Technological progress, human capital, and employment rate: an empirical analysis using P-ARDL models from 1960-2019^*

Progresso tecnológico, capital humano e taxa de emprego: uma análise empírica a partir de modelos P-ARDL no período 1960-2019

ABSTRACT

This paper aims to investigate the effects of technological progress, education, and human capital on employment in the short and long run using an econometric ARDL model for the period 1960-2019 across developed and developing countries. The hypothesis was that technological progress destroys jobs at a faster rate than investments in education and human capital can create. In the short run, the results confirmed the hypothesis that technological progress destroys relatively more jobs. However, in the long run, the results showed that the coefficients for education and human capital in developing countries outweighed those for technological progress, thus rejecting the hypothesis for this group of countries in the long run. For developed countries, coefficients for technological progress and investments in human capital were positive and significant.

KEYWORDS: Technological Progress; Human Capital; Employment

RESUMO

O trabalho buscou investigar os efeitos do progresso tecnológico, da educação e do capital humano sobre o emprego no curto e no longo prazo através de um modelo econométrico ARDL no período 1960-2019 para países desenvolvidos e em desenvolvimento. A hipótese é de que o progresso tecnológico destrói empregos a uma velocidade maior do que os investimentos em educação e capital humano é capaz de criar. No curto prazo, os resultados confirmaram a hipótese de que o progresso tecnológico destrói relativamente mais empregos. Entretanto, no longo prazo, os resultados demonstraram que os coeficientes da educação e do capital humano dos países em desenvolvimento superam os coeficientes do progresso tecnológico, invalidando a hipótese no longo prazo para este grupo de países. No caso dos países desenvolvidos, os coeficientes do progresso tecnológicos e dos investimentos em capital humano foram positivos e significativos para os países desenvolvidos.

PALAVRAS-CHAVE: Progresso Tecnológico; Capital Humano; Emprego

1. INTRODUCTION

Establishing a theoretical framework capable of explaining the factors causing unemployment is a daunting task. Each approach is based on different assumptions and analytical systems, but there is evidence that human capital creates jobs while technological progress tends to destroy them. The positive association among education, human capital, and employment is generally undisputed; however, the same cannot be said about the effects of technology.

Technological progress, while potentially leading to technological unemployment, has also created new businesses and opportunities, demanding more flexible workers with higher levels of education and better qualifications. It is essential to understand whether these new businesses and jobs are sufficient to counteract the trend of technological unemployment associated with labor-saving innovations. Thus, education and technology are clearly intertwined.

From the above, it is understood that education and human capital stimulate employment, while innovations and technological progress may promote technological unemployment but also create new business opportunities and jobs. This study aims to assess the net effects of these variables on employment, acknowledging their dependence on the structure and dynamics of social, economic, and institutional relations in which technology adoption occurs over the economic cycle.

Therefore, this research investigates the effects of technological progress, education, and human capital on employment in the short and long run using an Autoregressive Distributed Lag (ARDL) econometric model covering 65 countries from 1960 to 2019. These countries are divided into developed and developing groups. Additionally, the hypothesis that technological progress destroys jobs faster than education can create them will be examined.¹

To achieve the objective and test the hypothesis outlined, in addition to this introduction, the paper is structured into four sections. The second section provides a literature review. Following that, attention shifts to the data and methodology. The fourth section presents the empirical results. Finally, the last section contains the concluding remarks.

2. LITERATURE REVIEW

2.1. Education, Human Capital, and Employment

Knowledge is an important variable in economic models. When viewed from the perspective of the worker, it is referred to as human capital, defined as the skills, behaviors, and abilities that make workers more productive and generate positive social and economic impacts in society.

The theory of human capital was pioneered by Mincer (1958) and further developed by Schultz (1961) and Becker (1964). According to these authors, labor is more than just a factor of production; it is a type of capital that largely explains productivity gains and consequently, the monetary gains of workers.

Numerous empirical studies have estimated the effects of education and human capital on wage increases, productivity gains, and economic growth. However, fewer studies specifically analyze these effects on the employment variable.

Both classical economists and Keynesians, under the assumption of equilibrium in the economy, agree that real wages equal the marginal productivity of labor. However, mainstream economics suggests that cutting wages, by reducing production costs, could remedy unemployment. In contrast, Keynesians argue the opposite: reducing wages decreases effective demand, thus lowering employment levels (Keynes, 1996; Robinson & Wilkinson, 1977).

The effects of education on employment were not an explicit concern in Keynes’s theory of employment and effective demand. Keynesians believe that employment is determined by aggregate demand, with wages being one of its components. Assuming that education stimulates employment through real wage growth from a macroeconomic perspective, it can be seen as a private/public investment or government expenditure, thus contributing to aggregate demand.

Recently, some post-Keynesian theoretical-mathematical models have incorporated the role of public investments in education as components of aggregate demand and engines of economic growth (Carvalho, Lima & Serra, 2017; Lima, Carvalho & Serra, 2021).

Long run employment theory was not Keynes’s primary concern, but he recognized the need for a long run theory and briefly discussed the implications of expectations on employment (Eatwell, 1983; Asimakopulos, 1984). Investments in education are inherently long run because it takes time and effort to improve the educational level of a population, whereas labor productivity is more immediate. Perhaps for these reasons and others, Keynes did not explicitly demonstrate the potential impacts of education on employment levels.

Robinson (1947) analyzed the potential effects of increasing the school leaving age on employment. According to Robinson, in a community with fixed resources and production methods, the volume of production can only increase through higher investment rates or a reduction in the propensity to save. Therefore, an increase in years of education should be examined in this context due to its relationship with employment and production.

At the time, there were arguments that removing children from the labor market by increasing their years of education would raise production costs because older workers commanded higher wages, potentially reducing production. Robinson critiqued this argument, emphasizing that higher wages would actually stimulate production through increased aggregate demand.

The initial effect of increased education would likely be a reduction in the supply of workers.² Removing children and adolescents from the labor market statistics by opting to study for an additional year, as discussed by Robinson (1947), means that these individuals would not be counted as unemployed during that period, which can affect the statistical perception of unemployment.

Education and the development of human capital not only promote labor productivity and income growth but also stimulate employment through aggregate demand. They are essential for building a skilled workforce capable of handling new technologies and labor market demands.³

Technological progress has enhanced economies’ ability to produce more goods and services with less human labor, sparking discussions on jobless growth, technological unemployment, and the need for reducing working hours and implementing universal basic income. Despite these debates, employment cannot be entirely dissociated from economic output. In the long run, education and human capital can impact employment through economic growth, both via aggregate demand (public/private investments in education and human capital) and by increasing labor productivity.

Orthodox literature addresses the impact of human capital formation on economic growth from the perspective of aggregate supply, often overlooking potential impacts on aggregate demand components. Human capital significantly determines countries’ per capita GDP growth through its positive influence on labor productivity growth or technological change rates. Endogenous growth models emerged in response to dissatisfaction with the idea of economic growth being determined by exogenously given variables. These models expanded the concept of capital to include knowledge and education as crucial explanatory factors for economic growth.

Despite ample opportunities to study channels and mechanisms through which education and human capital impact economic growth, these discussions are less common in post-Keynesian models compared to mainstream economics. Nonetheless, there have been recent efforts to incorporate these issues from an aggregate demand perspective (Carvalho, Lima & Serra, 2017; Costa, 2016; Dutt, 2008, 2010; Lima, Carvalho & Serra, 2021; Renzi & Meirelles, 2014). Education creates opportunities and promotes social inclusion, but access to it can be restricted to higher-income social strata, making it not merely a personal choice and contributing to increased social inequalities and class stratification. Moreover, education can serve an ideological function by fostering acceptance of significant inequalities, creating a false impression of high income mobility (Dutt, 2008).

2.2. Technological Progress and Employment

The discussion between technological progress and employment is not recent and has been gaining relevance. According to Mattoso (2000), this debate is influenced by the economic cycle. During periods of prosperity, there is a tendency to emphasize the positive effects of technology on employment, whereas during crises, technological progress is blamed for unemployment. Generally, new technologies are labor-saving, although they can stimulate employment through, for example, the creation of new businesses and jobs.

Economic cycles, like technological progress, are inherent factors in capitalist economic development. The diffusion and adoption of technology have a dual dimension, as they promote technological unemployment while inducing long-run economic growth and the creation of new businesses. Thus, over the economic cycle, technological unemployment can be offset as a transient phenomenon (Schumpeter, 1997). Neo-Schumpeterians advance the view that the impacts of new technologies depend on historical, institutional, and structural factors. Therefore, technology and employment do not have a singular relationship among authors. Moreover, technological impacts are not uniform over time and space (Cardoso & Guedes, 1999; Proni, 2015).

Monteiro (2020) demonstrates through an empirical literature review that, regardless of the proxies used, product innovations tend to generate employment, while the impacts of process innovations vary between countries and periods analyzed.

In a conference in Madrid in 1930, Keynes (1999) stated that the accumulation of capital and technical inventions improves the standard of living but can lead to technological unemployment, especially disadvantaging countries not at the forefront of technological advancement.

Technological unemployment could occur because technical changes were outpacing the absorption of labor, but it would be only a temporary phase of adjustment, as:

Assuming there is no major war and no significant increase in population, the economic problem can be solved, or at least be within sight of a solution, within a hundred years. This means that the economic problem is not - if we look to the future - the permanent problem of the human species (Keynes, 1999, p. 4).

For post-Keynesians, unemployment is essentially a phenomenon resulting from deficiencies in effective demand, meaning its performance is determined by hypotheses and variables associated with the goods market. Short-run employment determination models in this theoretical approach include elements that enable discussions around technological unemployment caused by increased labor productivity. These models are divided into Marshallian (with flexible prices) and Kaleckian frameworks, where labor productivity is given. When considering an increase in labor productivity, ceteris paribus, this leads to technological unemployment if there is no corresponding increase in real wages in the same proportion (Lavoie, 2022).

In the long run, employment for post-Keynesians depends on the process of economic growth, which is observed when the rate of capital accumulation does not keep pace with the growth of the workforce and labor productivity (Shapiro, 1984).

Innovations increase the level of output, but their effects on employment levels in the long run are not straightforward to determine. The ultimate effect of inventions on long-run employment levels will depend on technical improvements⁴ and capital intensification.⁵ Every innovation results from an investment, but not every investment leads to an innovation. In the long run, where the capital stock is not fixed, investments serve a dual purpose: they are a component of aggregate demand, and they increase productive capacity. Thus, every innovation implies new productive capacity (Heller, 1999; Robinson, 1947).

Robinson (1947, p. 96-97) emphasizes that “in general, capital-using inventions have been the most frequent, there appears to be, from a long-period point of view, very strong grounds for the popular opinion that inventions tend to reduce employment.” This perspective led Harrod (1937) to argue that Robinson presents a pessimistic and bold view regarding the likely nature of inventions, criticizing the lack of a precise measure of the volume of capital which is not addressed in the analyzed work.

In the book Structural Change and Economic Growth, Pasinetti discusses the long-run unemployment issue through a growth model. Full employment occurs when the economy reaches potential output, meaning when effective demand equals potential demand.

Economic growth is fueled by technological progress, which stimulates increased productive capacity and labor productivity. However, the advancement of technological progress, by raising the equilibrium per capita output of the economy, can cause unemployment by pushing potential output growth beyond demand development. Therefore, to ensure full employment, demand must grow at a rate equivalent to the average growth rate of labor productivity.

This situation is unlikely due to the effects of technological progress on demand. Each sector and industry have specific characteristics, and the increase in equilibrium per capita output caused by technological progress will have different effects on the income elasticity of demand for each product, altering production composition over time. These effects are referred to by Pasinetti as structural dynamics, representing a natural cause of long-run unemployment (Shapiro, 1984).

3. DATA AND METHODOLOGY

The empirical analysis of this study is developed using the Autoregressive Distributed Lag (ARDL) econometric model proposed by Pesaran and Shin (1998) and Pesaran, Shin, and Smith (2001) for time series data. This model is dynamic with lags in both the dependent and independent variables, allowing for the examination of cointegration relationships (long run) among variables that are stationary I(0) and/or integrated I(1). The ARDL modeling provides consistent and efficient estimators in the presence of endogeneity among regressors, achieved by including lags in both endogenous and exogenous variables. The adaptation for non-stationary heterogeneous panels was developed by Pesaran and Smith (1995) and Pesaran, Shin, and Smith (1997, 1999), where the cross-sectional (N) and time series (T) observations should be large with T>N.

The literature presents three estimators for panel ARDL models (P-ARDL): Mean Group (MG), Pooled Mean Group (PMG), and Dynamic Fixed Effects (DFE). In these models, the main focus is typically on the long run, examining parameters and adjustment speeds. The time span of the time series should be large enough to ensure individual regressions for each group. For instance, the mean group estimator estimates regressions for each cross-section and presents coefficients as unweighted averages of these regressions, making it sensitive to outliers. The estimated coefficients vary heterogeneously between groups in both the short and long runs, without considering the possibility that certain parameters may be equal (Pesaran, Shin & Smith, 1999; Pesaran & Smith, 1995).

In the pooled mean group estimator, long-run coefficients are constrained to be identical across groups, while short-run coefficients, intercepts, and error variances differ. Coefficients are presented as a weighted average of individual group coefficients, with weights proportional to the inverse of their variance. This estimator serves as an intermediary between mean group and grouped fixed and random effects estimators, where intercepts may differ among groups while all other coefficients and error variances are constrained to be the same (Blackburne & Frank, 2007; Pesaran & Smith, 1995; Pesaran, Shin & Smith, 1997, 1999).

The dynamic fixed effects estimator, similar to the pooled mean group, constrains the slope coefficients and error variances to be equal across groups in the long run. However, it also restricts the short-run coefficients and adjustment speed. Only the individual intercepts differ across groups (Baltagi; Griffin; Xiong, 2000; Blackburne & Frank, 2007).

A P-ARDL model (p, q ₁ , q ₂, …, q _K )“can be expressed by Equation 1:

$y_{it}=\sum _{j=1}^p{\lambda }_{ij}{y}_{i,t-j}+\sum _{j=0}^q{\delta }_{ij}^{\text{'}}{x}_{i,t-j}+{\mu }_i+{\epsilon }_{it}$ (1)

in which:

• t=1, 2, …, T are time periods;

• i=1, 2, …, N represents the number of groups;

• x_it=vector of explanatory variables;

• μ_i =fixed effects;

• λ_ij =scalar;

• δ_ij =vector of coefficients (kx1).

In Equation 1, seasonal dummies, time trends, or other types of fixed regressors can be included. The model above can undergo reparametrization, as shown in Equation 2, to estimate long-run coefficients and the adjustment speed towards long-run equilibrium in an error correction framework:

$\Delta y_{it}={\varnothing }_{i}ECT\sum _{j=1}^{p-1}{\lambda }_{ij}^{*}\Delta y_{i,t-j}+\sum _{j=0}^{q-1}{\delta }_{ij}^{*\text{'}}\Delta x_{i,t-j}+{\mu }_i+{\epsilon }_{it}$ (2)

in case, ${\varphi }_i=-\left(1-\sum _{j=1}^p{\lambda }_{ij}\right)$ it is a parameter with the adjustment speed of the error correction term, ${\theta }_i=\sum _{j=0}^q\frac{{\delta }_{ij}}{\left(1-{\Sigma }_k{\lambda }_{ik}\right)}$, it is a vector with long-run relationships among the variables, $ECT=\left(y_{i,t-1}-{\theta }_i^{\text{'}}{X}_{it}\right)$ it is the error correction term and, ${\delta }_{ij}^{*}$ and ${\lambda }_{ij}^{*}$ are dynamic short-run coefficients expressed as:

${\lambda }_{ij}^{*}=-\sum _{m=j+1}^p{\lambda }_{im},j=1,2,...,p-1$ (3)

${\delta }_{ij}^{*}=-\sum _{m=j+1}^q{\delta }_{im},j=1,2,...,q-1$ (4)

A model of error correction implies that the short-run dynamics of the system variables are influenced by deviations from long run equilibrium. A significant and negative parameter indicating that, given a disequilibrium, the variables return to long-run equilibrium.

The short and long-runs relationships among the variables were estimated for the period 1950-2019. The choice of period aimed to encompass the largest amount of annual data for the selected variables. The estimations were conducted using a dataset comprising 65 countries, each with no more than five consecutive years of missing data. These estimations were also performed for two groups of countries: developed and developing countries.⁶ Table 1 presents the variables used in the study and their respective sources.

Table 1
Description of Variables and Data Source

Acronyms	Variables	Source
TXEMP	Rate of growth of the employment rate	Data was constructed from version 10.0 Penn World Table (PWT)1
TXCH	Rate of growth of the per capita human capital index
TXPT	Rate of growth of labor productivity
TXESC	Rate of growth in average years of schooling	The data were constructed from the databases Barro-Lee Educational Attainment Dataset2 and Cohen-Soto International Educational Attainment Database3

• Source: Own elaboration.
• Note: Further details can be found in ¹ Feenstra, Inklaar, and Timmer (2015), ² Barro and Lee (2013), and ³ Cohen and Soto (2007).

To verify the elasticity, all variables were transformed into growth rates, thereby avoiding issues stemming from high correlation among the regressors. The employment rate was calculated as the ratio of the number of workers to the population size. The number of workers is defined as all individuals aged 15 or older who worked during the reference week, even if only for one hour per week, or were not at work but had a job or business from which they were temporarily absent.

The variable TXPT refers to the growth rate of labor productivity defined as the quotient of the number of workers and the gross domestic product at constant national prices (US$ 2017). It would be advantageous to incorporate control variables related to the behavior of personal and functional income distribution; however, consistent series were not found within the timeframe of this study.

As proxies for human capital, the variables TXCH and TXESC were used. The per capita Human Capital Index is calculated based on the average years of schooling:⁷

$c{h}_{it}=e^{\varphi \left(s_{it}\right)}$ (5)

where, education return function, country, time and average years of schooling. The education return function is represented by:

$\varphi \left(s\right)=\left\{\begin{array}{cc}0.134.s& ses\le 4\\ 0.134.4+0.101\left(s-4\right)& se4<s\le 8\\ 0.134.4+0.101.4+0.68\left(s-8\right)& ses>8\end{array}\right.$

The phrase suggests that the first years of education or schooling have a higher return on human capital, reflected in higher wages, compared to later years.

The variable TXESC was calculated based on a linear extrapolation of the average years of schooling of individuals aged 25-64 from the Barro-Lee (BL) and Cohen-Soto (CS) educational performance datasets, which exhibit a high correlation coefficient of 0.9450 for the overall sample of all countries and the selected period.

The BL dataset is available for 146 countries at five-year intervals from 1950 to 2010, while the CS dataset covers 95 countries at ten-year intervals from 1960 to 2020. To fill in missing years of data, linear interpolations were performed per country within both the BL and CS datasets. Subsequently, a linear extrapolation per country was conducted to estimate the average years of schooling based on the BL dataset using data from CS.

4. RESULTS

The stationarity tests (unit root tests) for panel data by Levin, Lin, and Chun (LLC), Im, Pesaran, and Shin (IPS), ADF-Fisher, PP-Fisher, and Breitung were conducted for the variables used in this study and are presented in Table 2. Most variables are stationary at levels. Only TXCH exhibits a unit root in the sample of developing countries, and it is stationary after taking the first difference.

Table 2
Stationarity Tests (Unit Root Tests) for Panel Data in the Period 1960-2019

Variables/Tests	LLC	Breitung	IPS	ADF-Fisher	PP-Fisher
Developed Countries
TXEMP	-20,5112b (0,0000)***	-15,3997b (0,0000)***	-16,7696b (0,0000)***	361,353b (0,0000)***	341,316b (0,0000)***
TXPT	-25,2183b (0,0000)***	-16,1302b (0,0000)***	-25,1069b (0,0000)***	572,806b (0,0000)***	773,575b (0,0000)***
TXCH	-0,7657b (0,2219)	-2,8646b (0,0021)***	-1,4072b (0,0797)*	62,3981b (0,2024)	59,2482b (0,2901)
D(TXCH)	-29,4927b (0,0000)***	-24,9328b (0,0000)***	-23,9328b (0,0000)***	550,784b (0,0000)***	904,324 (0,0000)***
TXESC	0,1567b (0,5623)	-3,6553b (0,0001)***	-1,7987b (0,0360)**	69,0243b (0,0818)*	76,0760 (0,0255)**
Developed Countries
TXEMP	-17,3834b (0,0000)***	-17,1434b (0,0000)***	-19,7039b (0,0000)***	501,832b (0,0000)***	847,892b (0,0000)***
TXPT	-16,6298b (0,0000)***	-12,6543b (0,0000)***	-20,6227b (0,0000)***	528,513b (0,0000)***	1012,47b (0,0000)***
TXCH	-1,5119a (0,0644)*		-3,8966a (0,0000)***	121,095a (0,0005)***	120,402a (0,0005)***
TXESC	-1,1120b (0,1331)	-3,9068b (0,0000)***	-3,3559b (0,0004)***	108,599b (0,0055)***	133,187b (0,0000)***

• Source: Own elaboration. Notes: a) p-values in parentheses. b) (*), (**), and (***) indicate rejection of the null hypothesis at 10%, 5%, and 1% significance levels, respectively. c) 1 Lag. d) a=with constant, b=with constant and trend, c=without constant and trend.

The choice among DFE, PMG, and MG estimators was made using Hausman tests. To determine the appropriate number of lags, individual regressions were estimated for each country with a maximum of two lags. From the selected ARDL models for each country, an average was taken to define the P-ARDL model. Table 3 presents the estimation results for TXEMP across all countries, developed countries, and developing countries.⁸

Table 3
Estimates for the Employment Growth Rate using P-ARDL in the Period 1960-2019

Long Run
Variables/Model	Developing Countries		Developed Countries
	(1)	(2)	(3)	(4)
TXPT	-0,0439 (0,0000)***	-0,0535 (0,0000)***	0,2784 (0,0000)***	0,2666 (0,0000)***
TXCH	0,3648 (0,0000)***		0,1470 (0,1210)
TXESC		0,1349 (0,0000)***		0,0821 (0,0490)**
Short Run
Variables/Model	Developing Countries		Developed Countries
	(1)	(2)	(3)	(4)
Φ	-0,6979 (0,0000)***	-0,6973 (0,0000)***	-0,5163 (0,0000)***	-0,5247 (0,0000)***
D(TXEMP(-1))	-0,0118 (0,6590)	-0,0133 (0,6080)	0,1266 (0,0000)***	0,1355 (0,0000)***
D(TXPT)	-0,0772 (0,0000)***	-0,0738 (0,0010)***	-0,0965 (0,0000)***	-0,0940 (0,0000)***
D(TXCH)	0,1886 (0,4150)		-0,4089 (0,0740)*
D(TXESC)		-0,0332 (0,4980)		-0,0492 (0,4600)
C	-0,2791 (0,0000)***	-0,3432 (0,0000)***	-0,2629 (0,0000)***	-0,2695 (0,0000)***
Trend	0,0001 (0,0000)***	0,0002 (0,0000)***	0,0001 (0,0000)***	0,0001 (0,0000)***
Estimator	PMG	PMG	PMG	PMG
Hasuman Test	4,65a (0,0978)*	3,18a (0,2036)	1,24a (0,5376)	5,28a (0,0713)*

• Source: Own elaboration. Note: a) *, **, and *** indicate rejection of the null hypothesis at 10%, 5%, and 1% significance levels, respectively; b) Models (1) and (2) were estimated allowing for within-group correlation in standard error calculation; c) “a” indicates Hausman tests conducted for MG and PMG estimators, while “b” compares MG and DFE estimators.

In the short run, does not show statistical significance in developing countries, whereas in developed countries, its coefficients are positive and statistically significant at the 1% level. This suggests that is more volatile in developing countries, and an increase in the employment growth rate in the current period stimulates a proportional increase in the subsequent period. The variable TXESC does not demonstrate statistical significance in any of the selected samples, while TXCH shows a negative and statistically significant relationship with TXEMP in developed countries at the 10% significance level. This indicates that workers often need to temporarily leave the labor market to improve their qualifications. In all selected samples, TXPT has statistically significant negative coefficients. Thus, the short-run estimations confirm the hypothesis that technological progress destroys jobs at a faster rate than education creates them through human capital development.

In the long run, for developing countries, TXPT shows a significant negative relationship with TXEMP, whereas the opposite is observed for TXCH and TXESC, variables used as proxies for human capital, which exhibit positive and significant relationships with TXEMP. A 1% increase in TXCH and TXESC leads to respective effects of 0.3648% and 0.1349% on the dependent variable. It’s noteworthy that the coefficients of these human capital proxies are considerably larger than those of TXPT.

In the case of developed countries, there is a different scenario regarding the relationship with the labor productivity growth rate, as it shows a positive association with TXEMP. This implies that increases in TXPT stimulate the growth of TXEMP in the long run. Regarding the proxies for human capital, statistical significance was found for TXESC, which exhibited a positive coefficient, although relatively smaller compared to those observed for developing countries. Furthermore, the coefficients for the human capital proxies were smaller than those estimated for the labor productivity growth rate. A 1% increase in TXESC and TXPT results in an increase of 0.2666% and 0.0821%, respectively, in TXEMP.

The impact of the education growth rate in developing countries is at least 64% greater than in developed countries. This suggests that investment in education is a crucial tool to mitigate the negative effects of technological progress on employment in developing countries.

As expected, in all models, the estimated error correction coefficients, ϕ, were negative and statistically significant at the 1% level, ranging from -0.5163 to -0.6979. Thus, any long-run imbalance is adjusted within two periods/years.

In models (1) and (2), on average, 70% of the long-run imbalance is corrected each period/year in developing countries, while in models (3) and (4), a respective adjustment of 52% is observed, indicating that the return to long-run equilibrium is slower in developed countries.

5. FINAL REMARKS

This study aimed to investigate the effects of technological progress, education, and human capital on employment in the short and long run through an ARDL econometric model spanning from 1960 to 2019 for both developed and developing countries.

In the short run, the findings validate the hypothesis that technological progress destroys jobs at a faster rate than education and human capital create them. Specifically, for both developed and developing countries, an increase in the labor productivity growth rate reduces the employment growth rate, while increases in education and human capital growth rates do not yield statistically significant impacts. However, it is noted that human capital had a negative and significant effect on employment growth in developed countries, suggesting that workers often need to temporarily exit the labor market to enhance their qualifications.

In the long run, the literature review and empirical evidence strengthen the understanding that the impacts of technological progress on employment are complex and contingent upon the structure and dynamics of social, economic, and institutional relations. In developed countries, an increase in labor productivity growth is associated with higher employment growth rates, whereas the opposite holds true for developing countries. The impacts of education and human capital growth rates on employment growth are positive and significant in developing countries in the long run. In developed countries, only education growth rates showed positive effects. Moreover, the impact of education growth rates in developing countries is 64% higher than in developed countries.

Although labor productivity growth and education/human capital growth rates are respectively associated with a reduction and increase in employment growth rates, the coefficients for labor productivity growth are considerably smaller compared to those for education and human capital. Thus, the hypothesis that technological progress destroys jobs at a faster rate than education/human capital creates them was not validated in the long run.

These results indicate that investments in education and human capital are crucial in developing countries to counteract technological unemployment trends and reduce long run unemployment rates. Furthermore, it is through investments in education that developing countries can effectively position themselves at the forefront of technological progress.

REFERENCES

• ASIMAKOPULOS, Athanasios. “Long-period Employment” in The General Theory. Journal of Post Keynesian Economics, v. 7, n. 2, p. 207-213, 1984.
• BALTAGI, Badi H.; GRIFFIN, James M.; XIONG, Weiwen. To pool or not to pool: Homogeneous versus heterogeneous estimators applied to cigarette demand. Review of Economics and Statistics, v. 82, n. 1, p. 117-126, 2000.
• BARELLI, Walter et al. O futuro do emprego. São Paulo: Lazuli, 2002. 120 p.
• BARRO, Robert J.; LEE, Jong Wha. A new data set of educational attainment in the world, 1950-2010. Journal of Development Economics, v. 104, p. 184-198, 2013.
• BECKER, Gary S. Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. 1964.
• BLACKBURNE III, Edward F.; FRANK, Mark W. Estimation of nonstationary heterogeneous panels. The Stata Journal, v. 7, n. 2, p. 197-208, 2007.
• BOLDRIN, Michelle. A inovação destrói empregos com mais rapidez do que a educação os salva. 2017. Disponível em: <Disponível em: https://brasil.elpais.com/brasil/2017/07/19/economia/1500475025_052040.html#:~:text=A%20educa%C3%A7%C3%A3o%20pode%20mudar%20alguma,um%20produto%20de%20pior%20qualidade >. Acesso em 13 jul. 2022.
  » https://brasil.elpais.com/brasil/2017/07/19/economia/1500475025_052040.html#:~:text=A%20educa%C3%A7%C3%A3o%20pode%20mudar%20alguma,um%20produto%20de%20pior%20qualidade
• CARD, David; KLUVE, Jochen; WEBER, Andrea. Active labour market policy evaluations: A meta-analysis. The Economic Journal, v. 120, n. 548, p. F452-F477, 2010.
• CARDOSO, Sidnéia Reis; GUEDES, Cezar; DIEESE. Emprego e inovações tecnológicas no pensamento econômico. Emprego e desenvolvimento tecnológico: artigos dos pesquisadores. São Paulo: DIEESE, 1999.
• CARVALHO, Laura; LIMA, Gilberto Tadeu; SERRA, Gustavo Pereira. Debt-financed knowledge capital accumulation, capacity utilization and economic growth. University of São Paulo, Department of Economics Working Paper Series, v. 32, 2017.
• COHEN, Daniel; SOTO, Marcelo. Growth and human capital: good data, good results. Journal of Economic Growth, v. 12, n. 1, p. 51-76, 2007.
• COSTA, Júlia Mendonça da. Três ensaios em inovação tecnológica e crescimento econômico. Tese de Doutorado. Universidade de São Paulo. 2016.
• DUTT, Amitava Krishna. Education, growth and distribution: a heterodox macrodynamic perspective. Department of Economics, University of Notre Dame, 2008.
• DUTT, Amitava Krishna. Keynesian growth theory in the 21st century. In: 21st Century Keynesian Economics. Palgrave Macmillan, London, 2010. p. 39-80.
• DUTT, A. K.; VENEZIANI, R. A classical model of education, growth and distribution. Macroeconomic Dynamics, 2018.
• DUTT, Amitava Krishna; VENEZIANI, Roberto. Education and ‘human capitalists’ in a classical-Marxian model of growth and distribution. Cambridge Journal of Economics, v. 43, n. 2, p. 481-506, 2019.
• EATWELL, John. The long-period theory of employment. Cambridge Journal of Economics, v. 7, n. 3/4, p. 269-285, 1983.
• FEENSTRA, Robert C., Robert Inklaar and Marcel P. Timmer (2015), “The Next Generation of the Penn World Table” American Economic Review, 105(10), 3150-3182, available for download at http://www.ggdc.net/pwt
  » http://www.ggdc.net/pwt
• FEENSTRA, Robert C.; INKLAAR, Robert; TIMMER, Marcel P. User Guide to PWT 10.0 data files. 2021. Disponível em: Disponível em: https://www.rug.nl/ggdc/docs/pwt100-user-guide-to-data-files.pdf Acesso em: 11 fev. 2022.
  » https://www.rug.nl/ggdc/docs/pwt100-user-guide-to-data-files.pdf
• GUGER, Alois; MARTERBAUER, Markus; WALTERSKIRCHEN, Ewald. Growth policy in the spirit of Steindl and Kalecki. Metroeconomica, v. 57, n. 3, p. 428-442, 2006.
• GUGER, Alois; WALTERSKIRCHEN, Ewald. Josef Steindl’s Life and Work in Austria. PSL Quarterly Review, v. 65, n. 261, p. 135-149, 2012.
• HARROD, R. F. Review of Essays in the Theory of Employment., by J. Robinson. The Economic Journal, Oxford, v. 47, n. 186, p. 326-330, jun. 1937.
• HELLER, Claudia. Technichal Progress and Employment: Kalecki’s theorem and Joan Robinson’s model. Kalecki’s theorem and Joan Robinson’s Model. 1999. Economic Dynamics of Contemporary Capitalism: a Conference in Honour of Michal Kalecki’s Centenary. Disponível em: <Disponível em: https://www.researchgate.net/publication/215798922_Technichal_Progress_and_Employment_Kalecki_s_theorem_and_Joan_Robinsons_Model >. Acesso em: 15 jul. 2022.
  » https://www.researchgate.net/publication/215798922_Technichal_Progress_and_Employment_Kalecki_s_theorem_and_Joan_Robinsons_Model
• KALECKI, Michal. Os Aspectos Políticos do Pleno Emprego. In: KALECKI, Michal. Crescimento e Ciclo das Economias Capitalistas. São Paulo: Hucitec, 1987. p. 54-60.
• KAPLANIS, Ioannis. Local human capital and its impact on local employment chances in Britain. 2010.
• KARAYIANNIS, Anastassios D.; KATSELIDIS, Ioannis A. Technological Effects on Wages and Labour: Classical and Neoclassical Ideas. 2012.
• KEYNES, John Maynard. Perspectivas econômicas para os nossos netos. DE MASI, Domenico. Desenvolvimento sem trabalho. São Paulo: Ed. Esfera, 1999.
• KEYNES, J. M. Teoria geral do emprego, do juro e da moeda. São Paulo: Nova Cultural, 1996.
• LAVOIE, Marc. Post-Keynesian economics: new foundations. Edward Elgar Publishing, 2022.
• LIMA, Gilberto Tadeu; CARVALHO, Laura; SERRA, Gustavo Pereira. Human capital accumulation, income distribution, and economic growth: a demand-led analytical framework. Review of Keynesian Economics, v. 9, n. 3, p. 319-336, 2021.
• LUCAS JR, Robert E. On the mechanics of economic development. Journal of Monetary Economics, v. 22, n. 1, p. 3-42, 1988.
• MANKIW, N. Gregory; ROMER, David; WEIL, David N. A contribution to the empirics of economic growth. The Quarterly Journal of Economics, v. 107, n. 2, p. 407-437, 1992.
• MANNING, Alan. We can work it out: the impact of technological change on the demand for low-skill workers. Scottish Journal of Political Economy, v. 51, n. 5, p. 581-608, 2004.
• MARX, Karl. O Capital - Livro 1: Crítica da economia política. Livro 1: O processo de produção do capital. Boitempo Editorial, 2015.
• MATTOSO, Jorge. Tecnologia e emprego: uma relação conflituosa. São Paulo em perspectiva, v. 14, p. 115-123, 2000.
• MONTEIRO, Ana Bárbara Tomé Victor de Faria. Será que a inovação cria ou destrói emprego?: o efeito mediador das trajetórias tecnológicas. 2020. 40 f. Dissertação (Mestrado) - Curso de Economia, Faculdade de Economia, Universidade do Porto, Porto, 2020. Disponível em: <Disponível em: https://repositorio-aberto.up.pt/bitstream/10216/127861/2/408586.pdf >. Acesso em: 14 jul. 2022.
  » https://repositorio-aberto.up.pt/bitstream/10216/127861/2/408586.pdf
• NELSON, Richard R.; PHELPS, Edmund S. Investment in humans, technological diffusion, and economic growth. E, v. 56, n. 1/2, p. 69-75, 1966.
• PESARAN, M. Hashem; SMITH, Ron. Estimating long-run relationships from dynamic heterogeneous panels. Journal of Econometrics, v. 68, n. 1, p. 79-113, 1995.
• PESARAN, M. Hashem; SHIN, Yongcheol; SMITH, Ron P. Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association, v. 94, n. 446, p. 621-634, 1999.
• PESARAN, M. Hashem; SHIN, Yongcheol; SMITH, Richard J. Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, v. 16, n. 3, p. 289-326, 2001.
• PESARAN, M. H.; SHIN, Y. An autoregressive distributed-lag modelling approach to cointegration analysis. Econometric Society Monographs, v. 31, p. 371-413, 1998.
• PESARAN, M. Hashem et al. Pooled estimation of long-run relationships in dynamic heterogeneous panels. 1997.
• PRONI, Marcelo Weishaupt. Teorias do desemprego: um guia de estudo. Instituto de Economia - Unicamp, Campinas, n. 259, 2015.
• RENZI, Adriano; MEIRELLES, José Gabriel Porcile. Um modelo kaleckiano com gasto público em educação. Economia e Sociedade, v. 23, p. 319-339, 2014.
• ROBINSON, Joan. Essays in the Theory of Employment. Oxford: Basil Blagkwell, 1947.
• ROBINSON, Joan; WILKINSON, Frank. What has become of employment policy?. Cambridge Journal of Economics, v. 1, n. 1, p. 5-14, 1977.
• SASAKI, Hiroaki. Conflict, growth, distribution, and employment: a long-run Kaleckian model. International Review of Applied Economics, v. 25, n. 5, p. 539-557, 2011.
• SCHULTZ, Theodore William. Investment in human capital. The American Economic Review, v. 51, n. 1, p. 1-17, 1961.
• SCHULTZ, Theodore William. O capital humano: investimentos em educação e pesquisa. Rio de Janeiro: Zahar Editores, 1973.
• SCHUMPETER, Joseph. Teoria do desenvolvimento econômico: uma investigação sobre lucros, capital, crédito, juro e ciclo econômico. São Paulo: Nova Cultura. 1977 [1926]. (Os Economistas).
• SHAPIRO, Nina. Involuntary unemployment in the long run: Pasinetti’s formulation of the Keynesian Argument: a review article. Journal of Post Keynesian Economics, v. 7, n. 2, p. 235-245, 1984.
• SOLOW, Robert M. A contribution to the theory of economic growth. The Quarterly Journal of Economics, v. 70, n. 1, p. 65-94, 1956.
• STEINDL, Josef. Economic Papers 1941-88. Springer, 1990.
• STEINDL, Josef. Progresso técnico, distribuição e crescimento. GAREGNANI, P. Progresso técnico e teoria econômica. São Paulo: Hucitec, 1980.
• STOCKHAMMER, Engelbert; ONARAN, Özlem. Accumulation, distribution and employment: a structural VAR approach to a Kaleckian macro model. Structural Change and Economic Dynamics, v. 15, n. 4, p. 421-447, 2004.
• UZAWA, Hirofumi. Optimum technical change in an aggregative model of economic growth. International Economic Review, v. 6, n. 1, p. 18-31, 1965.

• *
  The authors thank CAPES for financial support.
• 1
  Inspired by the interview with the Italian economist Michele Boldrin titled “Innovation destroys jobs more quickly than education saves them” (Boldrin, 2017).
• 2
  In post-Keynesian literature, there is no construction of a labor supply function. However, Lavoie (2022) analyzes the effects of adopting different labor supply curves on short run employment levels determined by the principle of effective demand.
• 3
  Nelson and Phelps (1966), Steindl (1990) and Schultz (1973) point to human capital as an important variable for the generation of new technologies.
• 4
  An increase in output/income without changes in production factors/resources.
• 5
  Increase in the capital-labor ratio.
• 6
  ^{Appendix A} presents the selected countries.
• 7
  For further details, see Feenstra, Inklaar and Timmer (2021).
• 8
  The conduct of cointegration tests for variables is not strictly necessary in P-ARDL models. However, the results of the Pedroni tests for models (1)-(4) are presented in Appendix B, indicating that the variables are cointegrated and have a long run relationship.

• 20
  JEL Classification: E24; J24; O33.

APPENDICES

A - Selected Countries

Developed Countries (28 countries): Australia (AUS), Austria (AUT), Belgium (BEL), Canada (CAN), Switzerland (CHE), Cyprus (CYP), Germany (DEU), Denmark (DNK), Spain (ESP), Finland (FIN), France (FRA), United Kingdom (GBR), Greece (GRC), Ireland (IRL), Iceland (ISL), Israel (ISR), Italy (ITA), Japan (JPN), Republic of Korea (KOR), Luxembourg (LUX), Malta (MLT), Netherlands (NLD), Norway (NOR), New Zealand (NZL), Portugal (PRT), Sweden (SWE), Taiwan (TWN), United States (USA).

Developing Countries (37 countries): Argentina (ARG), Bolivia (BOL), Brazil (BRA), Chile (CHL), China (CHN), Democratic Republic of Congo (COD), Colombia (COL), Costa Rica (CRI), Dominican Republic (DOM), Ecuador (ECU), Egypt (EGY), Ghana (GHA), Guatemala (GTM), India (IND), Iran (IRN), Jamaica (JAM), Jordan (JOR), Kenya (KEN), Sri Lanka (LKA), Morocco (MAR), Mexico (MEX), Mauritius (MUS), Malawi (MWI), Malaysia (MYS), Pakistan (PAK), Peru (PER), Philippines (PHL), Paraguay (PRY), Thailand (THA), Trinidad and Tobago (TTO), Turkey (TUR), Uganda (UGA), Venezuela (VEN), South Africa (ZAF), Zambia (ZMB), Zimbabwe (ZWE).

B - Pedroni Test

Publication Dates

• Publication in this collection
  20 Dec 2024
• Date of issue
  2025

History

• Received
  08 Aug 2022
• Accepted
  16 Mar 2024