3.1 Sample and the digital ecosystem pillars

The data used in this study to assess the digital ecosystem (DPE) in the American continent come from the Digital Platform Economy Index databases made available by the Global Entrepreneurship and Development Institute (www.thegedi.org) (^{Acs et al., 2021}).

The DPE scores-and DPE constituents-were computed for a sample of 18 countries for the year 2019, including both North America and Latin America and the Caribbean islands: Argentina, Brazil, Canada, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico, Panama, Paraguay, Peru, United States, and Uruguay.

Concerning the output set, as we indicated in Section 2 the DPE index includes four sub-indicators that represent different constituents of the digital ecosystem: digital infrastructure governance, digital user citizenship, digital multisided platforms, and digital technology entrepreneurship. Summary statistics for the four DPE sub-indicators as well as the 12 DPE pillars are presented in Table 2. For illustrative purposes, the full data set used in this study is presented in the Appendix 2. Also, Appendix 3 includes the data for the 12 pillars included in the DPE index.

At this point, an important consideration is in order. For the purposes of this study, notice that the 12 DPE pillars were grouped into the four DPE sub-indicators (outputs) which are introduced in equation (1) to obtain the digital ecosystem performance measure (DPEBOD) (y=y1,…,yk ɅK=4): y1: digital infrastructure governance, y2: digital user citizenship, y3: digital multisided platforms, and y4: digital technology entrepreneurship. Because the sample is relatively small (i.e., 18 observations), an efficiency model with 12 outputs would produce biased results that would become evident in a large number of artificially efficient countries (^{Cook et al., 2014}). Thus, the proposed BOD model with four outputs seeks to reduce the potential loss of discriminatory power of BOD models with large numbers of inputs and/or outputs, relative to the number of units.¹

3.2 Method: Benefit of the Doubt (BOD) weighting

We propose that the digital ecosystem is the result of the mutual interactions between various economic agents (i.e., users, businesses, digital platforms, and public administrations) functioning in local settings with different institutional and environmental conditions. By taking this relevant property of this ecosystem into account, our analysis seeks to evaluate the digital ecosystem around the globe as well as its strengths and weaknesses with the objective to identify potentially optimal ecosystem-enhancing policies.

To construct the composite indicator, we employ the benefit-of-the-doubt (BOD) model (e.g., ^{Cherchye et al., 2007}). Rooted in non-parametric data envelopment analysis (DEA) techniques (for a survey on DEA see, e.g., ^{Cooper et al. (2011)} ), the BOD weighting is among the methods suggested by the ^{OECD (2008)} for computing objective, non-arbitrary composite indicators (CIs). The BOD model, originally proposed by ^{Melyn and Moesen (1991)} and further developed by, among others, Cherchye et al. (2007), ^{Sahoo et al. (2017)} and ^{Karagiannis and Karagiannis (2018)} , is a special case of the input-oriented DEA model (^{Charnes et al., 1978}) with a single constant input (vector of 1s) (^{Lovell and Pastor, 1999}; ^{Karagiannis and Lovell, 2016}).

Formally, for the 18 countries (i = 1,…,N = 18), the BOD weighting model used in this study considers the 4 DPE sub-indicators (y= y ₁,…,y _k ɅK=4) and employs a set of endogenous weights (w) to compute the weighted average of the seven indicators (y) that maximize the DPE score. The following linear program solves the BOD weighting problem and computes the optimal DPE value for each county (i):

(1)

sujeta to:

Equation (1) computes for each country a vector of endogenous weights for the 4 DPE sub-indicators (w _k = w ₁ ,…,w ₄ ) that maximizes the value of the DPE. The optimal DPE performance value is bounded (DPE ^BOD ≤1). For efficient countries on the frontier DPE ^BOD =1, while for inefficient territories DPE ^BOD <1 and 1-DPE ^BOD is the degree of inefficiency (i.e., the output expansion required to reach the frontier). Two properties of the BOD model (equation (1)) are worth highlighting. First, weights are constrained to be non-negative, which makes the composite indicator a non-decreasing function of the output set. However, the non-negativity constraint on the weights allows for extreme scenarios that might render DPE estimations inaccurate (e.g., high number of artificially efficient countries). Thus, additional restrictions on the weights are needed in order to account for the relative importance of all DEEBOD sub-indicators. In this study, we added to equation (1) a ‘pie share’ restriction (^{Cherchye et al., 2007}): L _k <Σ ^K _k w _ik y _ik <U _k . This proportional restriction is attractive because pie shares (w _ik y _ik ) do not depend on measurement units and they directly show the individual contribution of each pie share to the DEEBOD, while allowing for weight heterogeneity within and between countries. In equation (1), Lk and Uk are the lower and upper bound for each pie share, respectively. The endogenous weights are country-specific and the sum of the re-scaled pie shares equals the DEEBOD score (equation (1)).

The second property deals with the relative importance-i.e., weights and limits (L _k and U _k )-of the DPE^BOD sub-indicators. Instead of employing homogeneous ex-post weights resulting from the optimization model in equation (1), the DPE^BOD weights were generated via the data driven method described in equation (1) in which the DPE^BOD score is maximized based on the configuration of the efficient benchmark countries as well as on the position of the focal country in the technology hyperplane shaped by the output set (i.e., the 4 DPE sub-indicators).

4.1 Analysis of the digital ecosystem worldwide

This section presents the results of the evaluation of the digital ecosystem based on the measures computed via equation (1). We first analyze the results of the efficiency analysis of the digital ecosystem (DPE ^BOD model: equation (1)). The second part of the section deals with the analysis of the configuration of the digital ecosystem among the sampled countries.

Overall, from Table 3 we observe that the relative efficiency of the digital ecosystem is 36.43% (average DPE ^BOD =0.3643). This finding indicates that American economies have a large room for improving their digital ecosystem. Specifically, countries in the continent can enhance the quality of their digital ecosystem 63.57% by investing-in terms of physical resources and reforms-in the different constituents of the ecosystem. Obviously, significant variations in the quality of the ecosystem are detected among the sampled countries.

The United States is the only country reporting an efficient digital ecosystem, in terms of the BOD estimations (DPE ^BOD =1). According to our BOD model (equation (1)), Canada ranks second in our digital ecosystem composite indicator (DPE ^BOD =0.9202).

The group of Latin American countries lags behind the USA and Canada. Also, for this group we observe that only four countries report an above-average DPE ^BOD score, namely: Chile (DPE ^BOD =0.4514), Uruguay (DPE ^BOD =0.4028), Costa Rica (DPE ^BOD =0.4014), and Brazil (DPE ^BOD =0.3661). On the other hand, Honduras (DPE ^BOD =0.1595) and Guatemala (DPE ^BOD =0.1686) are the countries with the poorest digital ecosystem level (DPE ^BOD ).

To analyze the configuration of the digital ecosystem among the sampled countries, we employ the weights computed via equation (1) (Table 3) to aid in the identification of the strategic patterns that characterize the digital ecosystem. The descriptive results are summarized in Table 4. Additionally, Figure 1 shows the configuration of the digital ecosystem for the benchmark (efficient) peer-i.e., the United States-and for the group of Latin American countries.

The United States, the efficient country, has a balanced digital ecosystem (Table 4 and Figure 1). The results of the BOD model (DPE^BOD in equation (1)) indicate that the digital ecosystem in this country is primarily driven by the sub-indicator related to ‘user citizenship’ (0.3160) and ‘technology infrastructure’ (0.3100). On contrary, ‘technology entrepreneurship’ (0.2710) and ‘multi-sided platforms’ (0.2860) are relatively less important factors shaping the country’s ecosystem.

Although the descriptive result in Figure 1 suggests a balanced ecosystem, we observe two distinctive patterns in the configuration of the digital ecosystem of the group of Latin American countries. On the one hand, for the group of countries with an above-average DPEBOD score-i.e., Chile, Uruguay, Costa Rica-the ‘technology infrastructure’ sub-indicator is the most relevant aspect of their digital ecosystem whereas ‘user citizenship’, the top priority DPE indicator in the USA, is the weakest aspect of the ecosystem in these countries (Table 4). Because public trust is a critical prerequisite to user participation in the digital economy, this group Latin American countries should emphasize improvements in ‘user citizenship’ if the promotion of quality improvements in the digital ecosystem is the desired goal. For example, the development and sustainability of the digital ecosystem requires clear rules in terms of protection of personal data and enhanced privacy terms in digital platforms. Both aspects-which are included in the ‘user citizenship’ pillar-should be targeted by policy makers from these countries if the promotion of quality improvements in the digital ecosystem is the desired goal.

On the other hand, the DPE sub-indicator related to ‘multi-sided platforms’ is the most important factor of the digital ecosystem among the majority of countries with a below-average DPEBOD score, namely: Ecuador, Colombia, Mexico, Panama, Paraguay, and Peru (Table 4). Despite the reported similarity in the policy priority in these countries, we observe marked differences in what concerns their least relevant DPE factor.

Figure 1: Configuration of the digital ecosystem: United States vs. Latin America 

4.2 The connection between the digital ecosystem and countries’ economic performance

The supplementary analysis presented in this section explores the relationship between economic performance and the quality of the digital ecosystem among the sampled countries. Specifically, we assess if the quality of the digital ecosystem, measured via the BOD efficiency scores (equation (1)), correlates with countries’ economic results, measured as the GDP per capita. Notice that the GDP data was obtained from the World Bank databases and is expressed at 2011 prices in PPP international dollars. Summary descriptives for the two variables are presented in Table 5, while Figure 2 offers a visual representation of the connection between the DPEBOD score (equation (1)) and GDP per capita.

A first descriptive result indicates that there is a close connection between economic development and the digital ecosystem: Pearson correlation coefficient = 0.8732 (p-value < 0.000). The findings presented in Figure 2 further verify the reported strong correlation between the digital ecosystem and economic development. Notice that this result does not imply a causal relationship. We simply refer to the strong connection between economic development and the digital ecosystem among the sampled American economies.

Second, by observing the specific position of any given country below or above the computed trend line in Figure 2 we can offer a second descriptive analysis that is appropriate for comparison purposes. For example, the United States is the only efficient country in the American continent (DPE^BOD = 1.00), while Canada reports the second highest value for the digital ecosystem score (DPE^BOD = 0.92). But, both countries have a different position in the scatter plot presented in Figure 2. In the case of the USA, the descriptive result indicates a balanced configuration between economic development and the quality of the digital ecosystem. For Canada, the descriptive finding in Table 5 suggests that the country’s digital ecosystem is potentially above the mean level expected for developed economies with a similar economic development level.

Finally, for the two countries with the poorer results (Honduras and Guatemala) the result in Figure 2 indicates that the quality of the digital ecosystem is in accordance with the economic development level of these countries.

Figure 2: Connection between the digital ecosystem (DPEBOD) and GDP per capita 

The objective of this supplementary analysis was to show that there is a connection between economic development and the quality of the digital ecosystem, measured via the efficiency score proposed in this study (DPEBOD in equation(1)). The core finding of this descriptive exercise corroborates this intuition: the quality and configuration of the digital ecosystem can be seen as a consequence of specific policies that match the countries’ economic development level.