Entrepreneurial Ecosystem Diagnostics

Annex A. Methods and data

The national entrepreneurial ecosystem diagnostics presented in this report quantify each of ten entrepreneurial ecosystem “input” elements through composite indexes, while entrepreneurial ecosystem outputs are measured through a set of metrics external to those used to compute the element scores. This Annex sets out the main features of the methodology used to compute these ten composite indexes tracking the entrepreneurial ecosystem elements.

A complementary working paper details the methods and approaches used to produce the ten elements, how the indicators were selected, normalised, and aggregated and approaches to addressing missing values (Crotti et al., 2025 (forthcoming)[1])

Indicator selection

The choice of indicators to be included in this benchmarking report has been guided by the principles of the Statistical Quality Framework. The principles are identified in bold below, accompanied by a description of how each principle has been integrated into the choice of indicators for this benchmarking exercise:

1. Timeliness: Only indicators updated regularly and for which time series exist are considered. This is to enable the measurement of progress over time and to offer a current view of the situation in each country.

2. Coverage and accessibility: Only indicators available for significant share of OECD member countries are considered. To operationalize the principle, only indicators that contain at least a datapoint over the past 10 years for 80% of OECD countries are considered. This limits the presence of missing-values, enhancing international comparability.

3. Reliability and comparability: Only data computed according to high quality standards produced by reputable sources are considered. To make sure that statistics are internationally comparable, data are collected from a single source that applies consistent definitions and methods to all countries.

4. Relevance and accuracy: Only data that capture the theoretical concept well are considered. This involves a qualitative assessment of which metric is a reasonable representation of the concept among those available that match the other criteria above. Both direct and indirect measures of the underlying concepts are used. Abstract concepts are more easily measured indirectly or using proxies. Countable concepts are better measured with direct metrics. Some composite elements are more complex than others and require multiple indicators that capture different (non-overlapping) parts of an entrepreneurial ecosystem element.

5. Evidence base on “congruence”: When possible, indicators are selected based on credible studies that have shown a tendency of this measure to relate to productive entrepreneurship. The objective is of constructing the most robust composite measure possible using the lowest number of indicators (Leendertse, Schrijvers and Stam, 2022[2]).

A Principal Components Analysis (PCA) and a correlation analysis have been conducted to complement these conceptual and data-quality considerations with statistical considerations, testing the redundancy and fit of considered indicators. To respect the parsimony principle, limit potential collinearity, and avoid implicitly giving more weight to a particular aspect of a concept, indicators that are not unique and are highly correlated with one another are excluded from the computation. Notably, the Institutions element is an area where indexes are available that often capture overlapping aspects of governance or institutional quality. We seek to exclude overlapping indicators from the framework.

Although informed by statistical analysis, the selection and combination of indicators remains an area where quantitative results intertwine with conceptual considerations, which is why building composite indexes is often referred as both ’an art and a science’ (Nardo et al., 2005[3]). The resulting indicators selection reflects a synthesis of theoretical considerations, empirical considerations, literature review and results of the statistical analysis. In some cases, the paucity of indicators in some domains requires the use of proxy measures. Notably, there are very few available metrics measuring Networks, Leadership, and Intermediate services well. In these case proxies are used to measures these elements to the best extent as possible.

The overall, the selection of indicators for the assessment of the entrepreneurial ecosystem elements has been based on the OECD Statistical Quality Principles, outcomes of the statistical analysis, and a review of the entrepreneurship literature (e.g. (Leendertse, Schrijvers and Stam, 2022[2])).

Details of the final selection of indicators are provided in Table A A.1 and A A.2, while the list of indicators considered but discarded for not meeting the principles stated above is provided in Table A A.3.

Table A A.1. Indicators used to compute the entrepreneurial ecosystem elements

Element	Indicator (units)	Description	Pre-normalisaiton transformation	Source
1.Institutions	Rule of law, 0-100 best	Index scaled 0-100 built by combining eight questions. i. Extent to which violent crime pose a significant problem for government and/or business (0 no problem - 4 significant problem). ii. Extent to which organised crime likely to be a problem for government and/or business? (0 no problem - 4 significant problem). iii. Extent to which extent to there is a risk that the legal process/the courts can be interfered with or distorted to serve particular interests (0 verz low - 4 very high likelihood). iv. Extent to which there is a risk that contract rights will not be enforced (0 minimal risk - 4 high risk). v. Extent to which the judicial process is speedy and efficient (0 very speedy, 4 very slow). vi. Extent to which there is a risk of expropriation of foreign assets (0 no risk -4 high risk). vii. Intellecatual property protection (0 very good - 4 very low protection).viii. Private property rights (0 very high, 4 very low guarantee). Respondents are experts from a network of over 500 correspondents, reviewed for consistency by panels of regional experts.	Moving average	Economist Intelligence Unit (EIU) accessed via World Bank - Worldwide Governance Indicators
	Effective tax rate, % taxable income	Effective average tax rate, composite tax, computed under the country-specific interest and inflation rates scenario.	Moving average	OECD - Corporate Tax Statistics Database
	Product Market Regulation, Index 0-6 stringent	Product Market Regulation Index. It measures the degree to which policies promote or inhibit competition in areas of the product market where competition is viable. A higher value indicates a higher level of regulatory stringency. It is computed through multiple aggregation layers. The overall index is computed as the average of Distortions Induced by State Involvement sub-index and the Barriers to Domestic and Foreign Entry sub-index. The Distortions Induced by State Involvement sub-index includes Quality and Scope of Public Ownership, Governance of SOEs, Retail Price Controls and Regulation, Involvement in Business Operation in Network Sectors, Involvement in Business Operations in Services Sectors, Public Procurement, Assessment of Impact on Competition, Interaction with Stakeholders. The Barriers to Domestic and Foreign Entry sub-index includes Administrative Requirements for LLCs and POES, Communication and Simplification of Administrative and Regulatory Burden, Barriers to entry in Service Sectors, Barriers to entry in Network Sectors, Barriers to FDI, Barriers to Trade Facilitation, Tariff Barriers. Each measure captures de jure regulation: they reflect the status of existing laws and regulations. They are based on a questionnaire compiled by national authorities and vetted by OECD experts. For more details refer to https://www.oecd.org/en/topics/product-market-regulation.html.	Moving average	OECD
	Control of corruption index, 0-100 low incidence	Index of corruption based on population survey questions about frequency and spread of corruption.	Moving average	Varieties of Democracy Project (Videm) accessed via World Bank - Worldwide Governance Indicators
	Entrepreneurship as a good career choice, % 18-64 pop.	Percentage of 18-64 population who agree with the statement that in their country, most people consider starting a business as a desirable career choice.	Moving average	Global Entrepreneurship Monitor (GEM)
	High status to successful entrepreneurs, % 18-64 pop.	Percentage of 18-64 population who agree with the statement that in their country, successful entrepreneurs receive high status.	Moving average	Global Entrepreneurship Monitor (GEM)
	Trust in others, % respondents	Share of respondents who believe that most people can be trusted.	Moving average	World Value Survey (WVS)
3.Networks	SMEs collaborating on innovation, % total SMEs	SMEs with innovation cooperation activities as a share of all SMEs. The small and medium-sized enterprises (SMEs) with innovation cooperation activities include all enterprises that had any cooperation agreements on innovation activities with other enterprises or institutions in the three years of the survey period. It corresponds to the indicator "3.2.1 Innovative SMEs collaborating with others" in the EU scoreboard.	Moving average	European Commission - European innovation scoreboard
	University-business collaboration, 1-7 best	Average score across respondents who answred the question: in your country, to what extent do business and universities collaborate on research and development (R&D)? [1 = Not at all; 7 = To a great extent].	Moving average	World Economic Forum
4. Infrastructure	Fix broadband, subs. per 100 pop.	Total fixed broadband subscriptions per 100 population. Fixed broadband technologies corresponds to DSL, cable modem, fiber-to-the-home and other fixed technologies (such as broadband over power-line and leased lines).	Moving average	OECD - Telecommunications database
	Transport infrastructure quality, 1-5 high	Logistics performance index: Quality of trade and transport-related infrastructure (1=low to 5=high).	Moving average	World Bank - Logistic Performance Index (LPI)
	Mobile data use, Gb per subscrip./month	Mobile data use, Gigabits per subscription per month.	Moving average	OECD - Broadband and telecom databases
5. Markets	Gross domestic product, PPP$ million	Gross domestic product, expenditure approach, expressed in Purchasing Power Parity (PPP - international dollar).	Logarithm of moving average	OECD - Annual GDP and components
	Trade facilitation index, 0-2 best	Average trade facilitation performance computed as the average of the following aspects of border procedures: A. Information availability, B. Involvement of the trade community, C. Advance ruling, D. Appeal procedures, E. Fees and charges, F. Documents, G. Automation, H. Procedures, I. Internal border agency cooperation, J. External border agency cooperation, K. Governance and impartiality.	Moving average	OECD - Trade Facilitation Indicators
6.Finance	Early-stage VC investment, USD per capita	Total venture capital investments is the sum of seed, start-up and other early stage venture capital investments, in current USD, divided by the total population (USD dollars per person).	Moving average	OECD Entrepreneurship Financing Database
	Later-stage VC investment, USD per capita	Later stage venture capital investments, in current USD, divided by the total population (USD dollars per person). Growth stage investments is defined as total investments minus seed, start-up and other early stage investments.	Moving average	OECD Entrepreneurship Financing Database
	Outstanding SME loans, thousands USD per capita	Outstanding loans to SMEs divided by population.	Moving average	OECD - Financing SMEs and Entrepreneurs: An OECD Scoreboard
	Factoring, thousands USD per capita	Total value of factoring contracts, divided by population.	Moving average	OECD - Financing SMEs and Entrepreneurs: An OECD Scoreboard
7.Knowledge	Patents, per million pop.	Number of patents filed at triadic patent families per capita.	Moving average	OECD - Main Science and Technology Indicators
	R&D expenditure, % GDP	Gross Domestic Expenditure on R&D, % GDP.	Moving average	OECD - Main Science and Technology Indicators
	GitHub software uploads, per thousand people	Number of times developers in a country uploaded code to GitHub per thousand people.	Moving average	GitHub
8.Talent	Mean years of schooling, years	Average number of completed years of education of a country's population aged 25 years and older, excluding years spent repeating individual grade. In general, the indicator’s value denotes the level of skills and competencies of a country's population, which could be seen as a proxy of both the quantitative and qualitative aspects of the stock of human capital. A relative high value indicates great shares of the adult population according to the highest level of education attained or completed, and reflects a performing educational system. Caution is required when using this indicator for cross-country comparison. Countries do not always classify degrees and qualifications at the same ISCED levels, even if they are received at roughly the same age or after a similar number of years of schooling. Also, certain educational programmes and study courses cannot be easily classified according to ISCED levels. It is also important to note that this indicator is based on education levels attained or completed, in terms of years of schooling, and do not necessarily reveal the quality of the education (e.g. learning achievement or other outcomes).	Moving average	UNESCO
	Pisa, score	Average of Math, Reading, and Science of Pisa test scores.	Moving average	OECD - PISA
	Internet users, % pop.	Individuals who have used the internet as a share of the total population. Internet users are individuals who have used the Internet (from any location) in the last 3 months. The Internet can be used via a computer, mobile phone, personal digital assistant, games machine, digital TV etc.	Moving average	World Bank, World Development Indicators
	Perceived entrepreneurial capabilities, % 18-64 pop.	Percentage of 18-64 population who believe they have the required skills and knowledge to start a business.	Moving average	Global Entrepreneurship Monitor (GEM)
9.Leadership	Serial entrepreneurs, unit count	Number of people described as serial entrepreneurs in the crunchbase's individuals database.	Logarithm of moving average	Crunchbase
10.Intermediate services	Incubators, per million pop.	Total count of active incubators, accelerators, co-working spaces, entrepreneurship programmes, and startup competition programmes, divided by the total population.	Moving average	Crunchbase and OECD
	Coaches, unit count	Number of people described as mentors or coaches in the crunchbase's individuals database.	Logarithm of moving average	Crunchbase
	Technical employment, % total employment	Proxy measure for availability of experts in technical domains. It is measures in terms of employment in professional, scientific and technical activities as a share of total employment. Professional, scientific and technical activities correspond to the section M of UN's international standard industrial classification of all economic activities (ISEC, rev.4) and includes: (Division 69) Legal and accounting activities, (Division 70) Activities of head offices; management consultancy activities, (Division 71) Architectural and engineering activities; technical testing and analysis, (Division 72) Scientific research and development, (Division 73) Advertising and market research, (Division 74) Other professional, scientific and technical activities (e.g. design, photographic activities, other).	Moving average	OECD

Table A A.2. Indicators used to benchmark entrepreneurship output and variation

Element	Indicator (units)	Description	Pre-normalisaiton transformation	Source
Output	Birth rate of employer enterprises, % business pop.	New firm creation among employer firms (i.e. with at least one employee), as a proportion of active business population.	Moving average	OECD - SDBS business demography
	Medium and high-growth enterprises, %	Rate of medium and high-growth enterprises (10%+ growth based on employment) as a proportion of active business population.	Moving average	OECD - SDBS business demography
	Equity-based young firms, per million pop.	Number of companies aged 0-5 years old, divided by the population. Since the Crunchbase dataset is skewed towards equity-based firms, the indicator better represents equity-based firms that are 0-5 years old.	Moving average	Crunchbase, OECD
	Unicorns, per million pop.	Number of private companies with a valuation over USD 1 billion, divided by the population.	Moving average	CB Insights
	Enterprise churn rate, % business pop.	Sum of births and deaths of employer enterprises (firms with at least 1 employee) as a proportion of active business population.	Moving average	OECD - SDBS business demography
	2-year-old employer enterprises, % business population	2-year-old employer enterprises as a proportion of active business population.	Moving average	OECD - SDBS business demography
	3-year survival rate., % new employer enterp.	3-year survival rate of employer enterprises, as a proportion of new employer enterprises.	Moving average	OECD - SDBS business demography
	Expectation to create jobs, % entrepreneurs	Percentage of those involved in early-stage entrepreneurial activity who expect to create 6 or more jobs in 5 years.	Moving average	Global Entrepreneurship Monitor (GEM)
Variation	Geographical dispersion of start-ups, 0-100 high conc.	Herfindahl Hirschman Index, calculated using the share of equity-based start-ups (0-5 years old) located in different cities within a country. A higher value of this indicator implies a higher concentration of start-ups in the main cities.	Moving average	Crunchbase
	Women founders, % founders	Female founders as a share of total founders. Founders are defined as people whose job description refers to CEO, founder, co-founder, member of the board or partner. To avoid that extremely small samples drive results, the women founders share is computed only if the total number of entrepreneurs recorded in Crunchbase is at least 25.	Moving average	Crunchbase
	Missing entrepreneurs’ rate, % early stage entrepreneurs	Size of estimated "missing entrepreneurs" group divided by the population all early-stage entrepreneurs (missing entrepreneurs calculated as the additional number of entrepreneurs there would be if women, youth, seniors and immigrants created businesses at the same rate as males 30-49 years old).	Moving average	OECD Missing Entrepreneurs publication series

Table A A.3. Indicators considered but discarded

Element	Indicator (units)	Description	Source	Rationale
1.Institutions	SME income tax rate, % taxable income	SME Combined targeted corporate income tax rate	OECD	Too few countries covered
	Civil justice index, 0-100 best	The civil justice composite index is computed as the simple average of the following 7 variables: i. People can access and afford civil justice, ii. Civil justice is free of discrimination, iii. Civil justice is free of corruption, iv. Civil justice is free of improper government influence, v. Civil justice is not subject to unreasonable delay, vi. Civil justice is effectively enforced, vii. Alternative dispute resolution mechanisms are accessible, impartial, and effective. Each of these variables in turn, is a combination of one or more survey questions.	World Justice Project (WJP)	Too few countries covered
	Regulatory enforcement index, 0-100 best	The regulatory enforcement composite index is computed as the simple average of the following 5 variables: i. Government regulations are effectively enforced, ii. Government regulations are applied and enforced without improper influence, iii. Administrative proceedings are conducted without unreasonable delay, iv. Due process is respected in administrative proceedings, v. The government does not expropriate without lawful process and adequate compensation.	World Justice Project (WJP)	Too few countries covered
	SME implied tax subsidy rates, index	Implied tax subsidy rates on R&D expenditures of SME under a profitable profit scenario. The tax subsidy rate is defined as 1 minus the B-index, a measure of the before-tax income needed by a “representative” firm to break even on one additional monetary unit of R&D outlay (Warda, 2001). As tax component of the user cost of R&D, the B-Index is is directly linked to measures of effective marginal tax rates.	OECD - R&D Tax Incentives database	Too few countries covered
	Absence of corruption index, 0-100 best	The absence of corruption composite index is computed as the simple average of the following 4 variables: i. Government officials in the executive branch do not use public office for private gain, ii. Government officials in the judicial branch do not use public office for private gain, iii. Government officials in the police and the military do not use public office for private gain, iv. Government officials in the legislative branch do not use public office for private gain.	World Justice Project (WJP)	Too few countries covered
	Access to civil justice, 0-100 best	This indicator is a combination of survey questions that assess if the general population is aware of civil justice remedies, can afford legal representation, can access courts, how they assess procedural complexity to access civil justice . It corresponds to the sub-indicator "People can access and afford civil justice" of the Civil justice sub-index, from the WJP's Rule of law index.	World Justice Project (WJP)	Too few countries covered
	Government effectiveness, 0-100 best	Index scaled 0-100 built by combining two questions. i. How pervasive is red tape. ii. To what degree do vested interests/cronyism distort decision-making in the public and/or private sectors? Respondents are experts from a network of over 500 correspondents, reviewed for consistency by panels of regional experts.	Economist Intelligence Unit (EIU) accessed via World Bank - Worldwide Governance Indicators	Too few countries covered
2. Culture	Creative thinking PISA score, index	Average score obtained by students on the creative thinking PISA test, index	OECD	Too few countries covered
	Fear of failure, % 18-64 pop.	Percentage of the 18-64 population who agree that they see good opportunities but would not start a business for fear it might fail.	Global Entrepreneurship Monitor (GEM)	Volatile results
	Openness to technology, % respondents	Share of respondents who believe that in the future, more emphasis on technology is a good thing	World Value Survey (WVS)	Too few countries covered
	Propensity to imagination, % respondents	Share of respondents who believe that imagination is an important quality for a child	World Value Survey (WVS)	Conceptual miss-match
3.Networks	Participants to business events, per million pop.	Total number of all participants to business events that took place in a country in a given year as registered in Crunchbase. All types of participants are counted. It is divided by the total population.	Crunchbase	Conceptual miss-match
	Public-private co-patenting, % patents	Share of co-inventions (simple patent families) developed jointly by public-private partnerships across all technologies, as a share of total patents.	WIPO	Conceptual miss-match
	International co-patents, % co-inventions	Share of co-inventions (simple patent families) developed jointly by at least two inventors from different countries across all technologies. They are calculated based on inventor country - integer counts by country of residence of the inventors, and the first filing date worldwide, under the Paris Convention. The priority date is considered to be closest to the actual date of invention.	OECD - STI Micro-data Lab: Intellectual Property Database	Conceptual miss-match
	Domestic co-patents, % co-inventions	Share of co-inventions (simple patent families) developed jointly by at least two inventors from the same country, across all technologies. They are calculated based on inventor country - integer counts by country of residence of the inventors, and the first filing date worldwide, under the Paris Convention. The priority date is considered to be closest to the actual date of invention.	OECD - STI Micro-data Lab: Intellectual Property Database	Conceptual miss-match
4. Infrastructure	Density of rail network, Km per Sq. Mt.	Total Kilometers of rail roads divided by the country's surface area	OECD transport infrastructure indicators	Conceptual miss-match
	Density of road network, Km per Sq. Mt.	Total Kilometers of roads divided by the country's surface area	OECD transport infrastructure indicators	Conceptual miss-match
	Air transport, passengers per capita	Air passengers carried include both domestic and international aircraft passengers of air carriers registered in the country. It is divided by the country's population.	World Bank - World Development Indicators	Conceptual miss-match
	Mobile broadband, subs. per 100 pop.	Total mobile broadband subscriptions per 100 population. Mobile broadband subscriptions are mobile subscriptions that advertise data speeds of 256 kbit/s or greater.	OECD - Telecommunications database	Better alternative available
	Firms with 100 Mb/s connection, % firms	Percentage of enterprises with at least 10 employees with a fixed broadband connection with download speed at least 100 Mbit/s	OECD ICT Access and Usage by Businesses database	Too few countries covered
5. Markets	Net disposable income, PPP, per capita	Net disposable income per capita, PPP-adjusted US dollars, chain linked volume.	Per person	Conceptual miss-match
	Population, million	Country population expressed in millions.	Thousands persons	Conceptual miss-match
	Public procurement expenditure, million $ per capita	Total public procurement expenditure (USD million) by all levels of public administration.	PPP million	Too few countries covered
6.Finance	Interest rate spread, % large	Interest rate spread between interests applied to SME loans and interests applied to large company loans, expressed as a share of the interest rates paid by large companies.	OECD Scoreboard	Too few countries covered
	Leasing, USD thousands per capita	Total value of factoring expressed in thousands USD per capita.	OECD Scoreboard	Too few countries covered
7.Knowledge	Citable documents, per million pop.	Number of citable documents (articles, reviews and conference papers) published per million people.	Per million people	Conceptual miss-match
8.Talent	Tertiary education attainment, % working age pop.	Share of adults 25-64 years old who have attained at least tertiary education degree.	% adult population	Conceptual miss-match
	ICT technicians, % total employees	Percentage of ICT professional and technicians in total employment. It is computed as the sum of Professionals (ISCO-08) and Technicians and associate professionals (ISCO-08) in the Information and communication sector (Economic activity J. (ISIC-Rev.4)), divided by all employees in all sectors.	International Labour Organisation (ILO) - ILOSTAT database	Too few countries covered
	Adults TVET attainment, %	Share of population (both genders) aged 25-64 years old who have attained upper secondary vocational training education.	OECD	Conceptual miss-match
	PIAAC scores	Measure of adults' proficiency in literacy, numeracy and problem solving.	OECD	Too few countries covered
	Number of developer accounts on GitHub, per million pop.	Number of developer accounts on GitHub in a given economy. This count excludes users that are bots or otherwise flagged as “spammy” within internal systems. A three-year moving average is applied to the original data.	GitHub	Conceptual miss-match
9.Leadership	-	-	-	-
10.Intermediate services	-	-	-	-

Indicators time series, missing values and imputation

The diagnostics tool is computed for three periods 2016-2020, 2018-2022, 2020-2023. For each period, each indicator is computed as the moving average of data available within that period.

This approach has several advantages. First, it makes sure that data that are older than five years do not enter the calculation of the indicators, which allows to compare countries performances on a similar time horizon. Second, by averaging out yearly fluctuations, it reduces the effect of temporary swings. Third, by combining data from multiple years, moving averages reduce the incidence of missing values.

Missing values however remain an important empirical challenge for calculating composite indexes. This can be problematic because the presence of a missing values can bias the calculation of a country’s composite index score by implicitly imputing the value of the missing indicator(s) with the average of the non-missing variables in an element. To limit the issue, only indicators available for a large share of OECD countries are used.

Even after applying moving averages, missing values remain in the dataset. To address this issue, missing values are imputed. Whenever an indicator with sufficient explanatory power is available, missing values are imputed via pooled time series linear regressions. This allows to estimate most of the missing values present in the dataset. There are however a handful of data points for which it was not possible to use linear regressions to estimate missing values. In these few cases, missing values are imputed as the geometric mean of the non-missing values in each element.

Imputed values, even if based on well-defined linear models, must be used with parsimony to avoid producing element scores that are entirely built on estimated values. For this reason, for elements constituted by two variables, the element score is computed only if at least one true variable is available. For elements constituted by three of four variables, the element score is computed only if at least two true variables are available.

An alternative imputation technique is a machine-learning algorithm that estimates datapoints using information from all other indicators in the dataset. The non-parametric, entropy-based “MissForest” method has been tested (Stekhoven and Bühlmann, 2012[4]) and its results reported in a parallel working paper (Crotti et al., 2025 (forthcoming)[1]). However, the quality of the imputed values has been considered uneven across indicators, and the conservative approach of imputing using selected linear regression models only when good explanatory variables are available has been prioritised.

Indicator normalisation

The above indicators are aggregated into a composite index score for each entrepreneurial ecosystem element.

To aggregate indicators with different ranges and units on a composite score, all variables must be placed on a comparable scale. While multiple alternatives have been considered, all indicators in the entrepreneurial ecosystem diagnostics tool are normalised using a clipped min-max normalisation technique. This method allows to address potential distortions in transforming data from raw values to normalised scores due to the presence of outliers while maintaining the distribution of normalised scores consistent with the original distribution of the data.

The clipped min-max transformation is applied as follows. For each indicator, the minimum value is defined as the indicator’s cross-country mean minus two times the indicator’s standard deviation, and the maximum value is defined as the indicators cross-section mean plus two times the indicator’s standard deviation. Once these values are set, for each country i and indicator k, the normalised score is computed according to the formula:

${score}_{ik}=100\mathrm{*}\frac{{raw\_value}_{ik}-\mathrm{ }{min}_k}{{max}_k-{min}_k}$

where the raw value is the original raw data for a country, ${min}_k={mean}_k-\left(2*{stdev}_k\right)$ and ${min}_k={mean}_k-\left(2*{stdev}_k\right)$. Whenever a country’s raw value is below the minimum value, a normalised score of 0 is assigned, and whenever a country’s raw value is above the maximum value, a normalised score of 100 is assigned.

There are cases where the normalisation formula must be inverted. Whenever an indicator’s distribution is such that a high value corresponds to a negative outcome (e.g. taxation, debt), the normalisation is computed as:

${score}_{ik}=100\mathrm{*}\frac{{raw\_value}_{ik}-\mathrm{ }{max}_k}{{min}_k-{max}_k}$

In these cases, whenever a country’s raw value is below the minimum, a normalised score of 100 is assigned, and whenever a country’s raw value is above the maximum, a normalised score of 0 is assigned.

The normalised scores are computed with respect to 2020-2023 period’s minimum and maximum values for all time series instances. In other words, the minimum and maximum values for the 2020-2023 period are used also to compute normalised scores of variables in the 2016-2020 and 2018-2022 periods. This implies that the normalised scores of all indicators must be interpreted as performances relative to the data distribution of the 2020-2023 period.

This approach is motivated by the possibility of visualising progress over time. A potential drawback of this approach is that some indicators might evolve rapidly over time. In this case, 2020-2023 values can become outdated in future editions of the tool. For example, all countries tend to increase their mobile phones data over time, and 2020-2023 period’s minimum and maximum levels could become inadequate to compute future normalised scores. As a result, these values will be reassessed in future editions of this report. Another limitation is that missing values in the 2020-2023 period can cause the minimum and maximum of all years to be determined by the data distribution of the sample of countries for which data are available in this period. This is however a limitation that impacts few indicators, and the possibility of comparing performances over time by anchoring scores to one point in time has been prioritised.

Aggregation

There are multiple methods to aggregate single indicators into a composite index score. The most common methods include the arithmetic mean, the geometric mean, and the weighted average. The aggregate scores of all ten elements of the entrepreneurial ecosystem diagnostics tool are computed using the geometric mean of the underlying indicators’ normalised scores.

The main advantage of the geometric-mean method is that it does not compensate a low value on one indicator with a high value on another indicator. This is one of the main reasons why UNDP has produced the Human Development Index using geometric means (UNDP, 2015[5]). Limiting the possibility to compensate a low performance on one indicator with a higher performance on another is an attractive feature for an entrepreneurial ecosystems benchmarking where bottlenecks are expected to have adverse effects on the overall ecosystem. The fact that indicators are non perfect substitutes is a particularly binding condition for ecosystem analysis, which assumes that an incomplete ecosystem can hinder entrepreneurship even if other conditions are in place. For example, entrepreneurs may not respond to additional incubation services if there is limited venture capital availability in a country.

The use of geometric means shares common features with “penalty for bottlenecks (PFB) algorithms” (Acs, Rappai and Szerb, 2012[6]). PFB algorithms have been used in the literature to identify the weakest link in the ecosystem and reduce the index score by the presence of these weak links. Although less used today, the PFB idea that a low score in one area cannot be fully compensated by a high score in another area is valuable for constructing composite indices. The use of a geometric mean thus allows to penalise low scores, but to a lesser extent than PBF approaches. The arithmetic mean and weighted average aggregation have been considered but discarded. A full sensitivity analysis on the impact of the use of these different methods on countries’ scores is shown in a working paper (Crotti et al., 2025 (forthcoming)[1]).

The ten elements are not further aggregated into a single overall country-level entrepreneurial ecosystem index score. Although computing a single index may provide a comprehensible snapshot of the overall conditions for entrepreneurship in a country, such a score carries the risk of conveying an overly simplified depiction of the complexity of entrepreneurial ecosystems and may divert the attention from the elements.

References

[6] Acs, Z., G. Rappai and L. Szerb (2012), “Index-Building in a System of Interdependent Variables: The Penalty for Bottleneck”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.1945346.

[1] Crotti, R. et al. (2025 (forthcoming)), “Data and methods for entrepreneurial ecosystem”, OECD.

[2] Leendertse, J., M. Schrijvers and E. Stam (2022), “Measure Twice, Cut Once: Entrepreneurial Ecosystem Metrics”, Research Policy, Vol. 51/9, https://doi.org/10.1016/j.respol.2021.104336.

[3] Nardo, M. et al. (2005), “Handbook on Constructing Composite Indicators: Methodology and User Guide, OECD Statistics Working Papers”, Handbook on Constructing Composite Indicators: Methodology and User Guide.

[4] Stekhoven, D. and P. Bühlmann (2012), “Missforest-Non-parametric missing value imputation for mixed-type data”, Bioinformatics, Vol. 28/1, https://doi.org/10.1093/bioinformatics/btr597.

[5] UNDP (2015), Training Material for Producing National Human Development Reports.