Policy Significance Statement

The 2020 Communication “Shaping Europe’s digital future” of the European Commission puts forward as a key action the development of a “reinforced European Union (EU) governments interoperability strategy.” This article contributes to the development of the strategy by proposing two policy recommendations. First, efforts to improve e-Government services across the EU should be intensified to support the overarching digital agenda of the EU and increase benefits for European citizens. Second, interoperability should become a central element when designing new digital public services. In shaping a new interoperability strategy for the public sector, the Commission could further enhance a common approach to interoperability across the EU’s public sector by strengthening the governance of interoperability initiatives and encouraging the adoption of specific interoperability requirements.

2. Literature and Policy Context

3. Hypothesis Development

As discussed in Section 2, several models have been considered to assess the impacts of e-Government and interoperability. Building on previous research, this article devises a theoretical framework to assess one of the main impacts of the EIF on citizens. In a nutshell, by relying on publicly available data and information, the article analyses whether and how NIFs implementing the EIF affected the way citizens interact with the public administrations online in the country where they live. In addition, the article also assesses the role that e-Government services play in shaping the interactions between citizens and public administrations. The analysis thus tests two main hypotheses:

1. i) The introduction of NIFs that follow the EIF model facilitates the online interaction between citizens and public administrations;

2. ii) Better e-Government services encourage citizens to interact online with public administrations.

More specifically, the analysis measures the impact of the adoption of NIFs (represented by an independent, dummy variable showing whether a country has a NIF or similar documents implementing the EIF in place or not) and the development of e-Government services (represented by another independent variable capturing the quality of e-Government) on the way citizens interact with public administrations online (represented by a composite indicator).

The article contributes to the literature on interoperability and e-Government by proposing a novel theoretical framework, based on econometric methods, to test the two hypotheses. Testing hypothesis (i) is the first attempt to quantify econometrically the impacts on citizens of strategic actions coordinated across the EU to enhance the interoperability of digital public services. Testing hypothesis (ii) allows to estimate the impact on citizens of better e-Government services and to move from a descriptive analysis of the field (e.g., how many public services are available online) to an analysis of outcomes (whether such services are used), thus providing a richer perspective on the progress made in the field of e-Government. While both hypothesis (especially the second one) may be intuitive, the article provides a concrete basis for estimating the impact of existing EU and national initiatives in the fields of interoperability and the provision of digital public services and for shaping better policies in the future.

With respect to hypothesis (i), it could be argued that countries with a higher level of use of online public services are more likely to adopt NIFs. Nonetheless, this reverse causality is not observed. For instance, among the 10 leading countries in terms of the online interaction between citizens and public administrations between 2012 and 2019, four countries did not adopt a NIF and only Estonia has fully implemented a NIF since the beginning of the period under investigation (see Table 1). Based on the sample of countries analyzed, it cannot be concluded that countries with higher levels of use of online public services adopt NIFs.

Table 1.

Top 10 countries in terms of online interactions between citizens and public administrations over 2012–2019 (analysis limited to the countries analyzed in this article)

Note: Countries are ranked based on the average values of the online public services use index (see Table 2) over the period 2012–2019.

Source: Authors’ elaboration on Eurostat and the information presented in the Annex concerning the adoption of NIFs.

4. Data

The overall analysis is based on country-level data from 26 European countries (22 EU Member States, Iceland, Norway, Switzerland, and the UK)Footnote ⁷ collected over the period 2012–2019.

The dependent variable, that is online public services use, reflects the citizens’ use of online public services by relying on four different variables measured by Eurostat. More specifically, the dependent variable is computed as the simple average of the percentages of individuals that in each country: (a) interact with public authorities via the Internet; (b) obtain information from public authorities’ websites; (c) download official forms; and (d) submit completed forms to public administrations.

The first hypothesis presented in Section 3 is tested through a dummy variable, which represents one of the contributions of this work, as it was built by carefully examining official documents of the countries in the scope of the analysis, capturing strategic guidance on digitalization and interoperability in the public sector. The construction of the dummy variable is detailed in the Annex. The analysis differentiates between a complete adoption and an incomplete adoption of the EIF. In particular, two variables are considered:

• • NIF complete, which encompasses countries that have adopted NIFs that fully reflect the features of the EIF;

• • NIF complete and incomplete, which encompasses both countries that have adopted NIFs fully reflecting the EIF, as well as countries that have implemented only some parts of the EIF in their NIFs (or equivalent documents).

The second hypothesis is tested through the development of e-Government services variable, which is computed by relying on indicators included in the so-called “e-Government benchmark.”Footnote ⁸ This variable assesses the availability and delivery of online public services according to the exhaustiveness and relevance of the information provided online by public administrations as well as the range of services provided. This approach in line with UNDP (2014, p. 14), which emphasized the need to shift the focus from “e-Government maturity models” to the deployment of “a portfolio of e-services” when analyzing e-Government progress.

It is worth emphasizing that the analysis focuses on effects within countries, capturing the impact of NIFs and the development of e-Government services on how citizens interact with the public administrations in the countries where they live. The cross-border dimension of public services is taken, however, into account as part of the development of e-Government services variable, which is composed of a series of indicators including an indicator on public services provided to foreign citizens.

When exploring the nexus between e-Government development and citizens’ use of digital services, the limited empirical literature identifies several factors that need to be controlled for in order to properly isolate the effects of the independent variables. Zhao et al. (Reference Zhao, Wallis and Singh2014) and Lakka et al. (Reference Lakka, Stamati, Michalakelis and Anagnostopoulos2015) account for the available communication infrastructure to capture the extent to which citizens have access to communication lines that may allow them to interact with public administrations. Similarly, our model factors in the availability of infrastructure to access the Internet in each country. Furthermore, the model used in this article also includes general public services expenditure, a proxy that controls for all the effects that can arise from discrepancies between public spending in different countries for developing innovative public services. Finally, the influence of contextual factors, such as economic, cultural, political or educational ones, may also play a role in the relationship under investigation (Zhao et al., Reference Zhao, Wallis and Singh2014). Against this background, the model also accounts for:

• • Economic factors, by focusing on the real gross domestic product (GDP) per capita, which allows to control for any differences that may stem from the economic development of the countries under investigation (see, for instance, Taipale, Reference Taipale2013, who studies the role of sociodemographic, economic and geographical predictors on the e-Government use in Finland by relying on household income);

• • Educational factors, by including the national level of IT skills (in the same vein, to control for this factor, Lakka et al., Reference Lakka, Stamati, Michalakelis and Martakos2013 and Taipale, Reference Taipale2013, consider the level of education and education expenditures, respectively);

• • Political factors through a government effectiveness variable that captures the perception of the quality of public services, the quality and independence of the civil service, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies (Lakka et al., Reference Lakka, Stamati, Michalakelis and Martakos2013, rely on the same variable).

Table 2 summarizes the different variables included in the analysis, specifies the unit used and provides the sources. Table 3 presents descriptive statistics of the variables used in the econometric analysis. In the countries under analysis, on average, around 47% of the population aged 16–74 use the Internet to interact with public administrations. Substantial differences exist, however, among countries. For instance, Finland records the highest usage rate (80%) in 2019. In this context, Northern European countries (Sweden, Iceland, Denmark, and Finland) also record very high digital literacy scores and Internet use rates. On the other side of the spectrum, only 17% of Italians aged 16–74 used online public services in 2012, with very slight improvements over the study period, as the maximum recorded user rate in Italy reached 20% in 2019. With the highest value for the e-Government index in 2016, the Netherlands is a front-runner in the field. By contrast, Slovakia registers the lowest value for this index (22.5). When it comes to NIFs, almost half of the countries under analysis have adopted at some point in time a NIF that fully or partially reflect the features of the EIF (the full list is available in the Annex). When it comes to the control variables, the countries in the sample are quite heterogenous. While several countries (e.g., Austria, France, Latvia, and Luxembourg at the beginning of the period, and Iceland and Ireland over the entire period) invested little or nothing in research and development (R&D) related to general public services, Greece made considerable efforts over the whole period, which resulted in both a significant improvement in the quality of online public services and a greater use of these services. Overall, all the countries under investigation rank quite high in terms of government effectiveness, although Greece, Hungary, and Italy display lower scores than the other countries.

Table 2.

Description of the variables used in the econometric analysis

^a Data related to business mobility are missing for 2012 and 2016.

^b User centricity reflects the extent to which (information about) a public service is provided online, how the online journey is supported and if public websites are mobile friendly. Transparency evaluates the transparency of government authorities’ operations, service delivery procedures as well as consultation of personal data by public administrations. Citizen mobility mirrors the extent to which public services that aim at foreign citizens are available online, usable, and implement electronic identification and eDocument capabilities. In the same vein, business mobility assesses the extent to which public services that aim at foreign businesses are available online, usable, and implement electronic identification and eDocument capabilities. Finally, key enablers estimate the extent to which technical pre-conditions for e-Government service provision are used. The key enablers used for measuring the quality of the services to businesses and citizens are: (a) electronic identification; (b) electronic documents (eDocuments); (c) authentic sources; (d) digital post; (e) eSafe; and (f) single sign on.

Source: Authors’ elaboration.

Table 3.

Descriptive statistics of the variables

Source: Authors’ elaboration.

5. Statistical Analysis

To test the two hypotheses presented above, a panel data analysis covering 26 European countries over the period 2012–2019 is performed. The set of variables considered is, therefore, characterized by both the spatial and time dimensions, that is, $ i=26 $ and $ t=8 $ , respectively. The analysis assesses the ceteris paribus effects of each of the independent variables, described in Table 2 and measured at the country level, on the online interaction between citizens and public administrations. The corresponding econometric model is described by the following equation:

(1) $$ \mathrm{Index}\_{\mathrm{PA}}_{i,t}={\displaystyle \begin{array}{l}c+{\beta}_1\times \mathrm{Index}\_{\mathrm{egov}}_{i,t}+{\beta}_2\times {\mathrm{NIF}}_{i,t}+{\beta}_3\times \mathrm{l}\_\mathrm{gen}\_{\mathrm{pub}}_{i,t}+{\beta}_4\\ {}\times \hskip0.3em \mathrm{l}\_\mathrm{GDP}\_{\mathrm{cap}}_{i,t}+{\beta}_5\times \mathrm{gov}\_{\mathrm{eff}}_{i,t}+{\beta}_6\times {\mathrm{infra}}_{i,t}+{\beta}_7\times \mathrm{IT}\_{\mathrm{skills}}_{i,t}+{u}_i\\ {}+\hskip2px {\varepsilon}_{i,t},\end{array}} $$

where $ c $ is the intercept, $ {\sum}_{j=1}^7{\beta}_j $ are the coefficients associated to each explanatory variable accounted for in the model, $ {u}_i $ defines the fixed effects per country, and $ {\varepsilon}_{i,t} $ is the error term. $ \mathrm{Index}\_{\mathrm{PA}}_{i,t} $ is the dependent variable.

Before carrying out the analysis, it is crucial to ascertain that the independent variables (i.e., the explanatory variables) are not highly correlated with each other. The simultaneous introduction of two highly correlated variables in the model could lead to erratic results due to collinearity. To avoid this issue, in principle the correlations should be lower than 0.8.Footnote ⁹ In this context, Table 4 presents the pairwise correlations between the independent variables used in the analysis. Based on the model given by equation (1), only the positive correlations between the government effectiveness and the real GDP per capita could alter the robustness of the estimates as the correlation displays a value very close to 0.8.Footnote ¹⁰ Therefore, to ensure the robustness of the results, regressions are performed by dropping alternatively the government effectiveness and real GDP per capita (in logarithm) variables from the model.

Table 4.

Correlation matrix

Notes: All significance levels are at the p > .05, except where denoted by: *p < .05.

Source: Authors’ elaboration.

Specification tests are then conducted in order to apply the model that best fits the data. In panel data models, the question of the treatment of the fixed effects ( $ {u}_i $ ) often arises; in fact, it is important to determine whether to treat individual effects as constant over time or as a random variable. The Hausmann specification test between random and fixed effects is run with NIF_comp + incomp ( $ {\chi}^2(7)=14.65,p=0.04 $ ) and NIF_comp ( $ {\chi}^2(7)=14.66,p=0.04 $ ) and indicates that fixed effects are the best analytical approach. The Breush and Pagan Lagrangian multiplier (LM) test for random effects supports the use of fixed effects since the variance of individual specific effects ( $ {u}_i $ ) is significantly different from 0 ( $ {\chi}^2(1)=246.98,p=0 $ under the null hypothesis $ Var\left({u}_i\right)=0 $ ). The model, therefore, controls for country-level variables that are difficult to measure or observe and that do not vary over time (such as some types of cultural factors). Fixed effects isolate the impacts of time-invariant characteristics so that it is possible to assess the net effect of the explanatory variables on the selected dependent variable. In other words, the proposed model estimates an additional and single effect ( $ {u}_i $ ) for each country, which is assumed to be constant in the entire period of observation (2012–2019).

As panel models often violate the assumptions underlying the standard ordinary least square (OLS) regressions, two post-estimation tests for serial autocorrelation and heteroscedasticity are performed. The Wooldridge test for autocorrelation in panel data shows evidence of serial autocorrelation at the first order ( $ F\left(1,25\right)=5.12,p=0.03 $ ). The Modified Wald test for groupwise heteroscedasticity in fixed effect regression model indicates that errors are heteroskedastic ( $ {\chi}^2(28)=1150.44,p=0 $ ). Based on these results, the analysis is adjusted to correct biases that may result from serial autocorrelation and heteroscedasticity.

To increase the robustness of the results, the hypotheses are therefore tested using different model specifications:

1. i. First, the variable of interest, NIF_comp and NIF_comp + incomp are tested using a lag of 1 year. While the main model tests for a contemporaneous effect of the NIF on the online public services, the lagged model account for possible delays between the moment when the NIF is adopted and the moment when this translates in more online interaction between citizens and public administrations.

2. ii. Second, a model accounting for an interaction variable between the e-Government index and NIF_comp and NIF_comp + incomp, alternatively, is included in the analysis. This model allows us to test whether the effect of the e-Government index on the online interaction between citizens and public administrations depends on the adoption of NIFs based on the EIF and vice versa. If interaction effects can be detected, this means that the two variables combined have a greater effect than the sum of their individual effects.

6. Results and Discussion

By relying on the above-mentioned specifications, the results of the analysis support the two main hypotheses. Table 5 presents the results of the first model specifications (i.e., equation (1) which presents the main model with contemporaneous effects and the model in which NIF_comp and NIF_comp + incomp are lagged). Table 6 displays the results when the regression introduces interaction effects between the NIF and the e-Government index. The results associated with the NIF_comp variable are presented in (1) and (2) and those associated with NIF_comp + incomp are shown in (3) and (4). In addition, to address the problem of correlation presented in Table 4, the results when including the government effectiveness variable are described in (1) and (3) and those when including the real GDP per capita (in logarithm) variable are in (2) and (4). Several findings of the analysis are worth commenting as they can have substantial policy implications.

Table 5.

Regression results with contemporaneous and lagged effects

Notes: Robust standard errors are in parentheses.

^*** p < .01;

** p < .05;

* p < .1.

Source: Authors’ elaboration.

Table 6.

Regression results with interaction effects

Note: Robust standard errors are in parentheses.

^*** p < .01;

^** p < .05;

* p < .1.

Source: Authors’ elaboration.

Focusing on the first hypothesis (the introduction of NIFs facilitates the online interaction between citizens and public administrations):

• • On average and all other variables being equal, countries that have at least partially implemented the EIF (NIF_comp + incomp) register a higher level of citizens’ use of the Internet to interact with their public administrations compared to countries that have not adopted the EIF at all. The associated variable (NIF_comp + incomp) is significant at the 1, 5, or 10% level across all model specifications.

• • The dummy variable identifying countries that have implemented the EIF in full (NIF_comp), however, is significant only under one model specification at the 10% level (namely the second specification of the contemporaneous effects model). Given that only one specification led to significant results, it cannot be concluded that there is an overall effect associated with this variable. Therefore, considering the results associated with both dummy variables of interest (NIF_comp + incomp and NIF_comp), an effect from the partial implementation of the EIF can be confirmed through this analysis, but no effects can be confirmed in the case of the full implementation of the EIF based on the data and models available.

• • The key takeaway is that the true difference is not necessarily made only by the full implementation of the EIF. The main effect seems to come from the adoption of a well-crafted set of policies to foster interoperability in the public sector, which build to different extents on the EIF, and which also account for national specificities in their aim to enhance interoperability.

Moving to the second hypothesis (better e-Government services encourage citizens to interact online with public administrations):

• • All models (with the exception of one specification—namely specification (2) of the lagged-effects model) confirm that the availability and delivery of online public services (as captured by the index_egov variable) has a positive impact on citizens’ use of the Internet to interact with public administrations.

• • A 10% increase in the e-Government index, with all other variables being equal, leads to an increase of approximately 1–2% in citizens’ use of the Internet to interact with public authorities.

When it comes to the effect of the control variables, all model specifications indicate that higher levels of GDP per capita have a positive impact on the online interactions between citizens and public administrations. The coefficient should be interpreted as a semi-elasticity, that is, a 1% increase in GDP per capita leads, all else being equal, to an increase of about 0.3% in online interactions between citizens and the public administration services.

Finally, the model featuring interaction effects presented in Table 6 does not display a significant interaction term. In other words, the variable capturing the adoption of NIFs based on the EIF does not seem to impact the relationship between the e-Government index variable and citizens-government online interactions. Similarly, the level of the e-Government index does not seem to have an impact on the relationship between the variable capturing the adoption of NIFs implementing the EIF and the online interaction between citizens and public administrations. As reported in the key results related to the independent variables, the only interesting results lie in the individual effects of each variable, namely NIF_comp + incomp or NIF_comp and index_egov, on the dependent variable index_PA.

7. Limitations

The model and the analysis presented in this article test and validate the effect of the EIF and the general development of eGovernment services on how citizens interact online with their public administrations. There are, however, some limitations to the analysis that need to be discussed.

First, the analysis captures effects stemming from actions taken at the strategic level of interoperability. The guidance provided by the EIF is valid at the governance and strategic levels, aiming to increase coordination across the EU and improve the way EU countries design interoperable public services. Therefore, the model does not capture, at least directly, the effects stemming from the practical implementation of the interoperability requirements included in the EIF. Further research could consider analyzing the implementation of specific interoperability requirements to understand which concrete measures may yield more results. Further analysis on the effects stemming from certain interoperability requirements can also lead to more granular policy recommendations, thus also contributing to the ongoing debate on interoperability as a tool to foster competition in digital services.

Second, cultural factors are not directly accounted for in the analysis through dedicated explanatory variables as such factors are difficult to capture and measure. Cultural factors, that are specific to individual countries, may affect the way public services are developed, the extent to which they are interoperable (for instance, cultural factors may include resistance to technological change, Navarrete et al., Reference Navarrete, Gil-Garcia, Mellouli, Pardo and Scholl2010) and the citizens’ propensity to use digital public services. Nevertheless, the choice of the model mitigates this issue. The fixed effects model isolates the impact of factors that are country specific and that do not vary over time. Cultural factors are expected to be time-invariant over the limited period of analysis (2012–2019).

Finally, while time-invariant effects that are specific to individual countries are captured in the model via country fixed effects, additional time-varying effects that are not captured by the control variables are not included in the model. There are two main considerations that motivate the decision not to include time-fixed effects in the analysis. First, including additional dummy variables in the model to account for time-fixed effects could yield problems of collinearity.Footnote ¹¹ Second, the relatively short time period considered in the analysis could also be a limitation, meaning that only a limited set of data could be integrated in the analysis based on available sources. Over time, as more data become available, further research could rely on a longer timeframe and could apply a parametric approach to approximate time-fixed effects in case time dummy variables are hardly applicable as suggested by Gösser and Moshgbar (Reference Gösser and Moshgbar2020).

In spite of the limitations listed in this section, this article contributes to the academic and policy debates on the role of interoperability and its impact on digital innovation. In line with Hodapp and Hanelt (Reference Hodapp and Hanelt2022), it provides a more comprehensive perspective on interoperability that is not restricted only to the technical aspects but acknowledges the sociotechnical nature of interoperability. In addition, it also contributes to wider debates on the adoption of open source solutions as open source software can play a significant role in the public sector and in the development of digital public services (Spagnoletti and Federici, Reference Spagnoletti and Federici2011).