Ralph de Haas, Vincent Sterk and Neeltje van Horen
Anaemic productivity growth and limited business dynamism remain key policy concerns in Europe and the US. Policies to improve macroeconomic performance often target existing firms. Examples include tax measures to stimulate firm-level Research & Development and structural reforms to eliminate distortions in labour, financial, and product markets. In a new paper we investigate an entirely different policy lever, one that has so far remained largely unexplored: influencing the types of firms that are being started in the first place. Using a comprehensive new data set on European start-ups, we show how tax policies that shift the composition of new start-up cohorts could deliver meaningful macroeconomic gains.
The idea of improving the composition of new start-up cohorts (as opposed to ‘fixing’ already established firms) appears attractive for two reasons. First, because the rates of firm entry and exit are high, typically around 10% annually. This means that the majority of firms that will be in operation 20 years from now are yet to founded, while many current firms will no longer exist by then.
Second, forward-looking policies to shift the composition of start-up cohorts also appear attractive because start-ups are key drivers of job creation and productivity growth. Yet, start-ups are not a homogeneous group but come in all shapes and sizes. Some entrepreneurs are simply interested in starting a small, basic firm and do not have much ambition to grow their enterprise. Others have grander ambitions and try to scale-up their production as quickly as possible. Recent evidence shows that this ex-ante heterogeneity among newly established firms helps to predict their performance later in life. It follows that structural policies that successfully shift the mix of start-up types that enter the economy, may generate important macroeconomic impacts.
Not all start-ups are the same…
To better understand how start-ups differ, we collected unique new data on European start-ups in close collaboration with the Competitiveness Research Network (CompNet). The resulting data set contains detailed information on all start-ups established between 2002 and 2019 in Croatia, Denmark, Finland, France, Italy, Lithuania, the Netherlands, Slovenia, Spain and Sweden.
Because start-up types are not readily observed, we first have to classify start-ups into different types. We do so by using K-means clustering, an unsupervised machine learning algorithm. Clustering allows us to find and analyse groups of start-ups that form organically based on features that they share in a multidimensional space. The algorithm groups the data into k clusters and uses the distance between points as a measure of similarity. We feed the algorithm five key variables that entrepreneurs decide on when starting their business: employment; the capital-to-labour ratio; total assets; the leverage ratio and the cash-to-assets ratio.
The algorithm uncovers the presence of five well-separated clusters of start-ups, which we label large; capital intensive; high-leverage; cash-intensive and basic. This classification captures 50%–70% of the variation in the above mentioned variables. An interesting stylised fact is that these five types are present in all countries (Chart 1), in all (broad) economic sectors, and in all start-up cohorts – although their exact shares differ somewhat across countries, industries, and years. Furthermore, the initial differences between the types are persistent. For example, high-leverage start-ups (14% of all start-ups) start their operations on average with a leverage ratio of 1.2, much higher than other types. Over time, the excess leverage is reduced, but remains above that of the other types.
Chart 1: Distribution of start-up types by country
Notes: This figure illustrates the distribution of the start-up population for individual across the five start-up types. The start-up population comprises all cohorts available for each country.
The five start-up types perform very different over their life cycle. In particular, they display large and persistent differences in employment, productivity and exit rates. For example, the performance of the high-leverage type is consistently poor. These young firms are more likely to exit than any other start-up type and their productivity and profit levels are relatively low. By contrast, start-ups that are capital-intensive (7% of all start-ups) or cash-rich (26%) boast higher productivity levels and lower exit rates.
Corporate taxation as a policy instrument
Given the large differences across start-up types in how they develop over time, the mix of start-ups can potentially have significant macroeconomic effects. To provide insights into the economic relevance of this start-up composition channel we calibrate a simple firm-dynamics model in the tradition of Hopenhayn (1992). This model describes an economy with many firms that each have their own production function and level of productivity.
We use this model to evaluate the macroeconomic impacts of a budget-neutral change in corporate income taxation. More specifically, we analyse the impacts of a large number of possible policies that explicitly differentiate between start-up types in terms of the tax rate they face. Such changes obviously alter the incentives of different types to start operations and hence affects the start-up mix. We use this model to help us understand how much aggregate employment and labour productivity could in principle improve through this start-up composition channel.
This exercise shows that it is possible to reap substantial macroeconomic gains by actively influencing the mix of new startup cohorts. Table A provides two examples. The first two columns evaluate a policy that focuses on stimulating labour productivity. The first column shows how the tax rate changes for each start-up type. The basic start-ups, for example, will be paying a 3.1 percentage point higher rate, while the capital-intensive ones a 27.6 percentage point lower rate (for example, by replacing a 25% tax rate by a small subsidy). The second column shows how this affects the shares of the various types. Such change in taxation shifts the composition of new start-up cohorts towards more capital-intensive firms while reducing the share of basic start-ups. Since the former have much higher levels of labour productivity than the latter, aggregate labour productivity increases. Columns 3 and 4 show a similar exercise, except the focus is now stimulating employment. In this case, the policy stimulates the entry of large start-ups and discourages the entry of cash-rich start-ups. This shift in composition leads to an increase in employment of roughly 3%.
Table A: Policy experiment – tax differentiation and macroeconomic outcomes
Conclusions
Given high corporate entry and exit rates, policymakers aiming to improve macroeconomic performance may consider policies that explicitly target the composition of incoming generations of firms. The method outlined in this column is based on measurable criteria and therefore straightforward to implement. This not only makes it a potentially useful policy tool, but also a valuable complement to standard analyses evaluating the macro effects of tax reforms, which typically ignore impacts on the composition of new start-up cohorts.
Ralph de Haas works at the European Bank for Reconstruction and Development, Vincent Sterk works at the University College London and Neeltje van Horen works in the Bank’s Research Hub.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
