Innovations and income inequalities – a comparative study

Due to the complexity of relationships between innovations and income inequalities, the choice of measures to be taken in the course of their interaction is very important. This paper presents a regression analysis based on the selected measures of innovativeness (gross domestic expenditure on R&D, number of patent applications, the Creative Economy Index), income inequalities (Gini coefficient, top 3% and top 1% shares of national equalized income) and various control variables retrieved mostly from the Eurostat Database for 30 countries (European Union countries, Iceland, Norway) for the study period of 2005-2014. It has been found that higher gross domestic expenditure on R&D as a percentage of GDP tends to increase inequalities, while higher number of patent applications and higher value of the Creative Economy Index have the opposite effect. Besides, top income inequality is partly driven by different factors than broader measures of income inequalities.


INTRODUCTION
Since the 1970s many countries have witnessed the increasing income inequalities in their societies.There are several potential explanations for this trend.For instance, according to (Kuznets, 1955), increasing income inequalities may be the result of structural changes.Although Kuznets concentrated on the transition from agriculture to industry and the process of urbanization (noticing that rural populations can be characterized by lower average incomes and lower income inequalities, so an increasing share of urban population implies increasing income inequalities), his observations may also apply to the transition from industry to services, including financial ones.Similarly, transition from a centrally planned to a market-based economy meant significant increases in income inequalities for many countries that experienced this transition since late 1980s -early 1990s (Włodarczyk, 2013).
Currently, the literature identifies several other causes for income inequality such as skill-biased technological change, international trade and ongoing globalization, immigration, education, institutions

DATA AND METHODOLOGY
The analysis is confined to countries for which the values of the Creative Economy Index (CEI) were available.CEI has been calculated for 34 countries (European Union countries, Turkey, Iceland, Norway, Switzerland, Serbia and Macedonia) over the period 2005-2014.Therefore, annual data covering this period were collected for three measures of income inequalities (the Gini coefficient, top 3% and top 1% share of national equivalized income), three measures of innovativeness (gross domestic expenditure on R&D, number of patent applications and the Creative Economy Index) and several control variables (see Table 1).Due to unavailability of some data Turkey, Switzerland, Macedonia and Serbia were excluded from the sample.In case of remaining countries a few missing values were replaced with the earliest (latest) available observation or filled by linear interpolation.
The choice of control variables mostly reflects a standard practice.For instance, Aghion et al. (2015) control for the size of the government sector and financial sector, GDP per capita and the growth of total population, and suggest inclusion of data on marginal tax rates as taxation may affect both incentives to innovate and the top 1% income share.Due to the lack of data on marginal income tax rates, the analysis is based on total tax rates as a percentage of commercial profits which may explain some cross-country differences, but not necessarily the behavior of top income shares.Inclusion of inflation rate is motivated by the relationship between inflation and income inequalities discussed by Albanesi (2007).This list is supplemented with the unemployment rate, the percentage of working-age population with tertiary education (which was traditionally supposed to decrease income inequalities) and two variables referring to financial and trade openness to control for the impact of globalization.
As reported in table 2 in the appendix, all measures of income inequalities changed in the same direction in most countries with the exception of Estonia, Hungary (where increasing values of the Gini coefficient were observed along with decreasing values of top income inequality), and Lithuania (where an opposite tendency occurred).Gross domestic expenditure on R&D relative to GDP rose in the majority of countries over the analyzed period (with the most outstanding exception of Iceland where its value dropped from 2.7 in 2005 to 1.9 in 2014), but the other two measures of innovativeness changed in the same direction only in case of seven countries.Altogether, there was no universal pattern in terms of the relationship analyzed in the paper.There were countries where income inequalities and innovativeness were both increasing (e.g.Austria, Bulgaria, and Slovenia), both decreasing (Iceland), or the changes were in opposite directions (Czech Republic) for all or for some of the measures over 2005-2014.It is also worth mentioning that despite some country-specific differences, there is a common framework concerning innovation policy.All the countries belong to the European Economic Area with Research and Innovation constituting one of the core objectives of the Europe 2020 Strategy for smart, sustainable and inclusive growth.
In general, the sample used in the empirical analysis is a balanced panel of 30 states (European Union countries, Iceland and Norway) and a total of 300 observations (30 states over 10 years).For each combination of measures of innovations and income inequalities with all the control variables the Breusch-Pagan LM test was conducted.In each case the p-value was very close to zero providing evidence of significant differences across countries.Next, to decide on the character of individual effects (fixed or random) the Hausman test was run.In five cases (all the regressions including the gini and/or gerd variable) the null hypothesis that there is no correlation between regressors and effects (implying that both fixed and random effect estimators are consistent, but fixed effect estimator is inefficient) was rejected at the 0.05 level of significance.For the remaining four cases this hypothesis could not be rejected.Conducted analysis comes down to the estimation of a series of panel models including individual effects (fixed or random, dependent on the result of the Hausman test), regressing a measure of income inequalities in country i at time t, against a measure of innovativeness and a vector of control variables.For each combination of analyzed measures three models are estimated: (1) without time dummies, (2) with all time dummies, (3) with significant control variables and time dummies.

MAIN RESULTS
Main estimation results are reported in the appendix (see Table 3-5).As expected, conducted calculations demonstrate that the character of the relationship between innovation and income inequalities depends on the choice of the measure of innovativeness.In general, higher gross domestic expenditure on R&D as a percentage of GDP tends to increase inequalities, while a higher number of patent applications or a higher value of the Creative Economy Index has an opposite effect (however, for combinations: gini and patents as well as for top1 and CEI this relationship is not statistically significant in any specification).The strongest negative relationship is found for gini and CEI (observed in all three model specifications), implying that countries with better institutions supporting innovativeness are also more equal in terms of income distribution.
There is only one universal factor driving inequality in all specifications, namely the fraction of working age population with tertiary education.This confirms that a greater supply of high-skilled workers may not decrease income inequalities.Also a higher unemployment rate and a greater financial openness imply higher inequalities in most cases, but the impact of other factors on income inequalities depends on the choice of inequality measure (this contrast is especially pronounced between gini and narrow measures: top3 and top1).
As far as Gini coefficient is concerned, both the population growth rate and the level of GDP per capita in PPS have negative impact on its value (in the latter case the relationship is in fact nonlinear as squared values of GDP per capita had positive coefficients).For GDP per capita and financial openness conducted analysis confirms the results of Antonelli & Gehringer (2017), however, in the case analyzed here neither trade openness nor government expenditures play a significant role, although the signs of their coefficients are the same.
The main drivers of top income inequality are inflation and total financial assets, while trade openness tends to mitigate this phenomenon.Contrary to Aghion et al. (2015), no evidence is found for increased top income inequality to be driven by innovation, however, this result may be related to the specificity of the U.S. economy.Besides, as in Aghion et al. (2015), the negative relationship with GDP per capita seems to be weak at best, population growth plays an insignificant role (with a negative sign), while government expenditures decrease inequalities, but are often insignificant.Surprisingly, higher tax rates decrease top income inequality only in case of two models.
Obtained results are also similar to those of Jaumotte et al. (2013) who find a positive impact of financial globalization on income inequalities, negative for trade openness, positive for technology measured by the share of ICT in total capital stock, and a positive effect for population share with at least secondary education (albeit statistically insignificant).Furthermore, Peters and Volwahsen (2017) demonstrate positive impact of ICT investment as a percentage of total capital stock formation and unemployment rate on income inequalities, but report mixed results for financial openness.
Finally, the aim of including of time dummies was inter alia to capture the effects of the global financial crisis on income inequalities.However, in most specifications a significant decrease in income inequalities is observed in the years 2010, 2011 and 2012 (with 2005 as the reference year) which is the period of unfolding of the sovereign debt crisis in Europe.

CONCLUSIONS
On the whole, there is no single mechanism translating innovations into income inequalities.It is rather a dynamic interplay between capital and labor, their quality and quantity, their substitutability and complementarity, further complicated by measurement issues.
Empirical exercise conducted in this paper demonstrates that innovation can be the factor determining the scale of income inequalities.As already pointed by Jaumotte et al. (2013), innovation can have a potentially greater impact on income inequalities than globalization, because of two opposite pressures exerted by financial globalization and trade openness.
Different results obtained for different measures of innovativeness motivate to analyze various kinds of innovation and possibly include complementary measures of innovation in model specifications.For instance, as noted by Iacopetta (2008) faster technological change may increase income inequalities if it takes a form of product improvements, but cost-reducing innovations are more likely to decrease them.
The problem how to disentangle innovations reducing and increasing inequalities seems to be of particular interest for policymakers.Policy recommendations usually depend on chosen priorities.If the goal is the most efficient use of budget funds, the government may e.g.increase its expenditures on research and development (with income inequalities as a potential side effect).However, if the priority is given to low income inequalities, the government may consider implementing other, preferably more inclusive, instruments of innovation policy.

Table 1
Description of the dataset