When a company undergoes a technological change — such as introducing robots or reorganizing the workflow in manufacturing — does it mainly make individual workers more productive because they will be able to churn out more widgets per hour? Or does it make all factors of production equally more productive — workers can produce more widgets, state-of-the-art equipment can accomplish tasks faster, materials that can be sourced more efficiently are used, among others?

Macroeconomists assume that technological change is biased and increases labor productivity alone, while the literature on productivity assumes that it increases the productivity of all factors equally, or that technological change is Hicks neutral. A new paper, “Measuring the Bias of Technological Change,” from Ulrich Doraszelski, Wharton professor of business economics and public policy, and Jordi Jaumandreu, senior academic researcher at Boston University, shows that the truth is somewhere in between. “We are the first to provide numbers to back up this claim,” Doraszelski says. “Our paper has the potential to change how economists think about technological change and how they account for it in their models.”

The findings of the paper are important. “Biased technological change is a really big issue,” Doraszelski says. “It changes how the firm does things. If technological change is entirely Hicks neutral and the productivity of all factors of production goes up in lockstep, then it won’t change the ratio of capital to labor.” But if it is shown to be biased — that is, enhances the productivity of some inputs into the production process over others — then the company’s strategy will shift accordingly. A so-called labor-augmenting technological change means fewer workers are needed to produce the same output. The results buttress a longstanding concern of laborers that they will be displaced by new technology. “Ever since the Luddites [English textile workers] of the early 19th century, many people fear that this labor-augmenting technological change is to the disadvantage of workers,” Doraszelski adds.

But Doraszelski says this does not always have to be the case because companies can take advantage of lower labor costs to expand, resulting in the hiring of more workers. “Firms have an incentive to shed workers, but that’s not the end of the story. Because the workers are now more productive, the cost of the firm falls, so the firm expands output and uses more input,” Doraszelski says. “We’re trying to see which of these two factors is bigger: Does the firm want to shed labor, or expand output and add labor?”

“The truth is somewhere in between, and we are the first to provide numbers to back up this claim.”–Ulrich Doraszelski

Doraszelski adds that his research also can change the way people view productivity. “Historically, people have looked at productivity as a single number. What we’re showing is that it’s not. Technological change has multiple dimensions,” he says. “It changes the perspective people have on productivity. That means we need to rethink how we measure productivity and account for the fact that it’s multi-dimensional.”

Different Dimensions of Productivity

Doraszelski and Jaumandreu were able to tease out the different dimensions of productivity gains because they used panel data at the individual company level, and combined these data with advanced econometric techniques. “It’s the first firm-level study of biased technological change,” Doraszelski says. Prior research has always estimated technological change using highly aggregated data, such as the entire U.S. manufacturing sector. “At best, this may give us an average measure of technological change,” he adds. In contrast, “our approach is bottom up,” Doraszelski says. “We start from firm-level data and then estimate technological change at the level of the individual firm.… We can aggregate to the level of the industry, but we can also see what is behind this average.”

The authors used 1990 to 2006 data from 2,375 individual manufacturing firms spanning 10 industries in Spain that was culled from the “Encuesta Sobre Estrategias Empresariales,” a Spanish Ministry of Industry survey. The timeframe covered by the data is a period of rapid output growth accompanied by stagnant or slightly increasing employment but “intense” investment in physical capital, the paper said. The authors chose Spain because it is an industrialized economy that began fully integrating into the European Union from the late 1980s to early 1990s. “Any trends in technological change that our analysis uncovers for Spain may thus be viewed as broadly representative for other continental European economies,” the paper said.

“Our paper has the potential to change how economists think about technological change and how they account for it in their models.”–Ulrich Doraszelski

Doraszelski and Jaumandreu estimate that labor-augmenting technological change leads to increases in output by around 2% on average a year. For example, using advanced equipment on the factory floor can make workers produce on average 2% more widgets. This increase in labor productivity is not mainly due to the shift from unskilled to skilled workers, either. Rather, “in many industries, labor productivity grows because workers with a given set of skills become more productive over time,” the authors said.

So does this vindicate the view of macroeconomists that technological change increases labor productivity alone? Not entirely. Output also grows because all factors of production become more productive, not just because labor becomes more productive. Indeed, analysis of the data shows that because of Hicks-neutral technological change, output grew on average by additional 2% yearly.

“If you hold everything fixed, the industry will thus produce 4% more output growth because of technological change,” Doraszelski says. “It’s huge … If you think about it, you don’t change anything in the economy and output is going to grow by 4% a year using the same amount of input.” The figure is significant within the context of U.S. economic growth. Consider that in the second quarter of 2015, the U.S. gross domestic product grew by a seasonally adjusted 2.3%, according to the U.S. Commerce Department. That was up from a revised 0.6% in the first quarter.

“We need to rethink how we measure productivity and account for the fact that it’s multi-dimensional.”–Ulrich Doraszelski

Going forward, Doraszelski says future research will consider the employment implications of their findings. Firms can choose to either cut the workforce or hire more employees because labor costs have dropped. “The question is how these two forces balance,” he says.

The paper also shows that R&D efforts play a key role in determining the differences in labor-augmenting productivity across firms and how this component of productivity changes over time. “Technological change is not exogenous in the sense that a firm doesn’t wake up one morning to find itself more productive. At least to some extent, firms bring about technological change themselves through their R&D efforts, acquiring intellectual property, improving their internal organization,” among others, Doraszelski adds. “Now that we can disentangle different types of technological change in the data, the next step is to ask which activities of the firm are related to which type of technological change, and how these activities respond to economic incentives.”