Introduction
In Organization for Economic Cooperation and Development (OECD) countries, education investments and outcomes are increasingly focused on by governments in terms of providing incentives for greater efficiency in schooling and transferring resources to respond to rising demandsFootnote 1 As education has long been recognized as a prerequisite for economic growth around the world, it improves economic development by increasing skills,Footnote 2 stimulating innovation,Footnote 3 establishing an environment for more efficient governanceFootnote 4 and mitigating inequality between social classes.Footnote 5 Therefore, identifying the potential factors and mechanisms that influence investment in education is imperative. Previous studies have focused on the impact of separate factors on education investment. For example, tax revenue implies the total available funds for government expenditure, which could further affect public investment in educationFootnote 6 ; technological innovation and industrial structure upgrades raise the demand for human capital,Footnote 7 which in turn forces government expenditure on education development. However, evidence of how the policy environment itself determines education policies remains limited. For OECD countries, policy decision-making, including education investment, is fundamentally based on the political ideology of the ruling parties in the context of democratic political regimes.Footnote 8 To fill the gap in the literature, this paper aims to investigate the impact of political orientation on education investment.
In a democratic society, the impact of political orientation on education investment is determined by the voter base of left- and right-wing parties.Footnote 9 Previous literature suggests that a government dominated by a leftist party tends to implement policies that benefit the working class.Footnote 10 The worker-class-benefiting ideology makes the leftist ruling government reduce expenditure on providing public goods and increase transfer payments to labor welfare. As an important public service provided by the government, public education investment could be decreased by the government ruled by the left party. On the other hand, the rightist party represents the interests of wealthier classes who have more willingness to maintain education development.Footnote 11 The rightist-dominated government, therefore, tends to reduce social welfare but increase public expenditure, which eventually expands education spending policy. Due to the incentive of the ruling party to implement policies to maximize its re-election prospects,Footnote 12 education investment policy can be explained by political orientation.
Another important mechanism through which political orientation might influence education investment is that such investment could serve as a strategic tool for cultivating a party’s target electorate. A key study in this context demonstrates that education tends to shift individuals towards right-wing ideologies. Specifically, each additional year of education is associated with an approximately 5–6% increase in right-leaning political views.Footnote 13 This finding suggests that parties with differing political orientations may be strategically inclined to adopt distinct educational investment policies to optimize their electoral success rates. More specifically, left-wing parties may reduce education funding to maintain allegiance from voters with limited educational attainment, while right-wing parties may increase such investment to cultivate a more educated voter base, potentially aligned with their policies.
Attitude to innovation and technology is another channel through which political orientation affects education investment. Past evidenceFootnote 14 suggested that the adoption of new technology could benefit capital owners by decreasing human resource inputs. Technological progress potentially transfers the demanded type of human capital from workers to a professional labor force, such as researchers and managers. As education is known as a key factor for technological progress, leftist and rightist parties have different policy preferences on education investment to back up their represented social classes. By reducing education investment, a leftist government can limit the adoption of new technologies and therefore decrease the unemployment rate, which is a major goal of the left ideology.Footnote 15 Conversely, a rightist government tends to increase expenditure on education to encourage innovative activities. The promotion of innovation could benefit capital owners by promoting human capital efficiency and increasing the profitability of firms.Footnote 16
The current literature about the relationship between political orientation and education investment exhibits significant variation depending on the political and economic context. Additionally, leftist parties have often implemented tighter budgets, while rightist parties have increased spending, particularly in education, to alleviate economic hardships, as observed in Hungary and Poland.Footnote 17 This pattern is supported by an analysis of cabinet ideology’s influence on education spending in thirteen post-Communist democracies between 1989 and 2004. However, the current literature primarily focuses on specific countries or regions.
Building on the available evidence, this paper attempts to answer two key research questions by using detailed information about political orientation and associated education investment: (1) can political orientation explain cross-country differences in education investment; and (2) how leftist and rightist ideologies affect government policies on education resource development. Examining the impact of political orientation on education resources is not a trivial task. While attempting to investigate the impact of political orientation on education resources, the adverse impact of education on political orientation could result in an endogeneity problem. It may bias estimates of how political orientation affects education resources. More specifically, differences in education policies can influence voters’ preferences for political ideology, which further determines their political orientation.Footnote 18 For example, voters with lower education are traditionally associated with votes for left-wing parties.Footnote 19 However, a recent study also suggested a transformation that higher-educated voters are currently more inclined to support left-wing parties with more progressive ideologies such as environmental issues; voters with less education are more inclined to support right-wing parties with conservative policies.Footnote 20
The uncertainty of the reverse relationship between the two factors makes it challenging to identify the impact of political orientation on education resources. To address the endogeneity issue, we performed instrumental-variable (IV) regressions that used external communist influence as an exogenous variable to make a causal inference. The political orientation in OECD countries is highly associated with the far-left ideology of communism. As the foreign communist influence is solely determined by the communist revolution in other countries,Footnote 21 IV estimations allow us to capture the causal impact of political orientation on education resources through the exogenous shock of communism.
Using a panel sample of 21 OECD countries from 1970 to 2020, our research contributes to the current literature on the relationship between political orientation and education resources in several aspects. First, our analysis employed political orientation data from past studiesFootnote 22 that firstly grouped political ideologies by a binary classification (i.e., left- or right-parties), which allows us to directly analyze political orientation under the scenario of two-party systems. Compared with previous political studies that only considered ideological differences, our study presents more straightforward results of how political orientation affects education resources. Second, we investigated the impact of political orientation on education resources in terms of both educational inputs and educational outputs. For the inputs, we considered government expenditure on education, while we also considered measures of education quality and outcomes by using pupil-teacher ratio and enrollment rate as alternative explanatory variables. Third, we used foreign communist influence as an instrumental variable (IV), as well as other techniques including System GMMFootnote 23 and IV estimatorsFootnote 24 to address the endogeneity issue. Finally, we further explored the possible mechanisms through which political orientation can affect education by splitting samples based on country differences in technological development (i.e., the number of researchers).
The rest of the paper is organized as follows. Section “Data” describes the data structure and variable definition. Section “Empirical strategy” outlines the empirical strategy. Section “Baseline results” presents the baseline results. Section “Robustness checks” presents results from sensitivity checks, and Section “Conclusion” concludes.
Data
Our empirical analysis was performed based on unbalanced panel data on political orientation, education resources, revenue, population ages between 0 and 14, economic structural change, and external communist influence index for 21 OECD countries from 1970 to 2020. The detailed country list can be found in Appendix A1. We provide summary statistics and definitions for the key variables in the estimations in Table 1.
Table 1. Summary statistics of variables
Notes: Expenditure educ is the total government expenditure on education (% of GDP). Pupil-teacher is an indicator calculated by taking PCA of pupil-teacher ratio in each stage. Enrollment pca is an indicator calculated by taking PCA of enrollment in each stage. Left measures the share (%) of party orientation on left. Right measures the share (%) of party orientation in right. Dominant left is a dummy variable that takes value of 1 if left party was in a dominant position (i.e., share of left-wing was greater than 50%). Dominant right is a dummy variable that takes value of 1 if right-wing party was in a dominant position (i.e., share of right-wing was greater than 50%). Revenue is the government revenue excluding grants share of GDP. Pop age0-14 is the share (%) of population ages 0–14 in total population. Industry and Service represent the industry and service sector share of GDP (%), respectively. Communist is the index (%) measures external communist influence in OECD countries.
Political orientation
The political orientation data were sourced from Election Results Data from this past study.Footnote 25 The vote-results data are manually hand-coded for 21 OECD countries which are also Western democracies. The database was established based on multiple sources of political attitudes surveys as well as official election results, including the Manifesto Project Database, Eurobarometer, the European Social Survey, and the European Election Studies. The data usefully separate political orientation into two major groups, the “left” and “right,” which allows us to compare election results in two-party systems with highly fragmented party systems. Parties on the left side of the political spectrum generally implement expansionary fiscal policies to support the benefits of labor forces.Footnote 26 The data categorize social democratic, socialist, communist, green, and their affiliated parties such as Labour in the UK and Australia, as well as the Democratic Party in the US, as left parties based on party ideology. Parties with ideas of protecting capital owners by tightening fiscal policies are on the right side of the political spectrum.Footnote 27 For example, conservative parties in the UK, the Republican Party in the US, and anti-immigration parties such as the Danish People’s Party are identified as traditional right-wing parties.
We separately calculated the percentage of left- and right-wing voting (Left and Right) in each election in each country, which represents the society’s cleavage on political orientation. For a given country c, we filled in the missing observations between two elections with the previous election result because the election results would have a persistent impact on government policy until the next election campaign.Footnote 28 The proportion of left or right-wing voting reflects the social intention of political ideology. We also extracted two dummy variables that take the value of 1 if the proportion of left- or right-wing parties is greater than a threshold of 50 percent in an election (Dominant left and Dominant right ), which indicates a government and its policy is fundamentally dominated by one particular political ideology during the term.Footnote 29
Education investment
We collected education data from the World Development Indicators (WDI) by the World Bank. The education resources were measured in three ways. First, we used the percentage of general government expenditure on education relative to GDP (Expenditure educ ), where a high percentage of education expenditure reflects that the government has a higher priority for education resources.Footnote 30 Second, we used the pupil-teacher ratio (Pupil-teacher pca ) as a measurement of education resource abundance, which is generated by dividing the number of students by the number of teachers at the same level of education. A higher pupil-teacher ratio indicates lower-level education resource abundance. In the education literature,Footnote 31 the pupil-teacher ratio is also widely used as a predictor of education quality and student performance. The third measurement is the ratio of school enrollment (Enrollment pca ), which divides the number of students by the population of the age group corresponding to the same level of education. The school enrollment rate is widely used as an outcome variable of educational investment,Footnote 32 which could eventually determine the level of human capital. For the pupil-teacher ratio and school enrollment that were recorded at each level of education (i.e., primary, secondary, and tertiary), we used the first standardized principal component of the three levels of education to combine the data.
Other explanatory (control) variables
We used the share of government revenue excluding grants in GDP, the share of population ages 0–14 in the total population, as well as the industry and service sectors’ share in the economic structure as other explanatory variables to control their potential effects on education resources. These control variables were also obtained from the World Bank’s WDI database.
First, we used the share of government revenue in GDP (Revenue) to control the potential influence of the government’s economic situation on education expenditure. Previous studies suggest that a government with a good economic situation generally invests more financial resources in the public education sector to obtain long-term economic benefits from the growth of human capital.Footnote 33 The positive effect of government revenue on education resources is thus expected. Second, we set the ratio of population ages 0–14 to the total population (Pop age0-14 ) as an additional explanatory variable to control the potential population impact on education resources. The young age population under 15 years old is an important determinant of education resource allocation. The population structure with a larger proportion of young children requires the government to provide more education services, as suggested by this study.Footnote 34 Hence, a positive correlation between the young population and education resources is expected.
Finally, we used the share of industry and service sectors in GDP (Industry and Service) to control the potential effect of in-demand skills on education resources. More specifically, Industry accounts for value added in fields such as manufacturing, construction, and electricity, while Service refers to value added in fields such as retail trade, financial, professional, and personal services, including healthcare. Previous literature has found that economies concentrated in the industry and services sectors have a high demand for skilled human capital.Footnote 35 As education investment is a key driver to increase the capacity of skilled human capital,Footnote 36 we expected a positive effect of industry and service sectors on education resources.
External communist influence
To identify the causal relationship between political orientation and education resources, we used external communist influence (Communist), which has been suggested to be an important determinant of political orientation,Footnote 37 as an instrumental variable. The Communist was instigated by the labor movement in the 21 OECD countries, which was influenced by communist regimes in a total of 111 countries that cover approximately 95% of the world’s population. The data were sourced from this past study.Footnote 38 The external communist influence was calculated as follows:
$${Communist_{i,t}} = {{{\mathop \sum \nolimits_{j\,=\,1}^{111} \left( {D_{j,t}^{Com}}{Pop_{i,j,t}}/Dist_{i,j}^{Lin} \right)}}\over{{\mathop \sum \nolimits_{j\,=\,1}^{111} \left( {Po{p_{i,j,t}}/Dist_{i,j}^{Lin}} \right)}}}$$
where
$$Dist_{ij}^{Lin}\,=\,1 - {\left\{ {{{{\omega _{i,j}}}}\over{{\left[ {0.5\left( {{\omega _i} + {\omega _j}} \right)} \right]}}} \right\}^\lambda }$$
D
Com
is a dummy variable that takes the value of 1 if the government is dominated by a communist party and 0 otherwise; Pop refers to the population size. The communist influence is weighed by
$Dist_{ij}^{Lin}$
, which denotes the linguistic distance as a proxy for cultural distance. The spread of political (communist) ideology is more dependent on cultural and linguistic distance rather than geographic proximity. Based on the concept of a complex network analysis,
${\omega _{i,j}}$
is the number of nodes between languages of i and j. The linguistic distance is scaled by a parameter set to 0.5Footnote
39
. The data only cover a time span from 1870 to 2011. We kept the external communist influence constant for each country using the observations from the year 2011.
Empirical strategy
We estimated the impact of political orientation on education resources using the following equation:
$$Edu{c_{i,t}} = {\beta _0} + {\beta _1}Poli{t_{i,t}} + {\beta _2}{X_{i,t}} + {\delta _i} + {\varepsilon _{i,t}}$$
where the dependent variable Educ i,t represents education resources measured by government expenditure, pupil-teacher ratio, and enrollment in country i in year t; Polit i,t refers to political orientation measured by percentage and dummy for left and right parties; X i,t is a vector of control variables including revenue, population ages 0–14, and the share of industry and service in the economic structure. We included country fixed effects δ i to control for other potential factors associated with country differences. β 1 is the coefficient of interest, which represents the impact of the political orientation of left or right on education resources.
The OLS estimation of Equation (3) is likely to suffer from endogeneity problem due to the potential reverse causality. The endogeneity issue could bias the estimated effect of political orientation on education resources, but the overall direction could be either positive or negative. The abundance of educational resources could affect educational attainment, which may eventually hinder the success of left-wing parties in electoral campaigns. Previous studies have suggested that lower-educated electorate was historically associated with voting for social democratic parties that were ideologically based on defending the rights of less off-well classes.Footnote 40 By contrast, the literature also suggested a transformation in which higher-educated voters turned to parties with more progressive policies including equality and environmental issues, while lower-educated voters supported more conservative ideologies such as anti-immigration and nationalism.Footnote 41 Therefore, the consequence of the endogeneity issue remains ambiguous.
In order to address the endogeneity problem, an exogenous variation of external communist influence on the 21 OECD countries was employed as an instrumental variable for two-stage least squares. There are two reasons why Communist could be used as a valid instrument to make causal inferences. First, the external communist influence captures the political orientation of the government and society. In political economics studies, communist influence is commonly used as a predictor of the left party’s movement, as it has been recognized as a far-left ideology.Footnote 42 Therefore, we expected the communist influence to positively impact left political orientation in the first-stage regressions.
Second, the external communist influence has no direct impact on education resources but only through its effect on political orientation, which fulfills the exclusion restriction assumption of IV estimations. In the case of external communist influence, this influence predominantly shifts political ideologies and alignments, which is more relevant to ideological and political persuasion rather than direct intervention in specific policy preferences and implementations, including in education investment.Footnote 43 More specifically, in a communist system, education policies are shaped by the ideology’s emphasis on societal equality, leading to state-funded education systems that aim to provide equal access and opportunities for all. Conversely, when a country transitions away from communism, ideology may shift towards development, involving a balance between maintaining educational equality and enhancing quality, innovation, and alignment with global standards. The impact of external communist influence on education investment in OECD countries is an indirect effect, resulting from changes in political orientation, rather than a direct consequence of the communist influence itself. Unlike forms of economic aid or direct investment,Footnote 44 this influence generally lacks direct mechanisms for affecting education investment, except through the alteration of political orientation.
We utilized two alternative techniques to address the endogeneity problem as a robustness check for the main results: (1) we used the IV estimatorFootnote 45 to technically address the endogeneity issue, which provides the identification of a causal relationship on the condition that errors of exogenous variables are heteroskedastic; (2) we employed the System GMM,Footnote 46 which generates internal lags of endogenous variables as instruments and estimates a dynamic panel model. The System GMM model is estimated by the equation:
$$Edu{c_{i,t}} = {\beta _3} + {\beta _4}Edu{c_{i,t - 1}} + {\beta _5}Poli{t_{i,t}} + {\beta _6}{X_{i,t}} + {\varepsilon _{i,t}}$$
where
$Edu{c_{i,t - 1}}$
is a lagged factor in education resources. The country-fixed effects are removed from Equation (4) because this estimator uses a dynamic small T and large N panel, which already contains country-fixed effects.Footnote
47
Baseline results
Table 2 shows the results estimated by panel OLS and IV regressions with fixed effects. The results from columns (1) and (2) are estimated by the panel fixed-effects model, and columns (3) and (4) are estimated by panel IV regressions. There are consistent results of a statistically significant impact of political orientation on education resources across all model specifications. More specifically, an increase of votes on the left parties generates a decrease in government expenditure on education, while an increase of votes on the right parties can increase the government’s investment in education. As expected, the magnitude of the effect estimated by IV regressions is different from panel fixed-effects estimates, which provides further evidence that our IV models have addressed the potential endogeneity issue caused by reverse causality and omitted variables. These results also suggested that votes on the right parties have a larger impact on education resources than votes on the left parties.
Table 2. The effect of political orientation on education expenditure
Notes: The regressions are estimated by panel OLS and IV models with fixed effects. The heteroskedasticity robust z-values are reported in the parentheses. The year coverage ranges from 1970 to 2020. We reported Kleibergen-Paap Lagrange Multiplier (LM) statistics for under-identification in panel B. The significant p-values indicate the rejection of null hypothesis that IV models are under-identified. Significance at the 10%, 5%, and 1% levels are indicated by *, **, and ***.
Panel B from columns (3) and (4) shows the first-stage results of IV estimations. First-stage results confirmed that the external communist influence has a positive effect on left-voting in elections and a negative effect on right-voting at the 1% significance level, which is consistent with the literature.Footnote 48 We also performed a Kleibergen-Paap Lagrange Multiplier test to provide a diagnosis of the model under-identification. The null hypothesis is that IV models are under-identified, and the results rejected the null hypothesis at the 1% significance level, which indicates the validity of our IV model. In addition, the direction of estimated coefficients of all control variables is as expected where the coefficients were statistically significant: the revenue, population aged 0–14, and service share in economic structure have a positive impact on education resources.
Robustness checks
We perform three types of robustness checks against our core results. First, we check whether our core results are driven by the specific measures of political orientation and education development used in the core model (i.e., left and right percentage measures and government expenditure on education). For this purpose, we adopt alternative measures of these two variables. In particular, we create left and right dummies which equals one if the left/right orientation is dominant (i.e., percentage share greater than 50%) and zero otherwise. This is used as an alternative measure for political orientation. We then collect data on pupil-teacher ratios as well as enrollment rates from various educational attainment levels (i.e., primary, secondary, and tertiary) and combine them using the first principal component analysis (PCA) to form overall measures of pupil-teacher ratio and enrollment rates. These then are used as two alternative measures for education development. These alternative measures of political orientation and education development are commonly used in the current literature.Footnote 49
Second, we check whether our core results are driven by the specific estimators that we use (in particular, the specific instrumental variable we use to take care of endogeneity) in the core model. For this purpose, we adopt estimators that take care of endogeneity using artificial/constructed instruments such as the two-step System GMM and Lewbel 2SLS. The system GMM approach utilizes internal time lags as artificial instruments for the endogenous explanatory variables and is implemented in a dynamic time panel setting.Footnote 50 In doing so, we also adopt the instrument collapsing technique to ensure that the number of instruments we use is less than the number of countries.Footnote 51 The Lewbel 2SLS approach uses constructed instruments that are synthesized from exogenous control variables and the endogenous explanatory variables and is used to identify causal relationships in the event that the errors of exogenous variables exhibit heteroskedasticity.Footnote 52 The above two estimators have been widely used in the current literature to address endogeneity.Footnote 53
Third, we examine the relationship between political orientation and education development in different economies characterized by innovation. The reasons are two folds. On the one hand, the current literature has documented a significant causal relationship between political orientation (especially in the form of government ideology) and innovation.Footnote 54 On the other hand, innovation is long argued in the standard literature to be a potential driver for education development (albeit the opposite can also be trueFootnote 55 ). Therefore, it’s interesting to see whether innovation can be a potential moderator in the relationship between political orientation and education development. In particular, we aim to answer the question: does innovation enhance or weaken the impact of political orientation on education development?
Table 3 presents results on the alternative measures of education development. It’s quite evident that the significant impact of political orientation on education development is confirmed regardless of how we measure education development. In particular, on the one hand, the left political ideology exerts a positive impact on the pupil-teacher ratio (see Column (1) in Table 3), suggesting that this ideology weakens education development, in line with the core results note that the higher the pupil-teacher ratio, the lower the education development). On the other hand, the left political ideology exerts a negative impact on enrollment rates (see Column (3) in Table 3), which also suggests the same. It needs to be pointed out though that the impact of right political ideology on pupil-teacher ratio is insignificant. We think this may be a statistical artifact caused by the low number of observations for the pupil-teacher ratio regressions (441 versus 667 for the regressions involving enrollment, see Table 3 for details). Nevertheless, the positive impact of the right political ideology on enrollment rates is confirmed (Column (4) in Table 3), in line with expectations. In all regressions, the first-stage results once again confirm the validity of our core instrument (communist influence) with the only exception in Column (2), which has insignificant second-stage results likewise.
Table 3. The effect of political orientation on education expenditure: alternative measure of education development
Notes: The regressions are estimated by panel IV models with fixed effects. The heteroskedasticity robust z-values are reported in the parentheses. The year coverage ranges from 1970 to 2020. We reported Kleibergen-Paap LM statistics for under-identification in panel B. The significant p-values indicate the rejection of null hypothesis that IV models are under-identified. Significance at the 10%, 5%, and 1% levels are indicated by *, **, and ***.
Table 4 presents results from alternative measures of political orientation (Columns (1) and (2)), and alternative estimators (Columns (3) to (6)). It’s quite evident that regardless of how we measure political orientation (whether by percentage share or by dummies), the significant positive/negative impacts of the right/left political orientation on education development are once again confirmed. The absolute value of the positive impact of the right orientation is also greater than the absolute value of the negative impact of the left orientation, which is consistent with the core results. It’s also evident that regardless of what estimators we use to address endogeneity, we obtain the same conclusion on the relationship between political orientation and education development. In both the System GMM (Columns (3) and (4)) and Lewbel 2SLS (Columns (5) and (6)) regressions, the positive impact of the right orientation once again outweighs the negative impact of the left orientation, in line with core results. Moreover, in the system GMM regressions, there’s evidence of first-order autocorrelation, but no second-order autocorrelation. The Hansen tests of overidentification in all regressions from Columns (3) to (6) pass, which suggests the validity of the artificial and constructed instruments used in System GMM and Lewbel 2SLS.
Table 4. The effect of political orientation on education expenditure: alternative measure of political orientation and alternative estimator for endogeneity
Notes: The regressions are estimated by panel IV, two-step System GMM, and Lewbel IV models with fixed effects. The heteroskedasticity robust z-values are reported in the parentheses. The year coverage ranges from 1970 to 2020. For panel IV regressions, we reported Kleibergen-Paap LM statistics for under-identification in panel B. The significant p-values indicate the rejection of null hypothesis that IV models are under-identified. For System GMM regressions, we reported the number of instruments (lags) used in GMM process and ensured that the number of instruments was less than the number of countries. l. Expenditure educ is the lagged dependent variable. We also reported p-values of AR(1) and AR(2) for System GMM regressions. Significance at the 10%, 5%, and 1% levels are indicated by *, **, and ***.
Finally, Table 5 presents results from splitting the sample using innovation. We use the median value of number of researchers per 100,000, which equals 3,937 in our 21-country sample, as the criteria to split the sample. In particular, the impact of political orientation on education development is only significant in the high innovation sample (Columns (1) and (2)). However, it needs to be pointed out that the low innovation sample yields consistent direction of impacts (i.e., positive impact from the right, and negative from the left), albeit the impacts are insignificant with the impact from the left at the borderline in terms of significance at 10%. We think this result could suggest that innovation is an important moderator in the relationship between political orientation and education development. In other words, a country needs to be at certain innovation level for the political ideology to impact education development. The divergent views on politics from society may not translate into divergent views on education development if innovation is not high enough to generate such incentive. This result therefore is important for policymakers who may want to take advantage of this relationship between political orientation and education development, since innovation level may be an inhibitor/enabler of such relationship.
Table 5. The effect of political orientation on education expenditure: high innovation countries vs low innovation countries
Notes: The regressions are estimated by panel IV models with fixed effects. The heteroskedasticity robust z-values are reported in the parentheses. The year coverage ranges from 1970 to 2020. We split sample into high R&D and low R&D countries by the median number of researchers in R&D (per million people). We reported Kleibergen-Paap LM statistics for under-identification in panel B. The significant p-values indicate the rejection of null hypothesis that IV models are under-identified. Significance at the 10%, 5%, and 1% levels are indicated by *, **, and ***.
Conclusion
In sum, this paper explores the potential causal relationship between political orientation and education development using a panel data of 21 OECD countries from 1970 to 2020 by utilizing estimators that address endogeneity (i.e. 2SLS, System GMM, and Lewbel 2SLS) and have found empirical evidence to support the following:
First, using communist influence as a physical instrument for political orientation, our results find that political orientation has a statistically significant impact on education investment policies. An increase in leftist party votes can decrease government expenditure on education, while an increase in rightist party votes can increase government investment in education. In addition, we find that rightist orientation has a stronger impact on education than leftist orientation.
Second, our results are robust against a number of sensitivity checks including alternative measures of political orientations and education development, alternative estimators that take care of endogeneity (i.e., System GMM and Lewbel 2SLS), and the moderation effect of innovation in the relationship between political orientation and education development. In all these robustness checks, the positive/negative impacts of the right/left political orientation on education development are consistent with the core results. Moreover, the absolute value of the positive impact from the right outweighs the absolute value of the negative impact from the left, in line with the core results.
Third, the above results have profound implications for policymakers who aim to boost education investment and development in their countries. On the one hand, the significant impacts of political orientations on education suggest that policymakers should aim to avoid the impact of party politics on education policymaking. There needs to be perhaps an independent organization set up to oversee the making and implementation of education policies (much like how a country’s central bank should be independent from the political system) so that the impact of politics on education can be minimized. On the other hand, the above policy recommendations should only be used in the context of the level of a country’s innovation development. For countries that are currently experiencing low level of innovation development, the above imperatives may be less relevant for policymakers, as the impacts of political orientation on education investment in this case is insignificant. However, as a country develops further into a mature economy with high level of innovativeness, policymakers should think more about the impacts of political orientation on education investment. Moreover, the differing impacts of right versus left-leaning politics on education investment (which are opposite to each other), also have significant implications for policymakers. For example, in case a dominant left-leaning political environment is in place, policymakers may need to make stronger political advice to their leadership in order to maintain an adequate level of public education funding and investment. This need may be less for a political environment that’s dominated by the right-leaning politics. However, this is not to say that any type of political orientation should be a factor in education investment policymaking, but that these conditions should be considered by policymakers to avoid and potentially offset when making education investment policies.
Competing interests
The authors hereby declare that there’s no conflict of interest of any kind.
Open access
