Kumar and Woo (2010) are also critical
of R&R’s study as it does not take into account other growth determinants
and they claim that it also face reverse causality issues. When conducting
their own analysis, on a panel of advanced and emerging market economies over
the period of 1970-2007, on the impact of high public debt on long run economic
growth, they found that a 10 percentage point increase in debt-to-GDP ratio is
correlated with a decrease of 0.2 percentage point of economic growth annually.
Kumar and Woo (2010) pay particular attention to issues including outliers,
causality and endogeneity, one of which they critiqued in R’s study.
Further support for the debt threshold is also evident in Cecchetti, Mohanty
and Zampolli (2011)’s study where they discovered that when public debt is 85
percent of GDP, economic growth past this level begin to decline. “High debt is
bad for growth.” (Cecchetti, Mohanty and Zampolli, 2011).

Other studies have proposed
alternatives and extensions of R’s debt regimes. Extending R&R’s
Study, Herdon, Ash and Pollin (2014) revise the four debt-to-GDP ratio
categories and extend it by adding two additional categories. 90-120 percent
category, where results portrayed 2.4 percent average real GDP growth, and
greater-than-120 percent category, where results show a 1.6 percent average
real GDP growth. Swamy (2015)’s study seeks to compare dataset by utilising
similar debt regimes proposed by R&R while estimating the thresholds
endogenously utilising Hansen (1999) panel threshold regression model. Results
show that the study finds the debt thresholds to vary in the range of 84-114
percent of GDP for a sample of ranging from 10 to 50 years. While this study
supports the relationship between debt threshold and growth, threshold levels
need to be interpreted with great caution as the results are limited to the
period of study, the number of countries and the types of countries that are
considered (Swamy, 2015).

Chudik et al. (2015) studied the long
run impact of public debt on economic growth and analyses if the debt-growth
relationship will vary on the different level of indebtness. Examined 40
developed and developing countries over the period of 1965-2010 period. They
allowed for country-specific heterogeneity in dynamics, error variances as well
as cross-country correlations. Chudik et al (2015) assume homogenous threshold
parameters. They concluded that their empirical analysis found support for the
debt-threshold effects of 60-80 percent for the full sample. The debt-to-GDP
thresholds were lower for the developing economies when compared to the
advanced economies (Chudik et al, 2015).

The findings of a further study,
however, analyses the short run impact of debt-to-GDP ratios on GDP growth;
Baum, Checherita-Westphal and Rother (2013). They found a positive short-run
impact of debt on growth up to a debt threshold of 67 percent, where growth
declines thereafter. Baum, Checherita-Westphal and Rother (2013) employed a
dynamic threshold panel methodology on 12 Euro area countries between the
period of 1990-2010. A debt-to-GDP threshold level of 67 percent reflects
economic growth of around 0 percent.  Their results are robust in
the dynamic and non-dynamic threshold models and only deviate when data
accounts for before 1990 and the crisis years of 2008-2010.  For high debt-to-GDP ratios of above 95
percent, additional debt has a negative impact on economic growth.

Both studies by Chudik et al (2015)
and Baum, Checherita-Westphal and Rother (2013) found that debt thresholds
levels past 60 – 67 percent were seen to have adverse effects on economic
growth, despite the short run and long run empirical study difference. Chudik
et al (2015) claim the inclusion of inflation in empirical analysis is
important as deficit financing through domestic money creation in developing
economies may be more of an impacted factor in altering growth rather than
government debt.  Kaur and Mukherjee
(2012) also find similar results to Chudik et al (2015) and Baum,
Checherita-Westphal and Rother (2013) where threshold levels of debt above 61
percent is found to be detrimental to the upward relationship between the level
of debt and economic growth in India.

Égert (2015) identifies the
public debt thresholds endogenously and extends R&R’s time coverage back to
1790. Empirical analysis shows that the nonlinear relation between debt and
growth is not robust and should always consider that the results of the size of
thresholds vary across countries and across time. His results, which are
inconsistent with R’s (2010) claim, is that the negative coefficient in
the upper debt regime remains lower compared to the one in the middle debt
regime. The coefficient estimates indicate that public debt higher than 20
percent of GDP is associated with lower growth, others at 60 percent debt
threshold. However, they do support the claim that public debt beyond the 90
percent public debt is associated with significantly lower economic growth.
When looking at their threshold results, consideration should be taken into
account as they may vary across countries and across time.

Pescatori, Sandri and Simon (2014)
imposed a similar approach to R&R’s study, where analysis of the short run
relationship between debt and growth was explored. Supports R’s findings
as their analysis shows that GDP growth averages around 2 percent in countries
with below 90 percent threshold and -2 percent in countries whose debt ratio
increases above the 90 percent threshold level. However, omitted variables
could be the cause that results in debt higher than the 90 percent debt
threshold. Differently analysed from R&R, they focus on the long term
relationship between todays stock of debt over GDP and GDP growth. Utilizing a
long term analysis, they try to reduce reverse causality effects and potential
omitted variable bias. Pescatori, Sandri and Simon (2014) found that countries
that exceed the 140 percent debt threshold correlate to an average debt ratio
of 130 percent. Also supporting the idea that debt threshold beyond 90 percent
is detrimental to growth is Bilan and Ihnatov (2015) who utilized panel data
estimation methods on 33 EU Member States from 1990-2011. They identified that
the maximum debt threshold was 94 percent of GDP. Beyond the 94 percent
threshold, Bilan and Ihnatov (2015) claim that negative effects on economic
growth will be evident due to higher interest rates and sever budgetary
measures. Conversely, main limitations of this research stems from the lack of
comparable datasets for the developing EU Member states.

Kaur and Mukherjee (2012), expressed
in their paper that Cordella, Ricci and Ruiz-Arranz’s (2005) empirical study
also established that public debt beyond a certain threshold is negatively
correlated with economic growth. 
Cordella, Ricci and Ruiz-Arranz (2005) explores how the debt-growth
relationship varies with indebtedness level in a panel of developing countries,
HIPC, where indebtness has no effect on growth or investment, and non-HIPC,
where they found that high debt relates to a negative marginal effect on per
capita growth. Found the debt irrelevance threshold at around 70-80 percent of
GDP. But this highly depends on the individual country characteristics. When
debt exceeds 15-30 percent of GDP, countries with better institutions, policies
and easier access to private capital investment, face a debt overhang at this
threshold and debt irrelevance threshold is achieved at
70-80 percent of GDP (Cordella, Ricci and Ruiz-Arranz, 2005). Much worse
economic conditioned countries experience a debt threshold level at 0–20
percent of GDP and debt irrelevance at 15-53 percent of GDP. However, must be
cautious as these threshold levels are affected by considerable estimation
uncertainty and might reflect external factors.

Eberhardt and
Presbitero (2013) studied a long run relationship between public debt and
growth in a large panel of countries. They analysed cross country and
within-country effects of nonlinearity due to previous literature which claim
that policies from debt threshold analysis are applicable to all countries.
However, this is misleading and “growth-retarding at worst”. They concluded
that the commonly found 90 percent debt threshold, which roots from R
(2010) study, are outcomes of empirical misspecifications from pooled instead
of heterogeneous estimates that create misinterpretation of results provided.
Furthermore, Panizza and Presbitero (2013) surveyed theoretical and empirical
literature that examines the relationship between public debt and growth in
advanced economies. They conclude that the common debt threshold above which
growth collapses is not robust and that there are also causality issues to be
aware of. Suggests that future research should focus on cross-country
heterogeneity, although it may lead to large biases on why public debt may
hinder economic growth.

2.    Research Question

Previous literature studies have
explicitly stated the importance of not having a ‘one-size-fits-all’ policy for
debt crises (see for example Eberhardt and Presbitero’s
(2015), Presbitero (2008), Holland (2007) etc.). The aim of this study is to support policy formulation,
to prevent countries from defaulting or a debt-overhang by finding evidence and
examining the nonlinearity relationship between debt and economic growth. The current
empirical debt-growth literature on this topic have not yet explicitly explored
the 53 countries as a whole, under the classification of Lower-Middle Income
Economies, according to the World Bank Classification. Therefore, this paper
aims to analyse this country grouping and fills the gap in the literature.
Furthermore, this paper explores the point where debt becomes detrimental to
growth and questions whether the threshold regimes proposed in R&R’s (2010)
study is applicable for the ‘Lower-middle Income’ country grouping.
Specifically, utilizing the threshold estimation method developed by Hansen
(1999). This is most applicable as it explores the nonlinear effects between
the relationship of economic growth and debt and estimates the thresholds
endogenously. This is done by the application of two, dynamic and non-dynamic,
panel threshold regression estimations on the 53 lower-middle income economies.