How does the inequality of young people in the economy affect growth

across countries? (to amend title)

Section 1

- Summary stats

- Literature review
Section 2: Summary statistics
Need table of figure 1.2
The figure 1.2 illustrates the variables chosen from ‘the world inequality database (WID)’ the
dataset presents 109 different countries the observations are taken in the year 2020. The
table illustrates only the base variables of our model. In the model we are testing the
dependent variable GDP which is a measure of the countries’ outputs given a specific
period. The independent variables we have chosen are:

1. ‘Share of youth not in education, employment or training, total (% of youth population’

(WB58).
2. ‘Government expenditure on education, total (% of GDP)’ (WB17)
3. ‘Labor force participation rate for ages 15-24, total (%) (modelled ILO estimate)’
(WB21).

We felt that these variables are best able to illustrate the specific effect of how inequality of
opportunity for youth creates a negative impact on the economic development of a country.
The first variable shows how well the country is able to provide opportunities for its young
people, combining multiple factors such as the labour market problems, access to education,
as well as socioeconomic conditions. The second variable further illustrates the capacity for
youth to enter the labour force, and the third one is the best estimate we have available to
demonstrate the level of education available to the youth.

We have also chosen multiple control variables, mainly related to economic development,
which help us better explain the variation in GDPP to avoid ommitted variable bias.
We have decided to take GDP per capita as our dependent variable, as it is a universal
measurement of a country's economic development. As we can see on this histogram, the data is
skewed, as there are a couple of outliers - countries with incredibly high GDP per capita - and a far
larger number of countries with lower GDP. This means that the mean GDPP is inflated, meaning
that we need to use the natural log of GDPP in our regressions not to skew the results. This is
supported by a drastic decrease in standard deviation in proportion to the mean.
Another interesting relationship in the variables is between GDP and youth NEET (not in education,
employment or training). If we regress those variables, 40% of movement of GDP can be explained
by Youth NEET, with a coefficient of -0.08043, meaning that every increase in youth NEET by 1%
leads to a 7.7% decrease in GDP per capita. As we can see on the graph demonstrating this
relationship, there are some outliers to this trend, such as Turkey or Togo.

Another summary statistic we need to look at is the relationship between youth NEET and LFP
(labour force participation), as we want to use them as independent variables in our model, so we
have to ensure there is no multicollinearity.

As we can see on this graph, there is close to no relationship between the variables, which is
supported by running a regression, as on 4.2% of variation in NEET can be explained by LFP.
Therefore, we can use both of these variables in our model without having multicollinearity.

There is a slight relationship between trade and political stability, which is logical as the more
stable your country is, the more likely the other countries are to be willing to trade with you. On
this graph, we can also see the trade outliers such as Malta, Luxembourg and Ireland: small
developed nations that are unable to produce sufficient goods and are therefore forced to trade.
Variable Obs Mean Std. Min Max Varian Std error of Skewn Kurtos
dev ce mean ess is
GDP 109 23897.5 20888. 1217.4 11175 4.36E+ 2000.786 1.31570 5.2408
4 82 46 1.3 08 4 25
Share of youth 105 18.3315 8.9081 4.301 47.532 79.355 1.247623 0.78126 3.2031
not in education 2 71 52 5 39
Govt expenditure 96 4.73387 1.6000 0.3584 8.8095 2.5602 0.163308 0.12517 3.4607
on education 9 85 79 48 72 56 72
Youth Labour 105 42.2294 12.784 21.015 78.302 163.43 0.8693485 0.46111 2.4985
Force 1 33 92 18 84
Participation
Trade 106 83.9747 53.049 23.079 372.27 2814.2 2.817063 2.28927 11.060
6 26 78 14 24 9 49
FDI 108 5.41240 30.487 - 200.83 929.49 2.93367 4.53443 31.811
3 59 101.83 15 3 7 29
31
Exports 106 39.8187 29.003 9.2985 203.12 841.19 5.1526 2.56676 12.636
3 44 59 03 96 3 87

The model which we decided to go

Section 3: Regression analysis

This report conducts a series of multi-regression analyses using the

Ordinary Least Squares (OLS) method. OLS regression estimates the linear
relationship between a dependent variable and one or more independent
variables by minimizing the sum of squared residuals (Wooldridge, 2016).
As our concern is with how much young people expiation the variation of
economic development across the word, our base variables consist of the
following: GDP per capita being our dependent variable, the sum of gross
production of each country divided by its population. Specifically in this
model, we take the natural logarithm of GDP per capita ln(GDP).
The key explanatory youth variables, throughout are Government
expenditure on education (as a % of GDP), Share of youth not in
education, employment, or training (NEET), and Youth labour force
participation rate (ages 15-24, total %). It is worth noting that there are
many other variables that can capture the role of young people in an
economy but for this report our concern is only with these three.
As a result, the structure of the regression section is as follows: Model
1 begins with a baseline regression that includes only the youth-related
variables. Models 2 and 3 adds macroeconomic control variables to
improve model fit and reduce potential omitted variable bias. Models 4
and 5 then introduces a high-income country dummy to account for
inequality of economic development across countries. Splitting the sample
into high-income and non-high-income countries, we test the models on
different sub-samples. Models 6, 7 and 8 then introduce interaction terms
between our explanatory variables and income group to test whether such
impact young people have on GDP per capita significantly differs by
development level.

Gauss-Markov assumptions

Before moving onto the regression, it is important first to explain the Gauss-Markov theorem
which sets out key assumptions for the OLS estimator to be Best Linear Unbiased Estimator
(BLUE). In this report, our regression design accounts for these assumptions as we test and
pre-emptively refine models accordingly. We aim for our models to draw reliable statistical
inferences by considering and doing the following.

The dataset is assumed to be drawn from a random sample, and the model equations are
linear in parameters, satisfying the first two Gauss-Markov conditions.
Heteroskedasticity is when the variance of the error term is not constant across all levels of
the independent variables (Wooldridge, 2016). Testing for some of initial our regressions
produced p-values of less than the 5% significance level - rejecting the null hypothesis that
models are homoscedastic. This can be shown in table 1. To correct for this, we re-estimated
all the models using robust standard errors. Thus, the results in all regressions that are to
follow were ran using the robust command.
Multicollinearity is when two or more independent variables in a regression model are
highly correlated, making it difficult to isolate the individual effect of each predictor on the
dependent variable (Wooldridge, 2016). This can lead to inflated standard errors and
unreliable coefficient estimates. To assess this, we calculated the Variance Inflation Factor
(VIF) for each independent variable. Results showed that most VIF values, including our
youth variables, were close to 1 (below 5), indicating no evidence of problematic
multicollinearity. Table 1 shows each models mean vif value – while models 6 to 8 have high
values compared to previous regressions, given the nature of interactive models, this is to be
expected thus this does not pose a significant problem.
Lastly, the zero conditional mean assumption was assessed using the Ramsey RESET test to
detect potential omitted variable bias (Wooldridge, 2016). All models but one regression
produced p-values above the 5% level, indicating no significant evidence of OVB.
Furthermore, Regressions 2 and 3 included control variables that captured other relevant
influences on GDP, providing additional support for the validity of this assumption. While it
needs to be acknowledged that there is some capacity for reverse causality within the
model, however we have taken precautions of rigorously testing our regressions as well as
adding different control variables to ensure exogeneity within a reasonable degree.
The regression models that failed some of the Gauss-Markov tests will be discussed
individually, however our main findings remain robust, and we will interpret every result
with caution.
 Test Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Heteroskedastici
ty p-value
(before robust)
0.2649 0.0094 0.0095 0.0201 0.2482 0.0005 0.0003 0.0015

Multicollinearity:
mean vif-value
1.09 1.39 1.4 1.85 1.16 5.82 3.38 4.69

OVB (Ramsey
test) p-value
0.44 0.0952 0.1478 0.2873 0.5104 0.0849 0.0849 0.0224
Table 1: Results for each model testing for heteroskedasticity, multicollinearity (vif) and omitted variable bias (Ramsey test).

Variable Model 1 Model 2 Model 3 Model 4 Model 5 (Non- Model 6 Model 7 Model 8
(High-income) high-income)

Government .0719941 -.0773607 -.0850618 -.0312448 -.1202024 -.1159694 -.1121606 -.1176603

spending on
education

Share of youth -.0818929 -.0419818 -.038372 -.016031 -.024102 -.0252988 -.0228704 -.027552
NEET

Youth labour -.0116311 -.0062812 -.0070091 .0094555 -.016094 -.0050974 -.0054242 -.0178698
force
participation
 Trade - .0046598 .0042744 .0022147 .003433 .0031597 .0026231 .002911

.
Urban - .0346108 .0352443 .0070711 .0317084 .0281844 0289353 .0269666
population

FDI - -.0051384 -.004886 -.0007846 -.0074762 -.0038036 -.0039293 -.0034743

Inflation - - -.0196169 -.0498672 -.0120431 -.0100295 -.0112508 -.0144546

High Income - - - - - .4819639 1.045108 -.2259185

Interaction - - - - - .0478928 -.0242368 .0227235

term

B0 11.29738 8.491235 8.5454 9.865731 8.830393 8.567111 8.510617 9.261675

R^2 0.4783 0.7716 0.7786 0.5983 0.6580 0.8284 0.8302 0.8431

Obv. 95 92 91 41 50 91 91 91

Table 2: Results for each model showing coefficient values, R2 and number of observations.
 P-values
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Government 0.252 0.070 0.036 0.541 0.008 0.015 0.005 0.002

spending on
education

Share of youth 0.000 0.000 0.000 0.219 0.034 0.005 0.014 0.006
NEET

Youth labour 0.148 0.235 0.182 0.005 0.102 0.337 0.283 0.070
force
participation

Trade - 0.000 0.001 0.034 0.150 0.008 0.030 0.013

Urban - 0.000 0.000 0.097 0.000 0.000 0.000 0.000

population

FDI - 0.000 0.000 0.618 0.742 0.007 0.003 0.015

Inflation - - 0.088 0.001 0.298 0.389 0.331 0.167

High Income - - - - - 0.266 0.001 0.651

Interaction - - - - - 0.557 0.242 0.040

term
Table 3: P-value results for each coefficient in each regression model.
 Model Description
1 Baseline - youth related variables only
2 Adding control variables
3 Adding more control variables
4 High income countries only
5 Non-high-income countries only
6 Income group & education spending interactive term
7 Income group & NEET interactive term
8 Income group & Youth LF Participation interactive
term
Progression of models/ regressions.

Regression 1:

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 ( NEET )+ ^β 3 (YouthLFP)+û

Model 1 establishes the foundational relationship between youth-related variables and log
GDP per capita. The results indicate that the NEET variable is strongly and negatively
associated with economic development. A one percentage point increase in the share of
youth who are NEET is associated with an approximate 8.2% decrease in GDP per capita, a
statistically significant effect (p = 0.000). This suggests that disengaged youth represent a
substantial economic cost, consistent with broader findings in the development literature??.
Youth labour force participation is also negatively associated with ln GDP, though the
relationship is weaker. The coefficient implies that a one percentage point increase in
participation is associated with a 1.2% reduction in GDP per capita, marginally significant at
the 10% level (p = 0.087). This counterintuitive result may reflect the fact that youth
participation rates are often higher in lower-income economies due to weaker education
systems and informal labour markets, and thus may not directly reflect productive
employment however we will explore this in a later regression model.
Government spending on education shows a positive relationship but statistically
insignificant (p = 0.191), thus there is insufficient evidence to conclude that higher education
expenditure alone drives higher GDP per capita.
The model explains approximately 47.8% of the variation in log GDP (R² = 0.4783), which is
relatively strong for a baseline specification with only three predictors. This serves as a
useful starting point for the analysis, to be extended through the inclusion of other controls
variables in subsequent models.

Regression 2:

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 (NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6(FDI )+ û

Model 2 builds upon the initial regression by incorporating macroeconomic control variables
that are known to influence GDP per capita: inflation, trade, and foreign direct investment
(FDI). These factors provide a broader context for understanding income differences across
countries. For example, FDI act as an injection into the economy, meaning that countries
with greater inward investment may generate higher levels of income. By including such
control variables, this model helps isolate the independent effects of the youth-related
variables, making their estimated coefficients more accurate. In other words, FDI may now
be capturing some of the variation previously attributed to the youth variables.
The share of youth NEET remains a key negative determinant of economic development.
Although the coefficient decreases slightly from Model 1, it remains highly significant (p =
0.000), with a one-point increase in NEET corresponding to a 4.2% decrease in GDP per
capita.
Government expenditure on education now shows a negative coefficient -0.0774 with
marginally significant p-value (0.064) at a 10% confidence level. While unexpected, this
result may suggest inefficiencies in how education funds are allocated or reflect a time lag
between spending and returns to output.
Youth labour force participation, while still negatively signed, is no longer statistically
significant (p = 0.181), and its effect weakens relative to the baseline model. This indicates
that its earlier marginal significance may have been confounded by omitted macroeconomic
factors.
Overall, the addition of control variables improves model, with the R² rising to 0.772. This
highlights the importance of control variables in explaining differences in GDP per capita
across countries. *Should I also briefly explain the coefficients of the control variables?

Regression 3:

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 ( NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6( FDI )+ ^β 7

Model 3 explains approximately 78% of the variation in log GDP per capita (R² = 0.7786),
adding inflation as an additional control variable. This improved our model with the
following changes. The NEET variable coefficient weakens from -4.2% to -3.8%, though it
remains statistically significant (p = 0.000). The effect of government expenditure on
education becomes mildly stronger (to -8.5% coefficient) and now significant at the 5% level
(p = 0.036). The coefficient on youth labour force participation remains roughly the same (at
-0.007) and remains statistically insignificant (p = 0.182).
Including inflation as a control helps capture macroeconomic conditions that influence GDP,
and its coefficient is both positive and significant, suggesting a meaningful association with
log GDP per capita in this model.

Regression 4 and 5: High income vs non-high-income split sample

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 ( NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6( FDI )+ ^β 7

To examine whether the relationship between youth-related variables and GDP per capita
differs by development level, we now split the sample into high-income countries (Model 4)
and non-high-income countries (Model 5). We do this by adding high-income country
dummy variable to account for structural income group differences across the sample,
where high-income = 1 and non-high-income = 0. This allows for a clearer picture of how
these dynamics vary across economic contexts, particularly since earlier results suggested
income group may mediate these effects.
  Government Expenditure on Education:
In non-high-income countries (Model 5), the coefficient is negative and statistically
significant (-0.120, p = 0.008), suggesting that higher education spending is
associated with lower GDP per capita. This could point to reverse causality, where
countries with lower income levels increase education spending in response to
underdevelopment. It may also reflect inefficiencies in education systems, or the
delayed retunes to human capital investments.
In high-income countries (Model 4), however, the effect is much smaller and
statistically insignificant (-0.031, p = 0.541), consistent with the idea that diminishing
returns to education investment may exist in already-developed systems.

 Share of youth NEET:

In non-high-income countries, the NEET variable remains negatively associated with
GDP per capita (−0.024), and is statistically significant (p = 0.034), reinforcing the idea
that youth disengagement is particularly harmful in developing economies. In high-
income countries, by contrast, the coefficient is smaller (−0.016) and not statistically
significant (p = 0.219), suggesting that the economic consequences of youth
disengagement are less severe - likely due to stronger labour markets or better
education systems (university) that reduce the impact.

 Youth Labour Force Participation:

This variable shows a positive and statistically significant relationship with GDP per
capita in high-income countries (0.9%, p = 0.005), suggesting that increased youth
labour force participation contributes positively to economic output in advanced
economies. This likely reflects better employment opportunities, and ability to turn
young peoples inputs into more outputs. In contrast to non-high-income countries,
the coefficient is negative (−0.016) and not statistically significant (p = 0.102),
supporting the notion that youth participation in these contexts may reflect less
informal employment, which does not necessarily contribute to economic growth.

Regression 6, 7 and 8: Interaction terms

While Models 4 and 5 suggested that the impact of youth-related variables may differ by
income group, these regressions did not formally test whether those differences were
statistically significant. Models 6, 7, and 8 address this by including interaction terms
between the income group dummy and each of the three youth-related variables. These
models retain the full sample and enable direct tests of whether the effects of education
spending, youth disengagement, and labour force participation vary significantly between
high- and non-high-income countries. To clarify, the individual youth variable coefficients
under these following models represent the dummy variable equalling to 0, which is non-
high income our case. And the interaction coefficient represents the difference in slope of
high-income countries to non-high income.

Model 6: Government Expenditure on Education

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 (NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6(FDI )+ ^β 7

The coefficient on education spending is negative and statistically
significant (−0.116, p = 0.015), supporting earlier findings from Model 5
that higher spending is associated with lower GDP per capita in non-high-
income countries. This again may reflect inefficient allocation of
resources, or long-term investments that take time to produce growth.
The interaction term with high-income status is positive but not
statistically significant (p = 0.557), indicating no strong evidence that the
effect of education spending differs meaningfully across income groups.
This aligns with Model 4, where the coefficient was smaller and not
significant, suggesting that marginal returns to education spending may
level off in more developed contexts, but not in a way that’s statistically
distinguishable in this model.
However, our Gauss-Markov assumptions earlier suggested potential
multicollinearity for this model with a VIF value of 24.47 for the interaction
term and 19.74 for the income group dummy – the rest being close to 1 –
thus inflating the mean VIF value. While this is expected for models with
interaction terms it can still reduce standard errors and therefore the
accuracy of our coefficient estimates. But this does not mean our model is
consequently biased.

Model 7: Share of Youth Not in Education, Employment, or

Training

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 (NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6(FDI )+ ^β 7

The main coefficient on NEET is negative and statistically

significant (−0.029, p = 0.014), consistent with earlier findings that youth
disengagement is particularly harmful in less developed economic
contexts.
The interaction term, however, is not statistically significant (−0.024, p =
0.242), suggesting that there is no clear evidence that the relationship
between NEET and GDP differs significantly between high-income and
non-high-income countries. While the negative sign implies the effect may
be even stronger in high-income settings, the lack of statistical
significance means we cannot confidently conclude that this relationship
varies by income group. Overall, Model 7 supports the notion that youth
disengagement is economically damaging in lower income countries but
offers no strong evidence that the effect is different in higher income
countries.

Model 8: Youth Labour Force Participation

GDP(Ŷ )= ^β 0+ β^ 1(GovtEduc £)+ ^β 2 ( NEET )+ ^β 3 (YouthLFP)+ ^β 4 (Trade)+ ^β 5 (UrbanPop)+ ^β 6( FDI )+ ^β 7

The base coefficient for youth labour force participation is negative and
marginally significant (−0.018, p = 0.070), reinforcing the pattern seen in
model 5. This may reflect informal/ low quality, or necessity-based
employment, where young people enter the workforce early due to
economic pressure, often at the expense of education.
In contrast the interaction term is positive and statistically
significant (0.0227, p = 0.040), indicating that the effect is significantly
different in high-income countries. This supports earlier findings from
Model 4, where youth labour force participation had a positive and
significant relationship with GDP. These results suggest that developed
economies may better integrate young workers into productive roles, and
produce more output from them, and consequently GDP. However,
referring back to the Ramsey test for model 8 which produced a p-value of
0.0224 (less than 5%), suggests that OVB is a concern. Despite it’s
controls, other factors may also be affecting GDP per capita, meaning any
statistical inference must be taken with caution.

Overall, these interaction models confirm that the effects of our youth
variables are not uniform across all countries. While education spending
continues to show a negative and statistically significant association with
GDP per capita in non-high-income countries, there is no strong
evidence that this effect differs significantly in high-income settings.
Similarly, the NEET variable, although consistent in negative associated
with GDP, the interaction model does not provide statistically significant
evidence that this relationship varies across income groups. However,
labour force participation shows a strong and statistically meaningful
difference, with its impact on GDP significantly more positive in high-
income countries.
Section 3: Analysis of regression results

Goodness of fit

Model Description Adjuste

d R2
1 Baseline - youth related variables only 0.4783
2 Adding control variables 0.7716
3 Adding more control variables 0.7786
4 High income countries only 0.5983
5 Non-high-income countries only 0.6580
6 Income group & education spending interactive term 0.8284
7 Income group & NEET interactive term 0.8097
8 Income group & Youth LF Participation interactive 0.8424
term

The progression of the regressions shows a clear improvement in model fit

as more variables and interaction terms are added. For example, model 1
began by initially explaining 48% of the variation of DGP per capita but
increased to 78% in model 3 with the introduction of control variables.
When the sample is split by income group in Model 4 and Model 5, R²
drops to 0.5983 and 0.6580, respectively. This decline is expected due to
the reduction in sample size and variation, though both models still offer
valuable insight into how youth-related variables behave across different
income groups.
The inclusion of interaction terms in Models 6 - 8 retains the full sample
while allowing for differentiated effects by income group. These models
show the highest R² values overall.

F-tests

While previous models assessed the individual effects of youth-related

variables on GDP per capita, it is also important to evaluate whether these
variables are jointly significant - whether they contribute meaningfully to
explaining income differences when considered together. To do this, F-
tests are used to test the combined explanatory power of groups of youth
indicators within selected models (Wooldride, 2016). This provides a more
robust check of whether youth-related factors, even if only marginally or
individually significant, collectively influence GDP in a statistically
meaningful way.

F-test for Model 3

Model 3

H 0 : β 1= β 2=β 3=0
H 1: β 1≠ 0∨β 2 ≠ 0∨β 3 ≠ 0
 F stat :7.07
P−value :0.0003

In Model 3, the F-test assesses whether the three youth-related variables

are jointly significant in explaining variation in GDP per capita. The test
produces an F-statistic of 7.07 with a p-value of 0.0003, which is below the
5% significance level.
This result provides strong evidence that the youth variables, when
considered together, have a statistically significant joint effect on GDP per
capita. Even if some of the variables are only marginally significant on
their own, the F-test confirms that as a group, they contribute
meaningfully to the explanatory power of the model.

F-test for Model 6, 7 and 8

Model 6 Model 7 Model 8

H 0 : β 1= β 9=0 H 0 : β 2=β 9=0 H 0 : β 3=β 9=0

H 1: β 1≠ 0∨β 9 ≠ 0 H 1: β 2≠ 0∨β 9≠ 0 H 1: β 3 ≠0∨β 9≠ 0

F stat :3.95 F stat :4.92 F stat :2.19

P−value :0.0231 P−value :0.0096 P−value :0.1181

In Model 6, the F-test produces an F-statistic of 3.95 and a p-value of

0.0231, which is below the 5% significance level. This means we reject the
null hypothesis that education spending and its interaction with income
group have no joint effect on GDP per capita. Instead, our model predicts
that these variables, when considered together, are statistically significant
and help explain differences in GDP.

Similarly, Model 7 F-test gives an F-statistic of 4.92 and a p-value of

0.0096, which is also below the 5% significance level. Again, we reject the
null hypothesis which suggests that the result of the NEET variable and its
interaction with income group are jointly significant in explaining GDP per
capita.

However, in the model 8 F-test produces an F-statistic of 2.19 and a p-

value of 0.1181, which is above the 0.05. Thus, here we fail to reject the
null hypothesis, meaning youth labour force participation and its
interaction with income group are not jointly significant in the model – our
results in regression 8 must be therefore treated cautious
Conclusion of results:

As a result of the regression analysis, our findings can be summarised as

follows. This can prove to be useful to inform policy maker and wider
policy implementation surrounding young people.
*mention that corelation ≠ causation

 Non-high-income countries should prioritise reducing NEET rates, as

the economic costs of youth disengagement appears to be more
severe than in high income countries. Both the regression and F-test
results confirm its strong, negative association with GDP per capita.
*include specific stats?
 High-income countries may place less emphasis on NEET reduction,
as the data show no consistent or statistically significant
relationship with GDP per capita in these settings.
 Non-high-income countries should be cautious of high youth labour
force participation, which may reflect underemployment or early exit
of school…

 High-income countries show a more positive association between

youth labour force participation and GDP per capita, suggesting that
more developed economies may better integrate young people into
productive roles. However, these results should be interpreted
carefully, particularly in light of the F-test, which did not confirm
joint significance.
 Non-high-income countries should also evaluate the efficiency and
impact of education spending, as the regression analysis shows a
statistically significant negative relationship with GDP. This could
reflect inefficiencies or delays in returns.
 High-income countries show no strong evidence that the effect of
education spending significantly differs, and spending itself does not
display a robust impact on GDP, suggesting that education policy
alone may not drive further income gains at higher development
levels.

Abigeal Eludire (2023). THE IMPACT OF LABOR FORCE ON ECONOMIC GROWTH: A

GLOBAL VIEW. [online] ResearchGate. Available at:
https://www.researchgate.net/publication/376557034_THE_IMPACT_OF_LABOR_FORCE
_ON_ECONOMIC_GROWTH_A_GLOBAL_VIEW.

Eurostat (2021). Unemployment statistics and beyond - Statistics Explained. [online]

ec.europa.eu. Available at: https://ec.europa.eu/eurostat/statistics-explained/index.php?
title=Unemployment_statistics_and_beyond.

Jabr, F. (2021). John A. Long - Publications List. Publicationslist.org, 14(6).

Najaf-zada, F. (2024). The Impact of Education Spending on Economic Growth. [online]

Ssrn.com. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4705468.

Pologeorgis, N.A. (2023). Employability, the Labor Force, and the Economy. [online]
Investopedia. Available at:
https://www.investopedia.com/articles/economics/12/employability-labor-force-economy.asp