Tag Archives: Joseph Schumpeter

Innovation and economic growth in the early 20th century and lessons for today

Economic growth is vital for improving our lives and the primary long-run determinant of economic growth is innovation. More innovation means better products, more choices for consumers and a higher standard of living. Worldwide, hundreds of millions of people have been lifted out of poverty due to the economic growth that has occurred in many countries since the 1970s.

The effect of innovation on economic growth has been heavily analyzed using data from the post-WWII period, but there is considerably less work that examines the relationship between innovation and economic growth during earlier time periods. An interesting new working paper by Ufuk Akcigit, John Grigsby and Tom Nicholas that examines innovation across America during the late 19th and early 20th century helps fill in this gap.

The authors examine innovation and inventors in the U.S. during this period using U.S. patent data and census data from 1880 to 1940. The figure below shows the geographic distribution of inventiveness in 1940. Darker colors mean higher rates of inventive activity.

geography of inventiveness 1940

Most of the inventive activity in 1940 was in the industrial Midwest and Northeast, with California being the most notable western exception.

The next figure depicts the relationship between the log of the total number of patents granted to inventors in each state from 1900 to 2000 (x-axis) and annualized GDP growth (y-axis) over the same period for the 48 contiguous states.

innovation, long run growth US states

As shown there is a strong positive relationship between this measure of innovation and economic growth. The authors also conduct multi-variable regression analyses, including an instrumental variable analysis, and find the same positive relationship.

The better understand why certain states had more inventive activity than others in the early 20th century, the authors analyze several factors: 1) urbanization, 2) access to capital, 3) geographic connectedness and 4) openness to new ideas.

The figures below show the more urbanization was associated with more innovation from 1940 to 1960. The left figure plots the percent of people in each state living in an urban area in 1940 on the x-axis while the right has the percent living on a farm on the x-axis. Both figures tell the same story—rural states were less innovative.

pop density, innovation 1940-1960

Next, the authors look at the financial health of each state using deposits per capita as their measure. A stable, well-funded banking system makes it easier for inventors to get the capital they need to innovate. The figure below shows the positive relationship between deposits per capita in 1920 and patent production from 1920 to 1930.

innovation, bank deposits 1920-1940

The size of the market should also matter to inventors, since greater access to consumers means more sales and profits from successful inventions. The figures below show the relationship between a state’s transport cost advantage (x-axis) and innovation. The left figure depicts all of the states while the right omits the less populated, more geographically isolated Western states.

innovation, transport costs 1920-1940

States with a greater transport cost advantage in 1920—i.e. less economically isolated—were more innovative from 1920 to 1940, and this relationship is stronger when states in the far West are removed.

The last relationship the authors examine is that between innovation and openness to new, potentially disruptive ideas. One of their proxies for openness is the percent of families who owned slaves in a state, with more slave ownership being a sign of less openness to change and innovation.

innovation, slavery 1880-1940

The figures show that more slave ownership in 1860 was associated with less innovation at the state-level from 1880 to 1940. This negative relationship holds when all states are included (left figure) and when states with no slave ownership in 1860—which includes many Northern states—are omitted (right figure).

The authors also analyze individual-level data and find that inventors of the early 20th century were more likely to migrate across state lines than the rest of the population. Additionally, they find that conditional on moving, inventors tended to migrate to states that were more urbanized, had higher bank deposits per capita and had lower rates of historical slave ownership.

Next, the relationship between innovation and inequality is examined. Inequality has been a hot topic the last several years, with many people citing research by economists Thomas Piketty and Emmanuel Saez that argues that inequality has increased in the U.S. since the 1970s. The methods and data used to construct some of the most notable evidence of increasing inequality has been criticized, but this has not made the topic any less popular.

In theory, innovation has an ambiguous effect on inequality. If there is a lot of regulation and high barriers to entry, the profits from innovation may primarily accrue to large established companies, which would tend to increase inequality.

On the other hand, new firms that create innovative new products can erode the market share and profits of larger, richer firms, and this would tend to decrease inequality. This idea of innovation aligns with economist Joseph Schumpeter’s “creative destruction”.

So what was going on in the early 20th century? The figure below shows the relationship between innovation and two measures of state-level inequality: the ratio of the 90th percentile wage over the 10th percentile wage in 1940 and the wage income Gini coefficient in 1940. For each measure, a smaller value means less inequality.

innovation, inc inequality 1920-1940

As shown in the figures above, a higher patent rate is correlated with less inequality. However, only the result using 90-10 ratio remains statistically significant when each state’s occupation mix is controlled for in a multi-variable regression.

The authors also find that when the share of income controlled by the top 1% of earners is used as the measure of inequality, the relationship between innovation and inequality makes a U shape. That is, innovation decreases inequality up to a point, but after that point it’s associated with more inequality.

Thus when using the broader measures of inequality (90-10 ratio, Gini coeffecieint) innovation is negatively correlated with inequality, but when using a measure of top-end inequality (income controlled by top 1%) the relationship is less clear. This shows that inequality results are sensitive to the measurement of inequality used.

Social mobility is an important measure of economic opportunity within a society and the figure below shows that innovation is positively correlated with greater social mobility.

innovation, social mobility 1940

The measure of social mobility used is the percentage of people who have a high-skill occupation in 1940 given that they had a low-skill father (y-axis). States with more innovation from 1920 to 1940 had more social mobility according to this measure.

In the early 20th century it appears that innovation improved social mobility and decreased inequality, though the latter result is sensitive to the measurement of inequality. However, the two concepts are not equally important: Economic and social mobility are worthy societal ideals that require opportunity to be available to all, while static income or wealth inequality is largely a red herring that distracts us from more important issues. And once you take into account the consumer-benefits of innovation during this period—electricity, the automobile, refrigeration etc.—it is clear that innovation does far more good than harm.

This paper is interesting and useful for several reasons. First, it shows that innovation is important for economic growth over a long time period for one country. It also shows that more innovation occurred in denser, urbanized states that provided better access to capital, were more interconnected and were more open to new, disruptive ideas. These results are consistent with what economists have found using more recent data, but this research provides evidence that these relationships have existed over a much longer time period.

The positive relationships between innovation and income equality/social mobility in the early 20th century should also help alleviate the fears some people have about the negative effects of creative destruction. Innovation inevitably creates adjustment costs that harm some people, but during this period it doesn’t appear that it caused widespread harm to workers.

If we reduce regulation today in order to encourage more innovation and competition we will likely experience similar results, along with more economic growth and all of the consumer benefits.

More competition can lead to less inequality

Wealth inequality in the United States and many European countries, especially between the richest and the rest, has been a popular topic since Thomas Piketty’s Capital in the 21st Century was published. Piketty and others argue that tax data shows that wealth inequality has increased in the U.S. since the late 1970s, as seen in the figure below from a paper by Emmanuel Saez—Picketty’s frequent co-author— and Gabriel Zucman.

top-0-1-income-inequ

The figure shows the percentage of all U.S. household wealth that is owned by the top 0.1% of households, which as the note explains consists of about 160,000 families. The percentage fell from 25% in the late 1920s to about 7% in the late 1970s and then began to rise. Many people have used this and similar data to argue for higher marginal taxes on the rich and more income redistribution in order to close the wealth gap between the richest and the rest.

While politicians and pundits continue debating what should be done, if anything, about taxes and redistribution, many economists are trying to understand what factors can affect wealth and thus the wealth distribution over time. An important one that is not talked about enough is competition, specifically Joseph Schumpeter’s idea of creative destruction.

Charles Jones, a professor at Stanford, has discussed the connection between profits and creative destruction and their link with inequality. To help illustrate the connection, Mr. Jones uses the example of an entrepreneur who creates a new phone app. The app’s creator will earn profits over time as the app’s popularity and sales increase. However, her profits will eventually decline due to the process of creative destruction: a newer, better app will hit the market that pulls her customers away from her product, erodes her sales and forces her to adapt or fail. The longer she is able to differentiate her product from others, the longer she will be in business and the more money she will earn. This process is stylized in the figure below.

firm-life-and-profit2

If the app maintains its popularity for the duration of firm life 1, the entrepreneur will earn profits P1. After that the firm is replaced by a new firm that also exists for firm life 1 and earns profit P1. The longer a firm is able to maintain its product’s uniqueness, the more profit it will earn, as shown by firm life 2: In this case the firm earns profit P2. A lack of competition stretches out a firm’s life cycle since the paucity of substitutes makes it costlier for consumers to switch products if the value of the firm’s product declines.

Higher profits can translate into greater inequality as well, especially if we broaden the discussion to include wages and sole-proprietor income. Maintaining market power for a long period of time by restricting entry not only increases corporate profits, it also allows doctors, lawyers, opticians, and a host of other workers who operate under a licensing regime that restricts entry to earn higher wages than they otherwise would. The higher wages obtained due to state restrictions on healthcare provision, restrictions on providing legal services and state-level occupational licensing can exacerbate inequality at the lower levels of the income distribution as well as the higher levels.

Workers and sole proprietors in the U.S. have been using government to restrict entry into occupations since the country was founded. In the past such restrictions were often drawn on racial or ethnic lines. In their Pulitzer Prize-winning history of New York City, Gotham, historians Edwin G. Burrows and Mike Wallace write about New York City cartmen in the 1820s:

American-born carters complained to the city fathers that Irish immigrants, who had been licensed during the war [of 1812] while Anglo-Dutchmen were off soldiering, were undercutting established rates and stealing customers. Mayor Colden limited future alien licensing to dirt carting, a field the Irish quickly dominated. When they continued to challenge the Anglo-Americans in other areas, the Society of Cartmen petitioned the Common Council to reaffirm their “ancient privileges”. The municipal government agreed, rejecting calls for the decontrol of carting, as the business and trade of the city depended on in it, and in 1826 the council banned aliens from carting, pawnbroking, and hackney-coach driving; soon all licensed trades were closed to them.

Modern occupational licensing is the legacy of these earlier, successful efforts to protect profits by limiting entry, often of “undesirables”. Today’s occupational licensing is no longer a response to racial or ethnic prejudices, but it has similar results: It protects the earning power of established providers.

Throughout America’s history the economy has been relatively dynamic, and this dynamism has made it difficult for businesses to earn profits for long periods of time; only 12% of the companies on the Fortune 500 in 1955 were still on the list in 2015. In a properly functioning capitalist economy, newer, poorer firms will regularly supplant older, richer firms and this economic churn tempers inequality.

The same churn occurs among the highest echelon of individuals as well. An increasing number of the Forbes 400 are self-made, often from humble beginnings. In 1984, 99 people on the list inherited their fortune and were not actively growing it. By 2014 only 28 people were in the same position. Meanwhile, the percentage of the Forbes 400 who are largely self-made increased from 43% to 69% over the same period.

But this dynamism may be abating and excessive regulation is likely a factor. For example, the rate of new-bank formation from 1990 – 2010 was about 100 banks per year. Since 2010, the rate has fallen to about three per year. Researchers have attributed some of the decline of small banks to the Dodd-Frank Wall Street Reform Act, which increased compliance costs that disproportionately harm small banks. Fewer banks means less competition and higher prices.

Another recent example of how a lack of competition can increase profits and inequality is EpiPen. The price of EpiPen—a medicine used to treat severe allergic reactions to things like peanuts—has increased dramatically since 2011. This price increase was possible because there are almost no good substitutes for EpiPen, and the lack of substitutes can be attributed to the FDA and other government policies that have insulated EpiPen’s maker, Mylan, from market competition. Meanwhile, the compensation of Mylan’s CEO Heather Bresch increased by 671% from 2007 to 2015. I doubt that Bresch’s compensation would have increased by such a large amount without the profits of EpiPen.

Letting firms and workers compete in the marketplace fosters economic growth and can help dampen inequality. To the extent that wealth inequality is an issue we don’t need more regulation and redistribution to fix it: We need more competition.

The Economics of Regulation Part 3: How to Estimate the Effect of Regulatory Accumulation on the Economy? Exploring Endogenous Growth and Other Models

This post is the third part in a three part series spurred by a recent study by economists John Dawson and John Seater that estimates that the accumulation of federal regulation has slowed economic growth in the US by about 2% annually.  The first part discussed generally how Dawson and Seater’s study and other investigations into the consequences of regulation are important because they highlight the cumulative drag of our regulatory system. The second part went into detail on some of the ways that economists measure regulation, highlighting the strengths and weaknesses of each.  This post – the final one in the series – looks at how those measures of regulation are used to estimate the consequences of regulatory policy.  As always, economists do it with models.  In the case of Dawson and Seater, they appeal to a well-established family of endogenous growth models built upon the foundational principle of creative destruction, in the tradition of Joseph Schumpeter.

So, what is an endogenous growth model?

First, a brief discussion of models:  In a social or hard science, the ideal model is one that is useful (applicable to the real world using observable inputs to predict outcomes of interest), testable (predictions can be tested with observed outcomes), flexible (able to adapt to a wide variety of input data), and tractable (not too cumbersome to work with).  Suppose a map predicts that following a certain route will lead to a certain location.  When you follow that route in the real world, if you do not actually end up at the predicted location, you will probably stop using that map.  Same thing with models: if a model does a good job at predicting real world outcomes, then it sticks around until someone invents one that does an even better job.  If it doesn’t predict things well, then it usually gets abandoned quickly.

Economists have been obsessed with modeling the growth of national economies at least since Nobel prize winner Simon Kuznets began exploring how to measure GDP in the 1930s.  Growth models generally refer to models that try to represent how the scale of an economy, using metrics such as GDP, grows over time.  For a long time, economists relied on neoclassical growth models, which primarily use capital accumulation, population growth, technology, and productivity as the main explanatory factors in predicting the economic growth of a country. One of the first and most famous of such economic growth models is the Solow model, which has a one-to-one (simple) mapping from increasing levels of the accumulated stock of capital to increasing levels of GDP.  In the Solow model, GDP does not increase at the same rate as capital accumulation due to the diminishing marginal returns to capital.  Even though the Solow model was a breakthrough in describing the growth of GDP from capital stock accumulation, most factors in this growth process (and, generally speaking, in the growth processes of other models in the neoclassical family of growth models) are generated by economic decisions that are outside of the model. As a result, these factors are dubbed exogenous, as opposed to endogenous factors which are generated inside of the model as a result of the economic decisions made by the actors being modeled.

Much of the research into growth modeling over the subsequent decades following Solow’s breakthrough has been dedicated to trying to “endogenize” those exogenous forces (i.e. move them inside the model). For instance, a major accomplishment was endogenizing the savings rate – how much of household income was saved and invested in expanding firms’ capital stocks. Even with this endogenous savings rate, as well as exogenous growth in the population providing labor for production, the accumulating capital stocks in these neoclassical growth models could not explain all of the growth in GDP. The difference, called the Solow Residual, was interpreted as the growth in productivity due to technological development and was like manna from heaven for the actors in the economy – exogenously growing over time regardless of the decisions made by the actors in the model.

But it should be fairly obvious that decisions we make today can affect our future productivity through technological development, and not just through the accumulation of capital stocks or population growth. Technological development is not free. It is the result of someone’s decision to invest in developing technologies. Because technological development is the endogenous result of an economic decision, it can be affected by any factors that distort the incentives involved in such investment decisions (e.g., taxes and regulations). 

This is the primary improvement of endogenous growth theory over neoclassical growth models.  Endogenous growth models take into account the idea that innovative firms invest in both capital and technology, which has the aggregate effect of moving out the entire production possibilities curve.  Further, policies such as increasing regulatory restrictions or changing tax rates will affect the incentives and abilities of people in the economy to innovate and produce.  The Dawson and Seater study relies on a model originally developed by Pietro Peretto to examine the effects of taxes on economic growth.  Dawson and Seater adapt the model to include regulation as another endogenous variable, although they do not formally model the exact mechanism by which regulation affects investment choices in the same way as taxes.  Nonetheless, it’s perfectly feasible that regulation does affect investment, and, to a degree, it is simply an empirical question of how much.

So, now that you at least know that Dawson and Seater selected an accepted and feasible model—a model that, like a good map, makes reliable predictions about real world outcomes—you’re surely asking how that model provided empirical evidence of regulation’s effect on economic growth.  The answer depends on what empirical means.  Consider a much better established model: gravity.  A simple model of gravity states that an object in a vacuum near the Earth’s surface will accelerate towards the Earth at 9.81 meters per second squared. On other planets, that number may be higher or lower, depending on the planet’s massiveness and the object’s distance from the center of the planet.  In this analogy, consider taxes the equivalent of mass – we know from previous endogenous growth models that taxes have a fairly known effect on the economy, just like we know that mass has a known effect on the rate of acceleration from gravitational forces.  Dawson and Seater have effectively said that regulations must have a similar effect on the economy as taxes.  Maybe the coefficient isn’t 9.81, but the generalized model will allow them to estimate what that coefficient is – so long as they can measure the “mass” equivalent of regulation and control for “distance.”  They had to rely on the model, in fact, to produce the counterfactual, or to use a term from science experiments, a control group.  If you know that mass affects acceleration at some given constant, then you can figure out what acceleration is for a different level of mass without actually observing it.  Similarly, if you know that regulations affect economic growth in some established pattern, then you can deduce what economic growth would be without regulations.  Dawson and Seater appealed to an endogenous growth model (courtesy of Perreto) to simulate a counterfactual economy that maintained regulation levels seen in the year 1949.  By the year 2005, that counterfactual economy had become considerably larger than the actual economy – the one in which we’ve seen regulation increase to include over 1,000,000 restrictions.