Economists and sociologists have long been interested in studying the technologies involved in production. Going back to the seminal works by Ryan and Gross (1943) and Griliches (1957) on the diffusion of hybrid varieties of corn, the dominant approach to measuring technology has reflected whether a potential adopter uses some new or advanced technology. In addition, when studying the pace and barriers of technology diffusion, the main interests of researchers have focused on studying the effect of technology on productivity (e.g. Bartel et al. 2007, Juhasz et al. 2020), wages (Krueger 1993), and skills and job polarisation (Acemoglu and Autor 2011). This has motivated researchers to measure the use of (typically a few) advanced technologies by firms in numerous sectors. The most common examples are firm-level measures of information and communication technology (ICT) (e.g. the US Census Bureau, Information & Communication Technology Survey, Annual Business Survey, the Eurostat, and Community Survey of ICT Usage).
Despite all the progress, existing measures of technology still fall short of providing a comprehensive characterisation of technology within firms. From a technological standpoint, firms largely remain ‘black boxes’. First, the number of technologies typically covered is rather limited when compared to how many technologies are involved in the management and production processes of a firm. Second, because the focus of most surveys is on the use of particular advanced technologies, it is impossible to understand how production takes place in companies that do not use such advanced technologies (i.e. de facto most firms in developing countries). Third, since the unit of analysis is the firm, existing studies are not designed to study what business functions benefit from each technology. Finally, existing surveys do not reveal whether a technology is widely utilised or just marginally used.
The Firm-level Adoption of Technology (FAT) survey
To overcome these limitations, we have developed a new approach to measuring technology adoption that shifts the unit of analysis from the firm to the business-function level – the Firm-level Adoption of Technology (FAT) survey (Cirera et al. 2020). The survey was designed with the assistance of a large number of sector and technology experts who helped identify the key business functions, the technologies that companies can use to conduct the main tasks in each of the selected business functions, and the ranking of technologies according to their sophistication. The survey covers seven general business functions (GBFs) that are common to all companies regardless of the sector in which they operate. In addition, for ten large sectors, we have identified their key sector-specific business functions (SSBFs) and the main technologies that can be used to implement them. Figure 1 shows the general business functions identified and the technologies that can be used, from the most basic to the most sophisticated. Similarly, Figure 2 unpacks sector-specific production activities of the food processing industry into the key business functions and the technologies that can be used to accomplish them. The other nine sectors where we collect information on sector-specific technologies are agriculture, livestock, wearing apparel, automotive, pharmacy, wholesale and retail, land transportation, banking, and health services. In total, the FAT survey covers 59 business functions – general and sector-specific business functions – and 287 technologies associated with them.
Figure 1 General business functions and their technologies
Figure 2 Sector-specific business functions and technologies in food processing
The first round of the survey was implemented to a representative sample of firms in Senegal, Vietnam, and the Brazilian state of Ceará. For each country, the sampling frame is based on the most comprehensive and updated establishment census data available from the respective National Statistical Offices (NSOs) or similar administrative information. We collected data for 3,996 establishments, including 711 establishments in the State of Ceará in Brazil, 1,786 establishments in Senegal, and 1,499 establishments in Vietnam.
A new technology sophistication measure
The survey asks firms to first list all the technologies that are used in each business function, and then specify which one is the most widely used. With this information, we construct business function level measures of the sophistication of the most widely used technology and of the full array of technologies used in the business function. We illustrate how these measures of sophistication can characterise the technological landscape of firms by conducting a case study of two individual firms, with emphasis in two indices: (i) MOST, which reflects the sophistication of the most widely used technology in a business function; and (ii) EXT that measures the sophistication of the array of technologies used to conduct a business function.
Figure 3 presents, in four spider charts, the measures of EXT (right) and MOST (left) for each of the general (top) and sector-specific business functions (bottom) for two food processing firms (large in dashed blue, small in solid red). In general, the large firm uses more sophisticated technologies than the small one. However, the gap between the sophistication of technologies used in both firms varies considerably depending on the technology measure, the type of business function, and the specific business function we consider.
Figure 3 Example of two firms in food processing in Senegal
Note: Two firms in food processing in Senegal are selected to provide an example of the technology indices. The sizes of Firm A and B are nine and 100 employees, respectively.
New technology facts: Large within-firm technology variance and heterogeneity in ‘technology curves’
Our analysis reveals a large variation in the sophistication of technologies used in production at all levels of aggregation, but the variance in sophistication increases the more disaggregated they are. Specifically, we find greater variance in technology sophistication across the business functions of a firm (i.e. within-firm) than across firms, and greater variance across firms than across countries/regions. Both at the national and regional level, we document a positive relationship between cross-firm variance in technology sophistication and development. This large within-firm variance in technological sophistication debunks the notion that technology is uniform within firms, and that firms use either sophisticated or rudimentary technologies in all their functions. The within-firm variance in technology sophistication increases with the average level of sophistication in the firm and with firm size. These patterns suggest that heterogeneity in technology across business functions may be driven by heterogeneity in the value of more sophisticated technologies rather than heterogeneity in the costs of technology adoption.
To further explore this hypothesis, we study how the sophistication of technology in a business function varies with average firm-level sophistication. For each individual business function, we document stable cross-firm relationships between sophistication at the business function level and average firm sophistication. We name this relationship the ‘technology curve’. Technology curves account for a large share of the cross-firm variation in the sophistication of technology at the business function level, with an average R2 of 0.39. More importantly, there is large heterogeneity across business functions in the slope of technology curves. For general business functions (Figure 4, Panel A), the functions with steepest technology curves are business administration and planning, while the sales function has the flattest curve. For food processing (Figure 4, Panel B) the steepest technology curves are observed for packaging and food storage. These findings suggest that a critical driver of within-firm differences in technology are non-homotheticities in production.
Figure 4 The technology curves
Panel A: General business functions
Panel B: Sector-specific: Food processing
Technology and productivity
Our firm-level average sophistication measures are positively correlated with productivity. But perhaps a more surprising result is that productivity is also positively correlated with the within-firm variance of technology sophistication across business functions after controlling for average firm-level sophistication. This could be interpreted as evidence of the value of devoting more sophisticated technologies in more relevant business functions.
We conclude our analysis by conducting two development accounting exercises that shed light on two classical debates: (i) the drivers of cross-country productivity differences and (ii) the much larger cross-country productivity dispersion observed in agriculture than in non-agricultural sectors. We estimate that cross-firm differences in technology account for a third of the gap that exists between firms at the top 90% and bottom 10% of the productivity distribution. Cross-country differences in the technology of the average company in agriculture and in non-agricultural activities account for one fifth of the ratio between the cross-country gap in productivity in agricultural firms over the cross-country gap in productivity in non-agricultural firms.
Acemoglu, D and D Autor (2011), “Chapter 12 – Skills, Tasks and Technologies: Implications for Employment and Earnings”, in Handbook of Labor Economics, Vol. 4: 1043-1171.
Bartel, A, C Ichniowski and K Shaw (2007), “How does Information Technology Affect Productivity? Plant-level Comparisons of Product Innovation, Process Improvement, and Worker Skills”, Quarterly Journal of Economics 122(4): 1721-1758.
Cirera, X, D Comin, M Cruz and K M Lee (2020), “Technology Within and Across Firms”, CEPR Discussion Paper 15427.
Griliches, Z (1957), “Hybrid Corn: An Exploration in the Economics of Technological Change”, Econometrica 25(4): 501-522.
Juhasz, R, M P Squicciarini and N Voigtlander (2020), “Technology Adoption and Productivity Growth: Evidence from Industrialization in France”, NBER Working Paper 27503.
Krueger, A B (1993), “How Computers Have Changed the Wage Structure: Evidence from Microdata, 1984-1989”, Quarterly Journal of Economics 108(1): 33-60.
Ryan, B and N Gross (1943), “The Diffusion of Hybrid Seed Corn in Two Iowa Communities”, Rural Sociology 8(1): 15-24.