be feared….On thewhole, higher productivity has led to lower prices, expanding demand, andquickly to higher employment….” (2005, p. 18)
III. MEASURING METRO PRODUCTIVITY
Most studies of productivity have focused on the national or international level. Relatively few have explored the issue below the national level. There are some studies at the state level, such as Beeson (1989), Beeson and Husted (1989), Brock (2001), Domazlicky and Weber (1998), Hill (2001), Moomaw (1986), and Moomaw and Williams (1991), but very little at the level of the Metropolitan Area, which would be most useful for economic development and business location decisions. Any researcher coming to the field quickly discovers why: anapparent lack of data.
Previous work by this author explored different possible data sources, leading to the Economic Census as the best source. It provides data for all metro areas for a single point in time from a single data source, permitting the kind of cross-sectional study that we wish to do. The Census Bureau has a well-earned reputation for the quality of its data, and this is an important factor for any database. Along with productivity data, the Economic Census also provides data on capital stock additions for metro areas, which will be an important variable in the model.
Of course, the Economic Census is not a perfect database. We’d prefer to have data for as recent a period as possible, but the Economic Censuses are taken only every five years. The most recent data are for 2002. But what the Economic Census lacks in timeliness, it makes up for in data detail. Specifically, it is one of the few sources of geographically detailed data for industries that will let us perform analysis at the metropolitan level.
The year 2002 is the time point chosen for our study because it is the most recent year for which the Economic Census data are available. But does that bias our results? Where does 2002 fall in relation to the business cycle? Figure 1 shows data for real GDP from 1965 through the first quarter of 2007.[5] The National Bureau of Economic Research identifies July 1990 as a cyclical peak, followed by March 1991 as a trough. The next peak occurred in March of 2001 followed by a trough in November 2001.[6] This means that our 2002 data point covers the first year of the current expansion.
Figure 1
Real GDP in the U.S.
For this paper, we chose to measure productivity as value added in manufacture per hour of work by production workers. It would have been possible to calculate productivity as output per worker instead of per hour of labor. But that would be a less accurate measure since not all workers work full time. To the extent that some areas tend to use more part time workers, or to use overtime labor, their measures of productivity would vary. In fact, previous work demonstrates that productivity per worker and productivity per hour are highly correlated across metro areas. The correlation coefficient across 327 metro areas was 0.991 for 1997 data, and similar results obtained for both the “hours” and the “workers”measures in the regression analysis, suggesting that either approach is acceptable. We prefer the “hours” approach for the reason cited above.
Table 2 presentsnew productivity data calculated for this study, for the most and least productive metro areas in 2002 ranked by value added per hour worked. Appendices A and B present the productivity data for all 359 MAs in alphabetical order and ranked from most to least productive. The average for all 359 MSAs was $93.38 in 2002, the median was $85.20, and the average for the entire nation (not just metro areas) was $92.30. Our own Erie, PA produced $71.20 of value added for each hour of manufacturing work, nearly 24% below the mean, ranking it 249 of 359 U.S. metro areas . This table shows quite a broad range, from a high of $385.38 in Iowa CityIowa, to a low of only $20.26 in Jacksonville, North Carolina. Iowa City’s output per hour in manufacturing was 19 times that of Jacksonville’s in 2002. Of course, the key question is why productivity varies so much from place to place. And that’s the topic of the next section.
Table 2
Most and Least Productive MSAs, 2002
Ranked by Value Added per Hour in Manufacturing
Rank / MSA / VA per hour / Rank / MSA / VA per hour1 / Iowa CityIA / $385.38 / 340 / Houma-Bayou Cane-Thibodaux LA / $50.90
2 / MaconGA / 311.77 / 341 / Santa Fe NM / 50.54
3 / JanesvilleWI / 250.90 / 342 / Redding CA / 50.23
4 / RichmondVA / 233.36 / 343 / YakimaWA / 50.07
5 / LimaOH / 231.86 / 344 / Logan UT-ID / 49.07
6 / Raleigh-CaryNC / 224.76 / 345 / Hickory-Morganton-Lenoir NC / 48.90
7 / Baton RougeLA / 224.30 / 346 / Rapid CityND / 47.29
8 / Lake CharlesLA / 209.75 / 347 / Myrtle Beach-Conway-North Myrtle Beach SC / 46.72
9 / AlbanyGA / 206.39 / 348 / PrescottAZ / 45.77
10 / Austin-Round Rock TX / 202.98 / 349 / BangorME / 45.43
11 / Beaumont-Port Arthur TX / 192.98 / 350 / CorvallisOR / 45.23
12 / San Jose-Sunnyvale-Santa Clara CA / 192.96 / 351 / YumaAZ / 45.12
13 / Winston-SalemNC / 191.97 / 352 / College Station-Bryan TX / 45.04
14 / NapaCA / 190.45 / 353 / AmarilloTX / 44.73
15 / Hinesville-Fort Stewart GA / 182.53 / 354 / Coeur d'AleneID / 43.58
16 / Norwich-New London CT / 177.73 / 355 / St. George UT / 43.06
17 / Kalamazoo-Portage MI / 175.90 / 356 / DanvilleVA / 42.91
18 / RacineWI / 174.08 / 357 / GadsdenAL / 42.45
19 / Poughkeepsie-Newburgh-MiddletownNY / 172.95 / 358 / BristolVA / 39.24
20 / Bridgeport-Stamford-Norwalk CT / 171.29 / 359 / JacksonvilleNC / 20.26
IV. DETERMINANTS OF PRODUCTIVITY
Previous research and economic theory suggest a number of possible determinants of productivity.
1) Education
Education of the workforce might be expected to be a key determinant of the amount of output those workers produce. Beeson (1987), Beeson and Husted (1989), Moomaw and Williams (1991) and Brock (2001) all introduced education variables into their state-level studies of productivity differences.