In this methodology project, we have investigated how cohort analysis can be used to analyze entrepreneurship and new firms’ contribution to the economy.
Cohort analysis involves selecting a group of firms that started around the same time – the cohort – and following their development over time.
We discuss the pros and cons of cohort analysis, and how the method relates to previous research and existing statistical projects that focus on entrepreneurship. An important part of the project has also been to develop a method in order to solve the technical challenges that arise when following firms over long periods of time in business register data (the firm dynamics). To illustrate the potential of the cohort method in combination with Growth Analysis databases, and to test our method for defining firm dynamics, we conduct a cohort analysis of new firms that started in 2007 in Sweden. We follow several aspects of the cohort’s development over a ten-year period, or until 2017.
The technical discussions around methodology in this report are aimed primarily at those who work with statistics and analyzes of entrepreneurship and firm dynamics. The cohort analysis of 2007 years’ new firms may be of interest to anyone interested in entrepreneurship and its role in the economy.
In order to get a complete and reliable picture of firms’ development based on registry data, we need robust definitions for both the firm and the so called firm-dynamic events (referred to by Eurostat as demographic events) that a firm can undergo during its life cycle. Examples of such events are a firm’s start-up, closure and acquisition.
In order to achieve this, we have used Statistics Sweden's method for firm and establishment dynamics (FAD) as an inspiration. FAD is based on the RAMS linked employer-employee register. The method identifies firm-dynamic events by following the flows of employed individuals within and between firms, and thus produces more stable firm units than when only following corporate identity numbers.
Our method is based on FAD but with the following adjustments:
Our hope is that this method will facilitate future analyses of firm dynamics at Growth Analysis. However, the method should be regarded as a first draft, rather than a final proposal. In the future, we see opportunities to further develop and refine the method. Among other things, by adding data sources that Growth Analysis does not have access to currently. An in-depth discussion of such opportunities is, however, not included in this report.
The cohort in our analysis consists of new firms that started their business in 2007. We follow their development over a ten year-period, up to and including 2017. Given that this is the first time we are testing our method for defining firm dynamics, the results should be interpreted with caution.
The results from the analysis show that:
The main strength of a cohort analysis is that it controls for the economic and political environment in which the new firms grow up in (the framework conditions). This makes it easier for the reader to absorb the analysis and put the cohort's development in relation to the framework conditions, compared to aggregating firms that started over a wider time span. An example of a framework condition that is interesting to study is state support measures that were available to a given cohort.
A cohort analysis of entrepreneurship gives us a perspective of what firm life cycles tend to look like. It allows us to investigate questions such as: What is a normal life expectancy for firms in a particular industry? At what age do firms tend to be acquired? Which firms grow and which do not?
An important aspect of a cohort analysis is how the cohort has contributed to the development of the business sector as a whole. Cohort analysis provides the opportunity to study the contribution of new firms in different stages of development.
One disadvantage of cohort analyses is that the external validity is limited in case there are substantial differences between cohorts. In other words: to what extent can conclusions be drawn regarding entrepreneurship generally based on an individual cohort? In the report we show that some patterns are relatively stable between different cohorts, at least for cohorts that started within a span of six years. This applies, for example, to the survival rate of companies. In other cases there may be significant differences, especially on a more disaggregated level. An example is the spike in pharmacy entrepreneurship in the year when that sector was privatized.
Based on the cohort analysis which we undertake in the report, we conclude that it is often effective to switch between analyzing individual cohorts and aggregating multiple cohorts, in order to gain as full an understanding as possible of the dynamics of new firms in a given period.
We also discuss how the method that we focus on in this study relates to existing research and statistics. There are several international statistical projects that aim to develop internationally comparable statistics on entrepreneurship based on microdata. For example, the OECD and Eurostat run the Entrepreneurship Indicators Program and DynEmp. The strength of these projects is the international comparability. But that is also a limitation, as the method becomes a "lowest common denominator" that is feasible in multiple countries with different conditions. This means that the analysis will be more aggregated and of lower quality than it could be if more detailed data, similar to that which is available in Sweden, would be used.
We would like to emphasize that all statistics should be interpreted with caution. Correlation does not necessarily imply causality, and it is not always clear which conclusions should be drawn from the observed patterns. However, in order to understand the underlying mechanisms, one must first begin by gathering empirical evidence of what the patterns are. The policy discussion on how the statistics should be interpreted and what conclusions should be drawn can then proceed from there, on a solid empirical basis.
We see several opportunities to build on this method development project. For example, the method for defining firm dynamics can be further refined, in particular through access to more data sources and collaboration with stakeholders such as Statistics Sweden and the OECD. The cohort analysis can provide inspiration for in-depth studies of various aspects of entrepreneurship. Aspects from the method and the analysis could also be applied in the cohort analysis on entrepreneurship that is performed every three years by Growth Analysis as part of the official statistics production. Another interesting aspect is to investigate how the comprehensive statistical information in this report can be presented in new, more accessible ways, for example through interactive web-based applications.
Serial number: PM 2020:07
Reference number: 2019/097