Theoretical aspects

Technological progress plays a crucial role in both exogenous and endogenous growth models. However, while the latter develops explanations in this respect, the former assumes that the rate of progress is determined outside the model because their assumptions prevent the economy from allocating resources to finance technological progress. In particular, exogenous growth models assume a production function with constant returns to scale and perfect competition in the different markets so that the payment of each factor is equal to its marginal product. Using Euler's theorem, this implies that the total product of the economy is equal to the sum of the quantity of factors multiplied by their respective marginal productivities. Thus, there are no resources available to finance technological progress.

Broadly speaking, the link between ICT and economic growth can be seen in efficiency gains and productivity improvements by economic agents (^{Nairet al., 2020}). A simple way of incorporating the issue into the framework of growth models is presented by ^{Barro and Sala-i-Martin (2004)}, who use a simple leader-follower model to analyze the effect of innovation and technological imitation on the rate of economic growth. The importance of this model lies in the fact that it can be interpreted in a novel way to reveal how ICT penetration can drive economic growth in both the economy leading the innovation as well as in the follower economies, thus providing a useful point of reference and starting point for analyzing the phenomenon and obtaining policy recommendations.¹In this model, the growth of the leading economy (economy 1) is driven by its innovations, while the growth of the follower economy (economy 2) depends on imitating the innovations made in the leading economy.

According to ^{Barro and Sala-i-Martin's model (2004, p. 296)}, the growth rate of the leading economy can be written as follows:

whereθ> 0 and ρ > 0 since they are parameters related to household preferences. Meanwhile, 0 < α > 1 indicates the elasticity of the marginal product of the intermediate goods. An important assumption made by the authors is that these three parameters are similar for both economies (the leader and the follower). Furthermore,A₁ is a productivity parameter representing the quality of governance and the level of technology.L₁ is the labor endowment, andηis the unit cost of inventing a new variety.

With respect to the follower company, the growth rate will be:

WhereN₁ is the number of varieties of intermediate products developed in the leading economy andN₂ is the number of products introduced in the follower economy. At all times,N₂ ≤N₁. In turn, μ is a positive parameter that determines the convergence speed, and (N₂/N₁)* is a ratio that indicates when the two countries are in a steady state.

Several conclusions can be drawn given equations (1) and (2). First, the growth rate of the leading economy can be increased by reducing the cost of invention (η), by increasing the labor forceL₁ (to which ICT can directly contribute), or by increasing the productivity parameterA₁, the latter through actions that promote "e-governance, accountability, public-private partnerships, and learning from best practices around the world, ICT improves the quality of governance and thus productivity" (^{Vu, 2011, p. 360}).

Finally, the follower economy will have a growth rate that depends fundamentally on the growth rate of the leading economy (γ₁) and onμ. This last result is relevant when considering a conceptual breakdown of the current technological transformations and their impact on growth. On the one hand, there is Industry 4.0, which refers to a form of production based on the incorporation of 4.0 technologies, focusing mainly on automation, robotization of processes and real-time data; on the other hand, there is the digital services economy based mainly on the use of platforms (e-commerce). Some authors, such as ^{Capello and Lenzi (2023)}, show that the adoption of Industry 4.0 technology and automation could be associated with regional economic growth in the places where the transformation is taking place, while the effects of digitization more generally spread and diffuse across all regions, without giving those regions where the transformation of the digital services economy prevails a significant growth advantage over others.

Literature review

In general, the existing empirical literature regarding ICT, productivity and economic growth has been grouped according to different criteria (^{Fernández-Portilloet al., 2020}; ^{Vuet al., 2020}): effects of ICT on economic growth and/or total factor productivity (TFP) -including causality tests-, transmission channels, and differential impact of ICT, depending on the type of country.

However, it is also possible to find authors who divide the studies into two categories: those that examine the relationship between ICT and TFP -here making a subdivision between country-level studies and company-level studies- and those that examine the relationship between ICT and labor productivity (^{Shahnazi, 2021}).

In the case of Latin American countries, the empirical literature focuses on the analysis of individual cases at the country or firm level, except the papers by ^{Hofmanet al.(2016}) and ^{Quiroga-Parraet al.(2017)}.

^{Hofmanet al.(2016)} analyze the impact of ICT on economic growth in five Latin American countries (Argentina, Brazil, Chile, Colombia and Mexico) from 1990-2013. The authors are also interested in explaining the gap in GDP per capita relative to the US economy. They find that the main reason for the persistence of the labor productivity gap is the growing ICT gap, which is counteracting improvements in human capital in Latin America.

^{Quiroga-Parraet al.(2017)} conducted a comparative analysis of ICT use between six Latin American countries (Argentina, Brazil, Chile, Colombia, Peru and Mexico) and seven developed countries. The results show the significant technology, knowledge, productivity and innovation gaps in the region; the authors indicate this as the first evidence that ICT is a causal reason for the low productivity and quality of life in Latin America. Although they do not develop an econometric model, they carry out the main comparisons based on correlations for different measures of ICT access and use.

In contrast to these two papers, this paper considers a more recent and extended period when the Internet is no longer an emerging technology and a larger number of cross-sections at the country group and aggregate economy levels. In addition, the panel data methodology is helpful for international comparisons with other papers.

At the company level and for the Ecuadorian case, ^{Arévalo-Avecillaset al.(2017)} conducted a correlational study in which they reached three main conclusions. On the one hand, there is an evident influence between IT investment and the productivity of organizations. On the other hand, the adoption of a website shows a significant and positive relationship based on the performance of products and services, and finally, there is no relationship whatsoever with the performance of business processes. To conclude, the use of e-commerce does not show a significant relationship with productivity.

For the Mexican economy, ^{Figueroa Hernándezet al.(2021)} studied the impact of the Internet on economic growth for the period 1990-2018. Using a single-equation model estimated by Ordinary Least Squares (OLS), they find a positive effect of the variables: number of Internet users and fixed telephone subscriptions on growth.

Using a quadratic model, for the case of Nicaragua, ^{Brenes-Gonzalez (2023)} estimates the semi-elasticities of the percentage of households using the Internet on the economic growth of that economy. The estimated coefficient of the linear variable is 0.0373.

In the case of other economies, recent empirical contributions are generally more productive, perhaps due to the greater availability and specificity of data. In this respect, ^{Gopane (2020)} examines the impact of digitization on labor productivity using econometric modeling based on an endogenous growth model for the case of the so-called BRICS bloc (Brazil, Russia, India, China and South Africa). The study's findings confirm the new productivity paradox, as the author concludes that accelerated digitalization does not manifest itself in productivity growth.

For the case of the Organization for Economic Cooperation and Development (OECD) countries, ^{Kurniawati (2021)} works with a sample of 24 countries for the period 2000-2018. The methodology used is based on panel cointegration and vector error correction models. The results above show that innovation development is highly significant in promoting economic growth in the later years of his sample.

Previous studies for European or OECD countries are those of ^{Nairet al.(2020}) and ^{Pradhanet al.(2017b}, ²⁰¹⁸, ^2019a, and ^2019b). In all cases, using different methodologies, time periods, and country groups, they found positive effects of ICT on economic growth.

^{Adeleyeet al.(2022)}, for example, test the so-called leapfrog hypothesis for eight economies in the South Asian Association for Regional Cooperation (SAARC) from 2000 to 2020. This hypothesis is tested using unbalanced panel data. The key variables are real GDP per capita and four ICT indicators (cell phones, fixed telephones, fixed broadband and Internet users). The primary estimation method is the Quantile Method of Moments Regression (MM-QR). The results show that ICT (individual indicators and composite index) has a statistically significant positive effect on economic growth. Likewise, the leapfrog hypothesis is valid for cell phones and composite index models and only for the lower quantiles in the case of fixed broadband.

^{Odhiambo (2022)} suggests studying the relationship between ICT, income distribution and economic growth in the case of Sub-Saharan African countries for the period 2004-2014. The estimation technique used is the Generalized Method of Moments (GMM). The author finds evidence that, in general, an increase in ICT development leads unconditionally to increased economic growth in the countries studied.

^{Lyuet al.(2023)} recently found evidence of a positive relationship between the digital economy and ecological TFP in the case of the Chinese economy. Their principal findings show that the digital economy has, on the one hand, a positive direct impact and, on the other hand, a spatial spillover effect on green U-shaped TFP. The origin of these effects is mainly related to the role of the digital economy as an agent for promoting the progress of green technology. Furthermore, the reported results of the heterogeneity analysis show that the digital economy is the fundamental factor for cities whose economies are based on natural resources, thus breaking the so-called natural resource curse.

^{Haldaret al.(2023)} worked with a sample of 16 emerging markets from 2000-2018. Using GMM estimates with instrumental variable and fixed effect regressions, they find that ICT not only monotonically increases economic growth but also exerts an effect on growth by increasing the efficiency of financial development. However, the authors caution that ICT accentuates the adverse effects of trade on economic growth.

The literature on the relationship between ICT vs. labor productivity is extensive. In general, a positive effect of ICT use on labor productivity is documented, even for those with lower levels of education (^{Leeet al., 2020}).

Moreover, it emerges as another important issue related to the question of the sensitivity of jobs to computerization and the widespread diffusion of robotics in different sectors (^{Lorenzet al., 2023}; ^{Acemoglu and Restrepo, 2020}; ^{Frey and Osborne, 2017}). In this regard, it is important to note that "ICT can affect cost reduction by saving labor and capital. This can affect the flexibility of the production process and induce increasing returns to scale" (^{Shahnazi, 2021, p. 345}) (see Figure 1).

Source: prepared by the author, adapted (edited and completed) from ^{Shahnazi (2021)}.

Figure 1 Effects of ICT on productivity 

Finally, Table 1 provides a synthesis of empirical works regarding the impacts of ICT on productivity and growth:

Descriptive analysis

Table 3 shows the main statistics of the variables involved in the analysis.

As can be seen, there is a wide discrepancy in the PHBA variable, which reflects the marked difference between countries in terms of PHBA. In this regard, while countries such as Argentina and Chile have values between 85 and 90% for the last year of the sample, others such as Guatemala, Paraguay and Peru have values below 40%.

On the other hand, the TFP has not developed uniformly across economies (see Figure 2). Although a positive trend can be observed in some of them, the trend is clearly downward in the case of Brazil and Mexico.

Source: prepared by the author

Figure 2 Evolution of the TFP 

Estimates

Table 4 summarizes the main estimation results. The estimation was performed using MCO and robust standard errors based on the use of a White matrix, which resolves the dependence between different cross sections.

First of all, it can be observed that all the variables have the signs expecteda priori, according to the theoretical level, and at the same time, they show statistical significance. The coefficient of the variable PHBA shows that a 1% change in it is associated with a 0.064% change in the dependent variable (TFP). This indicates that the contribution, although in the direction suggested by the theory, has a strong inelasticity, implying that the increase in the percentage of households with broadband has a modest impact on productivity improvements. This value is very close to that reported by ^{Niebel (2018)}, who, although he calculates the elasticity of economic growth with respect to ICT capital, could consider these measurements to be close due to the significant impact of TFP on growth. The sample used by the author consists of countries with different levels of development. The overall value estimated in this study is 0.089 for the global sample; however, a more exhaustive analysis allows him to present the results according to groups of countries. He concludes that these values are 0.066, 0.059 and 0.084 for developing, emerging and developed countries, respectively. In the same direction, ^{Mashadihasanli and Zülfikar (2023)} obtain a similar value for the elasticity of economic growth with respect to the number of Internet users per hundred people. In particular, using a GMM estimation, they conclude that the value is 0.0477 for a sample of 35 countries (developed and developing) and for the period 2001-2021.

Moreover, all the control variables also have coefficients with the expected signs and some degree of statistical significance. For the first of these, FBC is used as a proxy for investment, and the result is also inelastic, 0.205, meaning that a 1% change in investment implies a 0.205% change in TFP. This result is reasonable and makes sense, among other factors, because of the law of diminishing marginal returns.

Something similar occurs in relation to the R&D expenditure variable, which has a coefficient value of 0.03. However, this variable was only significant at 10%. The reported value is in line with estimates for developed countries. In this respect, ^{Piekkola (2022)} estimates the elasticity of output with respect to R&D expenditure for Finland to be between 0.036 and 0.039.

Concerning the institutional control variable RG, it is also significant at 5%, and it should be remembered that in this case, its coefficient is semi-elastic as it enters the model at the level of assuming values between -2.5 and 2.5. The results, therefore, show a positive contribution of the RG to the TFP. This means that a 1% improvement in RG has an approximate impact of 0.0052% on the change in TFP.

In terms of Figure 1, the control variables FBC and RG positively impact productivity, FBC via inputs and RG via the process of obtaining outputs. In this respect, ^{Shahnazi (2021)} argues that ICT can affect the process through cost reduction by reducing or saving inputs (labor and capital). At this point, it is clear that the degree of flexibility of the production process will depend to a large extent on institutional aspects related to the "rules of the game."

Meanwhile, the validation tests of the model are satisfactory, there is residual normality, and the level of explanation is adequate (the joint variations in the explanatory variables explain 93% of the variations in the PFP). The problem of residual dependence between cross-sections is solved using the White procedure.

Finally, the annex presents the effects by cross-section and period, respectively. Concerning the former, only Bolivia, Brazil, Peru and Uruguay have negative values, which means that these idiosyncratic values are added to the intercept of their respective equations. At the same time, the effects by period reflect the changes associated with the intercept in each of the reference years.