In the recent past, Ghana's public debt became unsustainable, necessitating appropriate debt management strategies to restore economic stability. This study examines the nexus between public debt and macroeconomic performance, focusing on the government's fiscal vulnerabilities as key influencing factors, to reveal how government borrowing affects economic recovery policies.
We employ a hierarchical regression model and utilize the least squares method (linear) to analyze time series data (1983–2020) from the World Bank, IMF and Bank of Ghana databases. We further employ a Bayesian linear regression model to enhance the robustness of the analysis.
The results depict that public debt relates negatively with GDP, inflation and trade balance, but it relates positively with unemployment. All fiscal vulnerability indicators moderate the association between public debt and both GDP and inflation. Furthermore, all indicators (except external debt to exports and net international reserves to external debt ratios) moderate the relationship between public debt and unemployment. Only debt to domestic revenue ratio and interests to domestic revenue ratio moderate the association between public debt and trade balance.
The interaction effects reveal intricate relationships between public debt and macroeconomic performance. Our study innovatively applies Bayesian analysis to examine these interactions, offering a more robust and nuanced understanding of the complex dynamics at play and shedding new light on the relationships driving economic outcomes. This informs policymakers’ development of more effective policies that account for these nuanced relationships.
1. Introduction
Fiscal vulnerability, which refers to the susceptibility of a country’s fiscal system to shocks, stressors or uncertainties that could negatively impact its ability to achieve fiscal sustainability, maintain economic stability and promote macroeconomic performance, has become a critical issue following the global debt crisis in the early 2000s (INTOSAI, 2010; Donayre and Taivan, 2017), due to the excessive borrowing post Second World War by both advanced and emerging countries to finance fiscal balances (Hilton, 2021). The debt crisis was worse among emerging countries in sub-Saharan Africa, particularly Ghana, leading to a declaration of the country as a heavily indebted poor country (HIPC) in 2001 (Hilton et al., 2025). Currently, Ghana is at a crossroad of a severe economic crisis as its public debt stock hits a record high of GH¢742 billion as of June 2024 (Ministry of Finance, 2024). Many economists expressed worry and called on the government to halt or minimize its appetite for borrowing. However, some economists argued that government can borrow cautiously since it still has huge fiscal deficits to finance. Despite the successful attempts to introduce new taxes such as the COVID-19 recovery levy, sanitation levy, electronic transfer levy (e-levy), betting tax, emission levy and increment in petroleum taxes, it appears government still need to borrow to support its budgets. This reality concurs with those who posit that the Ghanaian economy is in crisis, so more revenue must be generated to salvage the situation. It means that though the public debt stock to GDP is unsustainably high, the size of the economy may accommodate additional borrowing to finance public spending.
The accumulation of public debt without proper management can adversely affect macroeconomic performance (Adom, 2016; Joy and Panda, 2020). The caution then must be to devise strategies to efficiently manage debt levels to ensure that the economy does not sink or crash further, which will create an economic recession, and consequently affect the people’s livelihood. In managing debt efficiently, researchers (e.g. Cochrane, 2011; Castro et al., 2015) postulate that fiscal vulnerability is a major concern because it exposes a country’s financial market to international shocks that can obstruct macroeconomic performance. For instance, high fiscal vulnerability can amplify the negative effects of public debt on macroeconomic performance, making it more challenging for a country to achieve economic stability; whereas low fiscal vulnerability can mitigate the potential benefits of public debt, such as financing productive investments, reducing the associated risks. Hence, attention must be paid to fiscal vulnerability indicators and how they interfere with the relationship between debt and macroeconomic performance. Castro et al. (2015) and Cochrane’s (2011) view underscores earlier strategies initiated by the IMF during the 1980s debt crisis. The strategies included a definition of whether a country is vulnerable to debt crisis and if so, to which extent. The attention was on emerging market economies, whose economic performance depends typically on external financing and other capital flows (INTOSAI, 2010).
Despite the plethora of literature on public debt, debt management and macroeconomic performance, empirical evidence on how fiscal vulnerability can either increase or decrease the impact of public debt on macroeconomic performance is unavailable, particularly from the Ghanaian perspective. This gap could be due to overreliance on the established relationship between public debt and macroeconomic performance and treatment of the fiscal vulnerability indicators as endogenous or exogenous variables. This approach does not indicate how fiscal vulnerability indicators, which are critical in debt management, could enhance or limit the unidirectional or bidirectional causal relationship between public debt and macroeconomic performance. It is, therefore, not empirically proven whether the rise or fall in public debt and macroeconomic performance could be interfered by the fiscal vulnerability indicators as hidden influencers.
Thus, this study investigates the interaction effect of fiscal vulnerability on the link between public debt and macroeconomic performance. This has been done by using historical data (1983–2020) and econometric methods such as least squares – OLS (frequentist approach) and Bayesian linear regression model (non-frequentist approach) in a hierarchical regression modeling. These methods address the inherent weakness of adopting one approach, thereby making the findings more robust, accurate, reliable and replicable (Briggs, 2023). The use of these methods closes a significant methodological gap in existing literature on public debt and macroeconomic performance, which largely used only the frequentist approach, raising substantial doubt about those findings in contemporary times (Briggs, 2023). The outcome of this study provides empirical evidence of how a careful consideration of fiscal vulnerability in debt management could possibly stimulate economic recovery, serving as a precursor to other developing countries in the new world of COVID-19. Furthermore, the findings give a sense of direction regarding debt management in economic recovery programs. Consequently, this study is novel as it indicates whether the mixed findings on public debt and macroeconomic performance could be explained with fiscal vulnerability indicators and whether the management of the fiscal vulnerability indicators should be of vital interest to any government in implementing strategies to manage its debt level efficiently.
The paper is structured as follows. The Introduction is followed by section 2 on Literature review entailing theoretical underpinnings, trends in Ghana’s public debt, macroeconomic performance, fiscal vulnerability, public debt and macroeconomic performance and public debt, macroeconomic performance and fiscal vulnerability, research gap and conceptual framework. Next, section 3 describes the Methodology including model estimation, model specification, data and source and unit root test. Section 4 presents the Results and discussion involving the preliminary and main results. Section 5 presents the Conclusion.
2. Literature review
2.1 Theoretical underpinnings
The following schools of thought and theories provide useful insights on the relationships between public debt, fiscal vulnerability and macroeconomic performance. We consider them relevant in explaining the study variables, thereby serving as a theoretical basis for this study. To begin with, the classical school of thought kicks against public debt, arguing that it thwarts economic performance since it limits the financial discipline of the budget process and the private sector’s access to credit (Broner et al., 2014). They claim that external debt crowds out economic performance by making it unfavorable for private investment and driving away potential foreign investors (Krugman, 1988; Diamond, 1965; Modigliani, 1961).
On their part, the Keynesian economists argue that funding public expenditures with debt has a fiscal multiplier effect on national output (Elmendorf and Mankiw, 1999). Propositions of this school of thought are premised on the “law of increasing state activity”, which states that increased government spending enhances the domestic economic activity and crowds in private investment (Ncanywa and Masoga, 2018). This implies that public debt take out cash from private investors but does not impact consumption since the borrowings are reintroduced into the economy to upturn total demand via wages, salaries and capital expenses (Onogbosele and Ben, 2016). Therefore, this school of thought overlooked the challenge of using tax cuts or borrowing to fund budget deficits (Nwannebuike et al., 2016). They advocate for recurrent government interference to lift aggregate demand, jobs and production through domestic or external borrowing (Nwannebuike et al., 2016).
The Ricardian equivalence hypothesis (REH) states that debt does not influence macroeconomic performance (Barro, 1979, 1990; Ricardo, 1951). The underpinning assumption of REH is that variants in government revenue and expenditure are equalized by variations in private savings (Kourtellos et al., 2013). Hence, irrespective of financing government deficits with a tax increase or borrowing, the total effect on demand is the same (Ricardo, 1951). The proponents elaborate that households will increase their earnings through the purchase of bonds issued by the government so that potential tax will allow debt repayment (Afzal, 2012; Barro, 1979, 1990; Ricardo, 1951). The proponents add that if a government cut taxes to finance budget deficits with loans, individuals are rational to upturn consumption knowing that in the future the government will impose taxes to payback debts; therefore, taxation or debt does not have permanent effect on the economy (Onogbosele and Ben, 2016).
The fiscal theory of price level (FTPL) explains the relationship between fiscal policy and inflation. It assumes that fiscal policy is the primary driver of inflation (Leeper, 1991). It posits that public debt is not inherently problematic, as long as the government’s fiscal policy is sustainable (Sims, 1994). It further states that the price level is determined by the government’s fiscal policy, rather than by monetary policy or other factors (Woodford, 1995). Based on these assumptions, the FTPL explains the mechanisms by which fiscal policy influences aggregate demand, which in turn affects the price level (Blanchard, 2000). It implies that if the government runs a large budget deficit, it will lead to an increase in the price level (Leeper, 1991). On the other hand, if the government’s fiscal policy is sustainable, it will not lead to an increase in the price level (Sims, 1994). In this respect, the FTPL suggests that monetary policy is not effective in controlling inflation, as the central bank cannot control the price level independently of fiscal policy (Sims, 1994). Meanwhile, based on the Ricardian equivalence, the FTPL assumes that taxpayers are forward-looking and adjust their behavior in response to changes in public debt (Barro, 1974). This means that an increase in public debt will lead to an increase in private saving, which offsets the effects of the debt on aggregate demand. Though the FTPL has been criticized for oversimplifying the relationship between fiscal policy and inflation (Krugman, 1999), it provides unique perspective on the relationship between fiscal policy and inflation, offering useful insights regarding the potential association between public debt and inflation (a component of macroeconomic performance).
Another important theory to consider with respect to trade balance is the intertemporal budget constraint theory, which states that a country’s trade balance is influenced by its fiscal policy and debt dynamics (accumulation and servicing of public debt) (Obstfeld and Rogoff, 1995). This theory suggests that a country’s trade balance is determined by its intertemporal budget constraint, which reflects the present value of its future trade balances. In the context of this study, this theory suggests that a country’s trade balance can be affected by its fiscal policy and debt dynamics in three ways: (1) through fiscal expansion, where an increase in government spending or a cut in taxes can lead to an increase in aggregate demand, which can cause the trade balance to deteriorate (Blanchard, 2000); (2) through debt accumulation, where an increase in public debt can lead to an increase in interest rates, which can cause the exchange rate to appreciate, making exports more expensive and imports cheaper, leading to a deterioration in the trade balance (Obstfeld and Rogoff, 1995) and (3) through fiscal vulnerability, where a country with high levels of public debt and fiscal vulnerability may experience a deterioration in its trade balance, as investors become less confident in the country’s ability to service its debt (Blanchard, 2000). Being an important component of macroeconomic performance, trade balance affects the overall balance of payments and the exchange rate, which in turn, affects the competitiveness of exports and the cost of imports. Thus, this theory is relevant to the topic because it highlights the importance of fiscal policy and debt dynamics in determining a country’s trade balance, which is a key component of macroeconomic performance.
Next is the natural rate of unemployment theory, which posits that there is a natural rate of unemployment that an economy tends towards in the long run (Phelps, 1967; Friedman, 1968). This natural rate is determined by structural factors, such as labor market institutions, demographics and technological change. This theory suggests that attempts to reduce unemployment below its natural rate through expansionary monetary or fiscal policy will ultimately lead to higher inflation, rather than lower unemployment (Friedman, 1968). In the context of this study, this theory provides a useful framework for understanding the relationship between public debt, fiscal vulnerability and unemployment. For instance, if a country’s fiscal policy is unsustainable, it may lead to higher inflation and higher unemployment in the long run, as the economy adjusts to its natural rate of unemployment.
2.2 Trend in Ghana’s public debt
After recording the first public debt in 1963, debt-financing became a part of the Ghanaian government revenue mobilization measures to fund its expenditures, leading to debt accumulation annually. By 1994, the debt stock rose to 37.06% of GDP from 9.37% of GDP in 1984 and hit 55.76% of GDP in 2000, then 79.19% in 2001, causing Ghana to be named as HIPC (World Bank, 2004).
After the implementation of several economic recovery policies such the HIPC initiatives and IMF bailouts, the debt stock dropped from 79.19% in 2001 to 22.64% in 2008. By 2016, the debt level rose from 22.64% in 2008 to 54.83% of GDP and steadily rose again to 92.4% as of December 2022, crossing the threshold 55% of GDP. Following the 2023 IMF bailout program, Ghana’s debt ratio declined from 92.4% in December 2022 to 72.5% as of December 2023, but the nominal debt rose from GH¢342.2 billon to Gh¢610 billon (Ministry of Finance, 2024). The trend of the rise and fall in the debt level between 1983 and 2020 is presented in Figure 1 below. From the graph, the highest debt level is 79.19%, recorded in 2001, and the lowest is 8.38% in 1983. On average, the debt to GDP ratio between 1983 and 2020 is 38.29%. It is worth noting that the main drivers for the rise in the public debt of Ghana are infrastructure expenditures (such as the construction of roads, hospitals, schools, factors and provision of social amenities to the citizenry), insufficient tax revenue to meet statutory obligations and depreciation of the cedi. The massive upsurge in the debt stock in recent years is attributed to the cedi depreciation, the financial sector bailout between 2017 and 2019, and the impact of COVID-19 (BoG, 2021).
The horizontal axis represents years and ranges from 1985 to 2020 in increments of 5 units. The vertical axis is labeled and ranges from 0 to 80 in increments of 10 units. The graph shows 38 bars. The data for the bars are as follows: 1983: 8.34. 1984: 9.04. 1985: 13.7. 1986: 15.1. 1987: 19.53. 1988: 28.63. 1989: 25.12. 1990: 24.42. 1991: 23.96. 1992: 24.42. 1993: 27. 1994: 37. 1995: 55.65. 1996: 49.36. 1997: 44.4. 1998: 46.56. 1999: 41. 2000: 56. 2001: 79. 2002: 62. 2003: 58. 2004: 53. 2005: 41.2. 2006: 33.7. 2007: 18.36. 2008: 22.5. 2009: 25.1. 2010: 27.2. 2011: 35.1. 2012: 31.4. 2013: 35.6. 2014: 43. 2015: 51.2. 2016: 55. 2017: 57.5. 2018: 58.6. 2019: 59.3. 2020: 63.11. Note: All numerical data values are approximated.Ghana’s public debt history (1983–2020). Source: World Bank Data, 2021
The horizontal axis represents years and ranges from 1985 to 2020 in increments of 5 units. The vertical axis is labeled and ranges from 0 to 80 in increments of 10 units. The graph shows 38 bars. The data for the bars are as follows: 1983: 8.34. 1984: 9.04. 1985: 13.7. 1986: 15.1. 1987: 19.53. 1988: 28.63. 1989: 25.12. 1990: 24.42. 1991: 23.96. 1992: 24.42. 1993: 27. 1994: 37. 1995: 55.65. 1996: 49.36. 1997: 44.4. 1998: 46.56. 1999: 41. 2000: 56. 2001: 79. 2002: 62. 2003: 58. 2004: 53. 2005: 41.2. 2006: 33.7. 2007: 18.36. 2008: 22.5. 2009: 25.1. 2010: 27.2. 2011: 35.1. 2012: 31.4. 2013: 35.6. 2014: 43. 2015: 51.2. 2016: 55. 2017: 57.5. 2018: 58.6. 2019: 59.3. 2020: 63.11. Note: All numerical data values are approximated.Ghana’s public debt history (1983–2020). Source: World Bank Data, 2021
2.3 Macroeconomic performance
Macroeconomic performance is described as an evaluation of how well an economy is doing in reaching critical government policy objectives (Hilton, 2022). Previous researchers have measured macroeconomic performance using different approaches. Over time, GDP growth, unemployment, trade balance and inflation become common indicators to measure macroeconomic performance (Lovell, 1995; Hilton, 2022). These indicators were based on the OECD’s “magic diamond” which was rebased on four macro-parameters such as GDP growth rate, trade balance, inflation rate and unemployment rate (Wang and Le, 2018).
The trends in Ghana’s macroeconomic performance (GDP, inflation, unemployment and trade balance) from 1983 to 2020 are displayed in Figures 2.1, 2.2, 2.3 and 2.4 below. In the case of GDP, the highest (14.04%) was recorded in 2011, while the lowest (−4.56%) was recorded in 1983. Regarding the inflation rate, the highest (122.87%) was in 1983, and the lowest (7.14%) was in 2019. Then, the unemployment rate was very high (10.46%) in 2000 and very low (3.49%) in 1991. Lastly, the trade balance recorded a high index (−0.43) in 1983 and a low (−20.58) in 1997. In terms of averages, the economy of Ghana experienced an average GDP growth rate of 5.11%, an inflation rate of 23.80%, an unemployment rate of 6.20% and a trade balance of −10.86.
The four graphs are shown in two rows and two columns. The first graph is titled “Figure 2.1 Trend in G D P.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from negative 10 to 15 in increments of 5 units. The curve begins at (1983, negative 4.6) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1984, 8.8), (1986, 5.3), (1987, 4.7), (1989, 5.1), (1998, 4.72), (2008, 9.37), (2011, 14.01), (2015, 2.17), and ends at (2020, 0.43). The second graph is titled “Figure 2.2 Trend in Inflation Rate.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from 0 to 140 in increments of 20 units. The curve begins at (1983, 124) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1985, 11), (1996, 60), (1999, 4), (2001, 42.44), (2003, 28.9), and ends at (2020, 8.7). The third graph is titled “Figure 2.3 Trend in Unemployment Rate.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from 0 to 12 in increments of 2 units. The curve begins at (1991, 3.3) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1998, 8.15), (2000, 10.47), (2006, 4.5), (2015, 6.69), (2017, 4.3), and ends at (2020, 4.5). The fourth graph is titled “Figure 2.4 Trend in Trade Balance.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from negative 30 to 0 in increments of 5 units. The curve begins at (1983, 1) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1987, negative 6.5), (1990, negative 9.19), (1993, negative 16.5), (1996, negative 7.9), (2002, negative 12.2), (2005, negative 25.2), (2009, negative 12.9), (2015, negative 8.6), and ends at (2020, negative 3.3). Note: All numerical data values are approximated.Trends in Ghana’s macroeconomic performance. Source: World Bank Data, 2021
The four graphs are shown in two rows and two columns. The first graph is titled “Figure 2.1 Trend in G D P.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from negative 10 to 15 in increments of 5 units. The curve begins at (1983, negative 4.6) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1984, 8.8), (1986, 5.3), (1987, 4.7), (1989, 5.1), (1998, 4.72), (2008, 9.37), (2011, 14.01), (2015, 2.17), and ends at (2020, 0.43). The second graph is titled “Figure 2.2 Trend in Inflation Rate.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from 0 to 140 in increments of 20 units. The curve begins at (1983, 124) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1985, 11), (1996, 60), (1999, 4), (2001, 42.44), (2003, 28.9), and ends at (2020, 8.7). The third graph is titled “Figure 2.3 Trend in Unemployment Rate.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from 0 to 12 in increments of 2 units. The curve begins at (1991, 3.3) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1998, 8.15), (2000, 10.47), (2006, 4.5), (2015, 6.69), (2017, 4.3), and ends at (2020, 4.5). The fourth graph is titled “Figure 2.4 Trend in Trade Balance.” The horizontal axis represents years and ranges from 1983 to 2019 in increments of 2 units. The vertical axis ranges from negative 30 to 0 in increments of 5 units. The curve begins at (1983, 1) and passes through many points forming sharp peaks and troughs, and some points are as follows: (1987, negative 6.5), (1990, negative 9.19), (1993, negative 16.5), (1996, negative 7.9), (2002, negative 12.2), (2005, negative 25.2), (2009, negative 12.9), (2015, negative 8.6), and ends at (2020, negative 3.3). Note: All numerical data values are approximated.Trends in Ghana’s macroeconomic performance. Source: World Bank Data, 2021
2.4 Fiscal vulnerability
Due to the monetary crisis afflicting many emerging market economies for the past decades, the study of fiscal vulnerabilities and their relation to indebtedness became a priority in economic policy discourse (INTOSAI, 2010). International institutions revealed the importance of controlling variables that might threaten debt sustainability. In this regard, the IMF implemented a wide-scope program to determine whether an economy is vulnerable to the debt crises and, if so, to which extent. IMF (2015) emphasizes that every country must be aware of the significance of evaluating and monitoring indebtedness and fiscal performance indicators.
According to INTOSAI (2010), fiscal vulnerability indicators are classified into three, namely: (1) debt indicators (which relate to maturity profiles, reimbursement schedules, sensitivity to interest rates and debt’s composition in foreign currency); (2) indicators on reserves sufficiency (which assess a country’s capability to avoid liquidity crises. It is the relationship between reserves, and short-term debt is a key parameter to assess the vulnerability of countries with considerable but limited access to capital markets); (3) financial soundness indicators (which examine a country’s financial sector strengths and weaknesses relative to the capitalization of financial institutions, assets quality and out-of-balance positions, profitability and liquidity, as well as credit growth’s rhythm and quality). Based on the above general classification of fiscal vulnerability indicators, the following specific indicators emerged in the literature: debt/domestic budgetary revenue, debt service/domestic budgetary revenue, interests/GDP, interests/domestic budgetary revenue, external debt/exports and net international reserves/external debt. In this study, the aforementioned indicators are used to measure fiscal vulnerability since they all provide helpful information about debt management strategies and tend to influence a country’s macroeconomic stability.
2.5 Public debt and macroeconomic performance
Prior studies reveal a link between public debt and macroeconomic performance. Donayre and Taivan (2017) discover that in highly market-driven economies, low economic growth influences public debt, whereas in more socialist states, there is bi-directional influence of low economic growth and debt buildup. Similarly, Arčabić et al. (2018) establish that, contrary to popular belief, the inter-temporal influence of economic growth on public debt is significant while the effect of public debt on economic growth is minimal. Omrane Belguith and Omrane (2017) also found that negative trade balances led to increased public debt in Tunisia. Additionally, Butkus and Seputiene (2018) investigated whether debt threshold levels were influenced by government effectiveness (one component of a country’s institutional quality) and trade balance and discovered that economies with higher public debt threshold levels had stronger institutional quality and a lower trade deficit.
Furthermore, rising debt stock reduces economic performance rates through discouraging private sector investment (Kobayashi and Shirai, 2017). Igberi et al. (2016) ascertain a relationship between rising public debt, inflation, GDP growth rate and unemployment in Nigeria. Omrane Belguith and Omrane (2017) also report that inflation has a statistically significant negative influence on debt and that inflation and investment lowered public debt. On the other hand, Aimola and Odhiambo (2021) assessed the relationship between public debt and inflation in Ghana and discovered a significant positive effect of public debt on inflation in both short and long run. Additionally, public debt and economic performance are positively related. That is, as debt level increases, economic performance increases (Eberhardt and Presbitero, 2015; Ewaida, 2017; Huang et al., 2018; Saungweme and Odhiambo, 2019).
Last but not least, there is neutral relationship between public debt and economic performance (Panizza and Presbitero, 2014; Hilton, 2021). Panizza and Presbitero (2014) submit that there is no relationship between public debt and GDP, inflation and trade balance. Hilton (2021) also indicates that there is no causal relationship between public debt and GDP and inflation in short-run. Given the unclear or direct association between public debt and macroeconomic performance, we set the following null and alternative hypotheses.
No relationship exists between public debt and macroeconomic performance
Relationship exists between public debt and macroeconomic performance
2.6 Public debt, macroeconomic performance and fiscal vulnerability
Literature suggests a link between public debt, macroeconomic performance and fiscal vulnerability. Gnangnon (2019) investigated the impact of emerging nations’ structural economic vulnerability on their public debt. The findings imply that structural vulnerability and overall public debt in emerging nations have a “U-shaped” relationship. The structural vulnerability appears to be a key driver of overall public debt accumulation, particularly in low-income countries (Gnangnon, 2019).
Further, Dufrénot et al. (2016) examined macro-financial imbalances that exposed European nations to fiscal stress before the European debt crisis. Rather than interpreting fiscal stress regarding fiscal sustainability, they focused on short-term fiscal fragility as indicated by debt refinancing circumstances in sovereign bond markets. They discovered that market-based indicators representing sovereign debt risk perceptions were impacted by indicators described in the European Macroeconomic Imbalance Procedure (MIP) and financial vulnerability characteristics. Holders of government debts consider macroeconomic imbalances and elements such as banking distress, corporate bond risk, liquidity hazards in the interbank market and stock price volatility when pricing the risk of sovereign bonds (Dufrénot et al., 2016).
Marti and Pérez (2016) contend that despite pre-crisis budgetary surpluses and low levels of public debt, Spain’s public finances have been under severe strain during the crisis. The crisis’s impact and an early period of counter-cyclical action worsened the existing (structural) fiscal vulnerabilities. A variety of bold policy moves were implemented to address the budgetary imbalances impacting taxation, public expenditure, national fiscal laws and the organization of the public sector (Mart and Pérez, 2016).
In Ghana, Matuka and Asafo (2018) establish a relationship between debt, total debt service to export ratio, GDP, trade balance, external debt to GDP and gross capital formation to GDP. Similarly, after assessing 72 developing countries using panel Granger-causality, Jalles (2011) reported no relationship between GDP and debt service to exports ratio. Therefore, we hypothesize that:
Fiscal vulnerability relates to macroeconomic performance.
There is an interaction effect of public debt and fiscal vulnerability on macroeconomic performance.
2.7 Research gap
While existing studies have investigated the relationship between public debt and macroeconomic performance (e.g. Huang et al., 2018; Saungweme and Odhiambo, 2019; Hilton, 2021), they largely focused on the direct impact of public debt on macroeconomic performance, without adequately considering the role of fiscal vulnerability in moderating this relationship. Meanwhile, fiscal vulnerability may potentially amplify the negative effects of public debt on macroeconomic performance. The existing studies have largely assumed a linear relationship between public debt and macroeconomic performance, without considering potential directional change in the effects that may arise from fiscal vulnerability. It follows that there is a need for more empirical research that examines the interaction effect of fiscal vulnerability on the relationship between public debt and macroeconomic performance. Furthermore, recent scholars have highlighted the importance of adopting both frequentist and non-frequentist econometric models to generate more robust, accurate, reliable and replicable findings (Briggs, 2023), yet available studies on the phenomenon utilized only the frequentist approach (mainly OLS and ARDL models). The Bayesian approach is efficient in checking the robustness of the frequentist findings (Thach, 2023, 2024). Thus, we seek to address these research gaps by investigating the interaction effect of fiscal vulnerability on the relationship between public debt and macroeconomic performance, using historical data of a developing economy (where is unavailable study) and econometric methods: least squares – OLS (frequentist approach) and Bayesian linear regression model in a hierarchical regression modeling.
2.8 Conceptual framework
Based on the literature above, we propose that the potential relationship between public debt and macroeconomic performance can be enhanced or impeded by the presence of fiscal vulnerability. Figure 3 illustrates the expected interaction effect of fiscal vulnerability on the potential association between public debt and macroeconomic performance. This will be so if the interaction effect is significant. Prior literature argues that poor debt management strategies, which are fiscal vulnerability issues, can lead to economic slump (Cochrane, 2011; Castro et al., 2015). Therefore, the assessment of the relationship between public debt and macroeconomic performance ought to include the moderating effect of fiscal vulnerability.
The flow begins with a box labeled “Public Debt” on the left. From “Public Debt,” a right-pointing arrow labeled “H10, H1a” points to a box labeled “Macroeconomic Performance.” At the top center, a box labeled “Fiscal Vulnerability” is present. From “Fiscal Vulnerability,” a downward arrow labeled “H3” points to the “H10, H1a” arrow. From “Fiscal Vulnerability,” another arrow labeled “H2” points to “Macroeconomic Performance.”Conceptual framework. Source: Authors’ construct (2024)
The flow begins with a box labeled “Public Debt” on the left. From “Public Debt,” a right-pointing arrow labeled “H10, H1a” points to a box labeled “Macroeconomic Performance.” At the top center, a box labeled “Fiscal Vulnerability” is present. From “Fiscal Vulnerability,” a downward arrow labeled “H3” points to the “H10, H1a” arrow. From “Fiscal Vulnerability,” another arrow labeled “H2” points to “Macroeconomic Performance.”Conceptual framework. Source: Authors’ construct (2024)
3. Methodology
3.1 Model estimation
We employed hierarchical regression model for the interaction effect of public debt and fiscal vulnerability on macroeconomic performance. The hierarchical approach allowed for the control of fiscal vulnerability, providing insight into its moderating effects on the relationship between public debt and macroeconomic performance. The hierarchical modeling allows for a step by step approach to the analysis, usually a three-model procedure. With this technique, the main effect of the independent variable can be examined, followed by potential controlling effect of the moderators, and finally, the interaction effect of the independent and moderating variables. This approach allows the researcher(s) to create interaction terms for the independent variable(s) and moderator(s). This model has been recommended by Aiken et al. (1991) and Cohen et al. (2003), and has been widely applied in social science studies (e.g. Hilton et al., 2024; Martins et al., 2024; Hilton et al., 2021, etc.). It is proven to be more efficient to assess interaction effect. We employed the ordinary least squares (OLS) method for our hierarchical regression model due to its suitability for assessing interaction effects in a frequentist framework. OLS assumes a linear relationship between the independent variables and the outcome variable, which is reasonable given the theoretical expectations in our study. It can effectively capture interaction effects between variables, enabling us to examine how fiscal vulnerability moderates the association between public debt and macroeconomic performance. It is a widely used and robust method, providing reliable estimates when assumptions are met. Compared to an alternative method like generalized linear models (GLMs), though GLMs offer flexibility, OLS is more suitable for continuous outcome variables and provides more straightforward interpretations. Thus, our use of OLS allows for a more traditional frequentist interpretation and comparison to the Bayesian method.
We further adopted the Bayesian approach, which does not rely on p-values to support the frequentist approach (i.e. OLS). We employed a Bayesian linear regression model as a robustness check, utilizing a normal prior distribution for the coefficients and an inverse-gamma prior for the variance. This approach allows for a comparison with the frequentist results and provides a more nuanced understanding of the relationships. The Markov Chain Monte Carlo (MCMC) simulation was employed to explore the posterior distribution of parameters (Gelman and Rubin, 1992; Kalia, 2024).
3.2 Model specification
The hierarchical regression models were specified as follows. Macroeconomic performance indicators (GDP growth rate, unemployment rate, inflation rate and trade balance) were specified as dependent variables. Public debt (total debt to GDP ratio) was specified as independent variable. The fiscal vulnerability indicators (debt/domestic budgetary revenue, debt service/domestic budgetary revenue, interests/GDP, interests/domestic budgetary revenue, external debt/exports and net international reserves/external debt) were specified as moderators. These variables were defined in Appendix 1. Applying Aiken et al.’s (1991) procedure, we created interaction terms for public debt and fiscal vulnerability indicators. The variables were centered around a mean of 0 to reduce the correlation between the interaction terms and the variables comprising the interaction to prevent multicollinearity. For simplicity, four (4) interaction effect equations were estimated further in the text.
where:
GDP is the annual GDP growth rate; INF is the inflation rate; UNE is the unemployment rate; TB is the trade balance; PD1 is the total public debt; PD_DDR2 represents the interaction term of public debt and debt/domestic revenue; PD_DSDR3 represents the interaction term of public debt and debt service/domestic revenue; PD_IGDP4 represents the interaction term of public debt and interests/GDP; PD_IDR5 represents the interaction term of public debt and interests/domestic revenue; PD_EDE6 represents the interaction term of public debt and external debt/exports; PD_NIR7 represents the interaction term of public debt and net international reserves/external debt; α1 – α4 are the respective constants and β1 – β7 are respective regression coefficients.
3.3 Data and source
We employed time-series data annually from 1983 to 2020. The period and frequency were used because there was no available data on the fiscal vulnerability indicators prior to 1983. Thus, to have consistent and stable data for the analysis, we chose the 1983–2020 periods where data on all the study variables exist.
Data on annual GDP growth rate as a measure of GDP, trade balance as a percent of GDP, inflation rate, unemployment rate and fiscal vulnerability indicators, were obtained from the World Bank Development Indicator database (World Bank, 2021) and BoG (2021). Data on total public debt as a percent of GDP were obtained from the IMF fiscal Affairs Department Database and WEO (IMF, 2021).
3.4 Unit root test
We ran a unit root test to confirm the stationarity of the series employed. Literature shows that time-series data exhibit non-stationary tendencies and therefore spurious correlations may show up among variables which are non-stationary over time (Granger and Newbold, 1974; Phillips, 1986). Thus, we adopted Kwiatkowski-Phillips-Schmidt-Shin (KPSS) to test the standard unit root in the data. KPSS test is appropriate for small sample sizes (less than 50 observations), whereas Augmented Dickey-Fuller (ADF) test is more suitable for large sample sizes (more than 50 observations). KPSS tests the null hypothesis that the series is stationary. We reject the null if the test statistics is higher than the critical value or the p-value is less than the significance level of 5%, indicating non-stationarity. Otherwise, we accept the null hypothesis, suggesting stationarity. The KPSS unit root test indicated that the series is stationary (Table 1).
KPSS unit root test
| Intercept and no trend | Intercept and trend | |||
|---|---|---|---|---|
| Variables | Levels | First difference | Levels | First difference |
| GDP | 0.122 [0.46] | 0.246 [0.46] | 0.116 [0.15] | 0.159 [0.15] |
| PD | 0.443 [0.46] | 0.196 [0.46] | 0.138 [0.15] | 0.110 [0.15] |
| UNE | 0.201 [0.46] | 0.277 [0.46] | 0.105 [0.15] | 0.124 [0.15] |
| INF | 0.852 [0.46] | 0.322 [0.46] | 0.353 [0.15] | 0.194 [0.15] |
| TB | 0.199 [0.46] | 0.383 [0.46] | 0.192 [0.15] | 0.191 [0.15] |
| DDR | 0.271 [0.46] | 0.355 [0.46] | 0.175 [0.15] | 0.143 [0.15] |
| DSDR | 0.432 [0.46] | 0.404 [0.46] | 0.197 [0.15] | 0.146 [0.15] |
| IGDP | 0.727 [0.46] | 0.268 [0.46] | 0.149 [0.15] | 0.072 [0.15] |
| IDR | 0.405 [0.46] | 0.349 [0.46] | 0.166 [0.15] | 0.139 [0.15] |
| EDE | 0.619 [0.46] | 0.133 [0.46] | 0.097 [0.15] | 0.106 [0.15] |
| NIR | 0.362 [0.46] | 0.094 [0.46] | 0.108 [0.15] | 0.090 [0.15] |
| PD_DDR | 0.234 [0.46] | 0.320 [0.46] | 0.113 [0.15] | 0.141 [0.15] |
| PD_DSDR | 0.147 [0.46] | 0.258 [0.46] | 0.116 [0.15] | 0.137 [0.15] |
| PD_IGDP | 0.303 [0.46] | 0.565 [0.46] | 0.193 [0.15] | 0.122 [0.15] |
| PD_IDR | 0.266 [0.46] | 0.329 [0.46] | 0.134 [0.15] | 0.139 [0.15] |
| PD_EDE | 0.313 [0.46] | 0.470 [0.46] | 0.214 [0.15] | 0.113 [0.15] |
| PD_NIR | 0.153 [0.46] | 0.178 [0.46] | 0.152 [0.15] | 0.115 [0.15] |
Note(s): Critical value is 5% or p-value > 0.05 as shown in the parentheses
4. Results and discussion
4.1 Stationarity, normality, serial correlation, heteroskedasticity and descriptive statistics
The KPSS test was run to check the unit roots or stationarity. Table 1 presents the unit root test with intercept and no trend as well as intercept and trend. The results illustrate that the series are stationary as the test statistics are low and the critical values or p-values are more than 5%. Furthermore, to ensure that the series are not suffering from serial correlation and heteroskedasticity, Breusch-Godfrey serial correlation and heteroskedasticity tests were run. The models are not serially correlated and there is no heteroskedasticity as the [Prob.] for the respective variables are not less than 5% significance level (Table 2). Additionally, we checked the normality of the series using Jarque-Bera. The p-value of Jarque-Bera for each model is above 5% significance level, affirming normality of the series (Table 2). It means that the findings are valid and reliable to provide empirical support to existing literature. The descriptive statistics for the variables are shown in Table 3.
Serial correlation, heteroskedasticity and normality test
| Dependent variable | Serial correlation LM F-statistic [Prob.] | Heteroskedasticity F-statistic [Prob.] | Normality Jarque-Bera [Prob.] | |
|---|---|---|---|---|
| GDP | 1.978756 [0.1940] | 0.662388 [0.7603] | 1.506443 [0.4708] | |
| INF | 1.957274 [0.1969] | 1.179733 [0.4034] | 0.705355 [0.7028] | |
| UNE | 0.030931 [0.9696] | 0.699819 [0.7294] | 0.799398 [0.6705] | |
| TB | 3.039603 [0.0980] | 0.680036 [0.7466] | 0.824476 [0.6621] |
Note(s): Null hypotheses are rejected at p-value > 0.05 as shown in the parentheses
Descriptive statistics
| Variables | Min | Max | Mean | Std. deviation |
|---|---|---|---|---|
| PD | 2.126 | 4.371 | 3.522 | 0.539 |
| GDP | −0.880 | 2.642 | 1.569 | 0.551 |
| INF | 1.582 | 4.811 | 2.923 | 0.675 |
| UNE | 1.249 | 2.347 | 1.787 | 0.273 |
| TB | −25.273 | −0.433 | −10.866 | 6.407 |
| EDE | −14.667 | −12.226 | −13.225 | 0.775 |
| DDR | −13.158 | 1.688 | −11.960 | 2.941 |
| DSDR | −22.849 | −8.442 | −19.390 | 3.776 |
| IGDP | −25.421 | −15.724 | −19.956 | 2.880 |
| IDR | −16.459 | −0.784 | −14.789 | 3.014 |
| NIR | 1.521 | 4.121 | 2.954 | 0.720 |
4.2 Main results
The frequentist analysis involved three-model approach. Model 1 represents the main effect (public debt and macroeconomic performance). Model 2 represents the controlling effect of the moderators (fiscal vulnerability indicators). Model 3 represents the interaction effect (public debt and interaction terms of public debt and fiscal vulnerability indicators). The coefficients, t-statistics, p-value, R-square, adjusted R-square and F-statistics as reported in the various tables, show that both public debt and fiscal vulnerability indicators explain significant changes in the variance of macroeconomic performance indicators. For easy comparison, the Bayesian analysis also involved three-step approach, where step 1 denotes the main effect, step 2 denotes the controlling effect of the moderators and step 3 denotes the interaction effect. We specified normal priors for the structural parameters or regression coefficients (β ∼ N(0, 10,000)) and inverse-gamma prior for variance (shape = 0.01, scale = 0.01), a non-informative prior setting. We assessed MCMC convergence using the Gelman-Rubin statistics (R-hat), which yielded values less than 1.1 (Table 7), indicating satisfactory convergence (Gelman and Rubin, 1992; Brooks and Gelman, 1998). Additionally, we examined MCMC simulation diagnostics, which generated acceptance rates between 35.1% and 36.7% across the three model specifications (Table 7), suggesting reasonable mixing. The average efficiency decreased from 13.2% in the main effect model to 2% and 2.1% in the controlling effect and interaction effect models, respectively, likely due to increased model complexity (Table 7). These diagnostics provide evidence of adequate MCMC performance and convergence (Roberts and Rosenthal, 2001).
Table 4 presents the main effects for the frequentist analysis. There is a significant relationship between public debt and macroeconomic performance indicators. Hence, we reject the null hypothesis (H10) of no relationship and accept the alternative hypothesis (H1a) that relationship exists. Specifically, public debt has negative relationships with GDP, inflation and trade balance. However, public debt positively correlates with unemployment. These findings are consistent with Igberi et al.’s (2016) study which submitted that a relationship exists between public debt, inflation, GDP and unemployment in Nigeria. However, these findings contradict Panizza and Presbitero’s (2014) finding that there is no relationship between public debt, GDP, inflation and trade balance.
Main effects (public debt and macroeconomic performance)
| Estimated | Estimators | Coefficient | t-stat | Prob | R2 | ΔR2 | F-stat |
|---|---|---|---|---|---|---|---|
| GDP | PD | −0.342 | −1.90* | 0.065 | 0.094 | 0.068 | 3.625 |
| INF | PD | −0.336 | −1.67* | 0.102 | 0.072 | 0.047 | 2.805 |
| UNE | PD | 0.323 | 2.59** | 0.014 | 0.193 | 0.165 | 6.733 |
| TB | PD | −3.224 | −1.69* | 0.098 | 0.074 | 0.048 | 2.868 |
Note(s): ***p < 0.01, **p < 0.05 and *p < 0.10
A negative relationship between public debt and GDP suggests that as public debt increases, GDP decreases, ceteris paribus. High public debt may lead to a debt overhang, where the burden of debt repayment limits the government’s ability to invest in growth-enhancing activities, thereby reducing GDP (Krugman, 1988). Similarly, high public debt may crowd out private investment, leading to reduced government spending, as a large portion of the budget is dedicated to debt servicing, thereby reducing GDP (Broner et al., 2014). High public debt may also lead to lower economic growth, as the government may struggle to finance investments in public goods and services, infrastructure and human capital. Ultimately, high public debt may lead to a debt crisis, where government struggles to pay its debts, leading to a sharp contraction in GDP. This is evidence in Ghana’s struggle to repay both domestic and external and had to resort to IMF bailout, domestic debt exchange program and plead for external debt forgiveness between 2022 and 2023 fiscal years. Our finding affirms previous studies (e.g. Reinhart et al., 2012; Anning et al., 2016; Kobayashi and Shirai, 2017) that revealed an inverse relationship between public debt and GDP, but opposes other researchers (e.g. Ewaida, 2017; Huang et al., 2018) who indicated that GDP level rises as public debt level rises.
Furthermore, a negative relationship between public debt and inflation means that as public rises, inflation falls. High public debt may lead to increased efforts to reduce debt, which can result in reduced government spending, lower aggregate demand and subsequently lower inflation. Again, as high public debt crowds out private investment, it reduces economic activity and lower inflation. Likewise, high public debt may decrease investor confidence, causing a decrease in spending and investment, and subsequently lower inflation. This present finding does not support Aimola and Odhiambo’s (2021) study, which reported positive relationship between public debt and inflation in Ghana in short and long runs. The difference may be due to estimation methods.
Moreover, a negative relationship between public debt and trade balance suggests that as pubic debt increases, the trade balance worsens, ceteris paribus. High public debt is associated with larger budget deficits, which can lead to larger trade deficits as the government relies more heavily on foreign capital to finance its spending. Again, high public debt may crowd out private investment, reducing competitiveness and exports, and subsequently worsening trade balance. High public debt may also lead to a depreciation of the currency, making imports more expensive and exports less competitive, thus worsening the trade balance. It is worth noting that high public debt may reduce national savings, leading to a decrease in investment and exports, and a worsening trade balance. This present finding is consistent with Omrane Belguith and Omrane’s (2017) study that established an inverse relationship between public debt and trade balance in Tunisia.
Meanwhile, a positive relationship between public debt and unemployment suggests that as public debt increases, the unemployment rate also increases, ceteris paribus. This indicates that government spending and taxation might be ineffective in stimulating economic growth and reducing unemployment. High public debt may crowd out private investment, leading to reduced job creation and higher unemployment. Additionally, high public debt may lead to higher interest rates, making borrowing more expensive for businesses and individuals, thereby reducing investment and increasing unemployment. This finding is also reported by Igberi et al. (2016).
We found consistency in both frequentist and Bayesian results. Disparity only occurs when the statistical significance level of p < 0.05 is extended to p < 0.10 (i.e. in the case of GDP, inflation and trade balance) under the frequentist approach. In the Bayesian analysis, the significance of an estimated coefficient is determined by examining its credible intervals. A coefficient is considered significant if its credible interval does not cross zero, indicating that the posterior distribution of the coefficient lies entirely on one side of zero (Gelman et al., 2013; Kruschke, 2014; McElreath, 2020). This approach allows us to quantify the uncertainty surrounding our estimates and provides a more nuanced understanding of the relationships between variables. This outcome proves the robustness of the Bayesian approach, making its adoption in our study novel.
Table 5 presents the controlling effect of the fiscal vulnerability indicators (frequentist results). It can be observed that fiscal vulnerabilities as control variables do not have significant relationship with GDP and inflation. However, they have significant relationship with unemployment rate and trade balance. It means that whereas these fiscal vulnerabilities do not contribute to explaining criterion variance regarding GDP and inflation, they do in the case of unemployment and trade balance. Precisely, debt to domestic revenue ratio, debt service to domestic revenue ratio and interests to GDP ratio have negative significant relationship with unemployment, while interests to domestic revenue ratio and external debt to exports ratio positively correlate with unemployment. Meanwhile, the relationship between net international reserves to external debt ratio and unemployment is insignificant. Debt to domestic revenue ratio, debt service to domestic revenue ratio and interests to GDP ratio have negative significant association with trade balance, while interests to domestic revenue ratio positively correlates with trade balance. In the case of external debt to exports, net international reserves to external debt and trade balance, the correlation is insignificant. In contrast, the Bayesian results in Table 7 indicate that debt to domestic revenue ratio and net international reserves to external debt are significantly related to GDP. None of the control variables show significant relationships with inflation. For unemployment, most control variables (except debt to domestic revenue ratio and interests to domestic revenue ratio) are significant. Lastly, only the external debt to exports ratio shows a significant relationship with trade balance.
Controlling effects of fiscal vulnerability
| Estimated | Estimators | Coefficient | t-stat | Prob | R2 | ΔR2 | F-stat |
|---|---|---|---|---|---|---|---|
| GDP | PD | 0.382 | 0.77 | 0.450 | 0.743 | 0.631 | 6.623 |
| GDP | DDR | 0.694 | 0.80 | 0.432 | 0.743 | 0.631 | 6.623 |
| GDP | DSDR | −0.178 | −0.73 | 0.473 | 0.743 | 0.631 | 6.623 |
| GDP | IGDP | −0.217 | −0.54 | 0.594 | 0.743 | 0.631 | 6.623 |
| GDP | IDR | −0.681 | −0.71 | 0.487 | 0.743 | 0.631 | 6.623 |
| GDP | EDE | −0.421 | −0.86 | 0.399 | 0.743 | 0.631 | 6.623 |
| GDP | NIR | 0.061 | 0.15 | 0.880 | 0.743 | 0.631 | 6.623 |
| INF | PD | 0.478 | 0.90 | 0.380 | 0.397 | 0.133 | 1.504 |
| INF | DDR | −1.146 | −1.24 | 0.233 | 0.397 | 0.133 | 1.504 |
| INF | DSDR | −0.195 | −0.75 | 0.464 | 0.397 | 0.133 | 1.504 |
| INF | IGDP | −0.523 | −1.22 | 0.239 | 0.397 | 0.133 | 1.504 |
| INF | IDR | 1.332 | 1.29 | 0.213 | 0.397 | 0.133 | 1.504 |
| INF | EDE | 0.624 | 1.19 | 0.250 | 0.397 | 0.133 | 1.504 |
| INF | NIR | 0.121 | 0.29 | 0.779 | 0.397 | 0.133 | 1.504 |
| UNE | PD | 0.489 | 4.22*** | 0.000 | 0.887 | 0.835 | 16.906 |
| UNE | DDR | −0.634 | −3.11*** | 0.007 | 0.887 | 0.835 | 16.906 |
| UNE | DSDR | −0.515 | −8.57*** | 0.000 | 0.887 | 0.835 | 16.906 |
| UNE | IGDP | −0.758 | −7.91** | 0.000 | 0.887 | 0.835 | 16.906 |
| UNE | IDR | 1.147 | 5.07*** | 0.000 | 0.887 | 0.835 | 16.906 |
| UNE | EDE | 0.428 | 3.62*** | 0.002 | 0.887 | 0.835 | 16.906 |
| UNE | NIR | −0.154 | −1.59 | 0.132 | 0.887 | 0.835 | 16.906 |
| TB | PD | −12.378 | −2.02* | 0.060 | 0.564 | 0.374 | 2.966 |
| TB | DDR | 12.773 | 1.19 | 0.248 | 0.564 | 0.374 | 2.966 |
| TB | DSDR | 9.645 | 3.21*** | 0.005 | 0.564 | 0.374 | 2.966 |
| TB | IGDP | 17.441 | 3.53*** | 0.002 | 0.564 | 0.374 | 2.966 |
| TB | IDR | −22.525 | −1.90* | 0.075 | 0.564 | 0.374 | 2.966 |
| TB | EDE | −8.922 | −1.47 | 0.158 | 0.564 | 0.374 | 2.966 |
| TB | NIR | −5.060 | −1.03 | 0.318 | 0.564 | 0.374 | 2.966 |
Note(s): ***p < 0.01, **p < 0.05 and *p < 0.10
Table 6 shows the interaction effect results for the frequentist approach. For GDP, the interaction term of public debt and debt service to domestic revenue ratio is positively significant. It means that public debt level and government debt-servicing capacity (i.e. the ability to pay both interest and capital with domestic revenue) may lead to high GDP growth. It also suggests that debt restructuring plays a crucial role in shaping the impact of public debt on GDP. The current finding refutes Saungweme and Odhiambo’s (2019) results of no relationship between debt servicing and GDP. The other interaction terms have no significant association with GDP. In contrast, the Bayesian results show that all interaction terms except public debt and debt service to domestic revenue ratio are significant (Table 7). It means that changes in these interaction terms may affect GDP. For instance, the possibility of government facing unproductive expenditures because of interest payments is consequential to GDP growth. Similarly, the financial cost as a proportion of tax revenue and public debt stock can significantly affect GDP growth, because of an increase in unproductive expenditures. Likewise, external debt to exports ratio inflates the inverse effect of public debt on GDP, ceteris paribus. This finding rebuts Jalles (2011) that there is no relationship between GDP and external debt to exports ratio, but supports Matuka and Asafo (2018) who revealed a relationship between external debt to exports ratio and GDP.
Interaction effects of fiscal vulnerability
| Estimated | Estimators | Coefficient | t-stat | Prob | R2 | ΔR2 | F-stat |
|---|---|---|---|---|---|---|---|
| GDP | PD | −2.559 | −1.63 | 0.121 | 0.745 | 0.633 | 6.684 |
| GDP | PD_DDR | −0.872 | −1.31 | 0.761 | 0.745 0 | 0.633 | 6.684 |
| GDP | PD_DSDR | 1.483 | 2.34* | 0.032 | 0.745 | 0.633 | 6.684 |
| GDP | PD_IGDP | 1.139 | 1.29 | 0.215 | 0.745 | 0.633 | 6.684 |
| GDP | PD_IDR | −0.850 | −0.29 | 0.772 | 0.745 | 0.633 | 6.684 |
| GDP | PD_EDE | −0.606 | −1.15 | 0.267 | 0.745 | 0.633 | 6.684 |
| GDP | PD_NIR | 0.285 | 0.41 | 0.687 | 0.745 | 0.633 | 6.684 |
| INF | PD | 3.226 | 1.73* | 0.100 | 0.261 | 0.061 | 0.811 |
| INF | PD_DDR | −1.169 | −0.35 | 0.732 | 0.261 | 0.061 | 0.811 |
| INF | PD_DSDR | −1.475 | −1.96* | 0.068 | 0.261 | 0.061 | 0.811 |
| INF | PD_IGDP | −1.788 | −1.70* | 0.108 | 0.261 | 0.061 | 0.811 |
| INF | PD_IDR | 2.623 | 0.76 | 0.457 | 0.261 | 0.061 | 0.811 |
| INF | PD_EDE | 1.624 | 0.97 | 0.344 | 0.261 | 0.061 | 0.811 |
| INF | PD_NIR | −0.414 | −0.49 | 0.624 | 0.261 | 0.061 | 0.811 |
| UNE | PD | 1.600 | 3.67*** | 0.002 | 0.848 | 0.777 | 11.976 |
| UNE | PD_DDR | 0.413 | 0.51 | 0.618 | 0.848 | 0.777 | 11.976 |
| UNE | PD_DSDR | −0.622 | −3.17*** | 0.006 | 0.848 | 0.777 | 11.976 |
| UNE | PD_IGDP | −0.820 | −3.32*** | 0.004 | 0.848 | 0.777 | 11.976 |
| UNE | PD_IDR | 0.209 | 0.26 | 0.799 | 0.848 | 0.777 | 11.976 |
| UNE | PD_EDE | 0.318 | 0.82 | 0.423 | 0.848 | 0.777 | 11.976 |
| UNE | PD_NIR | −0.298 | −1.53 | 0.146 | 0.848 | 0.777 | 11.976 |
| TB | PD | −13.680 | −0.84 | 0.415 | 0.691 | 0.556 | 5.126 |
| TB | PD_DDR | −16.951 | −0.57 | 0.573 | 0.691 | 0.556 | 5.126 |
| TB | PD_DSDR | 7.316 | 1.10 | 0.285 | 0.691 | 0.556 | 5.126 |
| TB | PD_IGDP | 12.271 | 1.32 | 0.202 | 0.691 | 0.556 | 5.126 |
| TB | PD_IDR | 9.370 | 0.31 | 0.760 | 0.691 | 0.556 | 5.126 |
| TB | PD_EDE | 0.117 | 0.01 | 0.993 | 0.691 | 0.556 | 5.126 |
| TB | PD_NIR | 3.944 | 0.54 | 0.595 | 0.691 | 0.556 | 5.126 |
Note(s): ***p < 0.01, **p < 0.05 and *p < 0.10
Bayesian regression results
| Variables | GDP PM [95% C.I] | INF PM [95% C.I] | UNE PM [95% C.I] | TB PM [95% C.I] |
|---|---|---|---|---|
| 1. Main effect | ||||
| PD | −0.15 [−0.48, 0.18] | −0.07 [−0.39, 0.26] | 0.37 [0.06, 0.68] | 0.03 [−0.29, 0.36] |
| 2. Controlling effect | ||||
| PD | 0.21 [−0.44, 0.92] | 0.26 [−0.42, 0.93] | 0.13 [−0.12, 0.37] | 0.36 [−0.11, 0.81] |
| EDE | 0.45 [−0.55, 1.43] | −0.11 [−1.14, 0.91] | 0.52 [0.11, 0.91] | −1.06 [−1.76, −0.35] |
| DDR | −1.58 [−2.74, −0.30] | 0.18 [−0.92, 1.34] | −0.35 [−1.04, 0.29] | 0.84 [−0.21, 1.93] |
| DSDR | −0.06 [−1.37, 1.19] | −0.31 [−1.57, 0.91] | 0.53 [0.08, 1.04] | 0.30 [−0.48, 1.19] |
| IGDP | 0.24 [−0.65, 1.16] | 0.76 [−0.12, 1.64] | 0.50 [0.26, 0.74] | 0.02 [−0.48, 0.53] |
| IDR | 0.56 [−0.82, 1.99] | −0.30 [−1.66, 1.03] | 0.03 [−0.52, 0.61] | 0.40 [−0.77, 1.26] |
| NIR | 1.14 [0.06, 2.11] | 0.27 [−0.80, 1.27] | −0.39 [−0.73, −0.03] | 0.34 [−0.33, 1.06] |
| 3. Interaction effect | ||||
| PD | 0.39 [−0.02, 0.88] | 0.35 [0.05, 0.64] | 0.04 [−0.26, 0.37] | −0.24 [−0.80, 0.22] |
| PD × EDE | −20.15 [−25.6, −15.4] | −26.99 [−49.3, −3.99] | −7.16 [−20.04, 5.53] | 1.20 [−21.2, 20.9] |
| PD × DDR | −31.23 [−41.1, −21.3] | 44.71 [36.26, 52.92] | 25.48 [18.54, 32.30] | −37.87[−53.9, −22.5] |
| PD × DSDR | 5.56 [−1.17, 12.0] | −14.39 [−21.6, −6.62] | 8.60 [3.25, 13.84] | 1.59 [−6.65, 10.34] |
| PD × IGDP | −13.69 [−18.3, −8.81] | −7.49 [−16.1, 1.06] | 6.75 [3.06, 10.66] | 3.63 [−1.25, 8.97] |
| PD × IDR | 28.78 [17.92, 40.40] | −32.46 [−36.2, −28.5] | −35.23 [−39.7, −31.2] | 33.60[23.35, 44.89] |
| PD × NIR | 15.95 [5.01, 26.96] | −39.04 [−46.9, −31.2] | 2.60 [−5.20, 9.36] | −2.63 [−14.76, 8.39] |
| MCMC Simulation Diagnostics | Value (step 1) | Value (step 2) | Value (step 3) |
|---|---|---|---|
| Acceptance rate | 0.351 | 0.355 | 0.367 |
| Min efficiency | 0.123 | 0.011 | 0.011 |
| Max efficiency | 0.151 | 0.038 | 0.031 |
| Average efficiency | 0.132 | 0.020 | 0.021 |
| MCMC Convergence Diagnostics | |||
| Gelman–Rubin Rc | 1.001 | 1.044 | 1.024 |
Note(s): PM = Posterior Mean and C.I. = Credible Interval
Regarding inflation, while the Bayesian results show that all interaction terms except interests to GDP ratio are significant (Table 7), the frequentist results depict that only the interaction terms of public debt and debt service to domestic revenue ratio, and public debt and interests to GDP ratio have negative significant relationship with inflation (Table 6). This suggests that as interests to GDP ratio and public debt level rises, inflation rate is expected to reduce. Similarly, government’s ability to pay both interest and capital with domestic revenue) may lead to lower inflation. This is because high public debt may lead to increased efforts to reduce debt, which can result in reduced government spending, lower aggregate demand and subsequently lower inflation. Furthermore, a negatively significant interaction effect of public debt and external debt to exports ratio on inflation suggests that the debt burden level over exports or the capability of acquiring currencies might affect inflation rate. Consequently, the interaction effects results enhance the established impact of public debt on inflation rate in the short-run (Hilton, 2021), and long-run (Aimola and Odhiambo, 2021).
The Bayesian results reveal that only the interaction terms of public debt and external debt to exports ratio, and public debt and net international reserves to external debt ratio are not significantly related to unemployment rate (Table 7), whereas the frequentist results show that only the interaction terms of public debt and debt service to domestic revenue ratio, and public debt and interests to GDP ratio have negative significant association with unemployment rate (Table 6). This illustrates that when public debt interacts with interests to GDP ratio and debt service to domestic revenue ratio, they produce an inverse relationship with unemployment, even though public debt as independent variable in the model has positive significant effect on unemployment. This implies that indeed public debt and unemployment rate are positively related as recorded in model 1 (main effect), and also reported in extant literature (Igberi et al., 2016). On the contrary, the interaction of public debt and external debt to exports ratio has insignificant effect on unemployment. It implies that as external debt burden level rises together with public debt stock, unemployment rate will be constant or rise insignificantly, ceteris paribus.
In the case of trade balance, the interaction terms are not significant for the frequentist approach. They have no significant effect on trade balance. It means that there is no interaction effect of the fiscal vulnerability on macroeconomic performance. For instance, government’s ability to maintain favorable net international reserves to external debt ratio may not affect the relationship between public debt and trade balance. Similarly, government’s control of the number of times that external debt exceeds reserves may insignificantly affect public debt and trade balance. However, two interaction terms: public debt and debt to domestic revenue ratio (negative effect) and public debt and interests to domestic revenue ratio (positive effect) are significant for the Bayesian approach.
5. Conclusion
We employed six fiscal vulnerability indicators as moderators in the association between public debt and macroeconomic performance. Public debt has a relationship with macroeconomic performance. Precisely, public debt relates negatively with GDP, inflation and trade balance, but it relates positively with unemployment. The frequentist approach highlights specific interaction terms that are significantly related to macroeconomic performance. Notably, the interaction between public debt and debt service/domestic revenue is significantly related to GDP, inflation and unemployment. This suggests that debt restructuring plays a crucial role in shaping the impact of public debt on these macroeconomic outcomes. Additionally, the interaction between public debt and interests/GDP is significantly related to inflation and unemployment, implying that government spending influences the effects of public debt on these outcomes. In contrast, the Bayesian approach yields a more nuanced understanding of the interaction effects. Most interaction terms are significant for GDP, inflation and unemployment, highlighting the complex relationships between public debt and these macroeconomic outcomes. Furthermore, the Bayesian approach reveals that the interactions between public debt and debt/domestic revenue, and public debt and interests/domestic revenue are significant for trade balance, suggesting that debt dynamics play a crucial role in shaping the impact of public debt on trade balance.
The comparison of frequentist and Bayesian results highlights the importance of considering multiple statistical approaches. The divergent results between the two approaches emphasize the need for robustness checks and careful consideration of methodological choices. Moreover, the interaction effects demonstrate that the impact of public debt on macroeconomic performance is context-dependent, varying with different levels of control variables. This has important implications for policy decisions, as it suggests that policymakers should consider the specific conditions and context in which public debt operates. Overall, the interaction effects reveal intricate relationships between public debt and macroeconomic performance. The findings highlight the importance of considering the complex interplay between public debt and control variables, as well as the need for careful methodological consideration. By understanding these interaction effects, policymakers can develop more effective policies that account for the nuanced relationships between public debt and macroeconomic performance.
The interaction effects analysis has provided helpful insights about public debt and macroeconomic performance. This study pertinently contributes to the present literature by revealing that the impact of public debt on macroeconomic performance, particularly inflation and unemployment is enhanced by fiscal vulnerability. In some cases, while a positive relationship exists between public debt and unemployment, the interaction of public debt with some fiscal vulnerability indicators resulted in an inverse relationship. Similarly, whereas negative relationship exists between public debt and GDP, the interaction effect of debt service to domestic revenue resulted in a positive relationship. This proposes that in debt management, the government must focus keenly on managing these fiscal vulnerability indicators to limit the negative impact of its debt level on macroeconomic performance. The ministry of finance should ensure that the central government borrows within its ability to pay based on domestic revenue and not just the general size of the economy. With no fiscal space to borrow, the government should resort to new tax introductions or increments. In exercising their oversight responsibility, parliament must ensure the ministry of finance upholds financial discipline by requesting a periodic update on the fiscal vulnerability indicators.
Furthermore, this study reveals that the fiscal vulnerability indicators are potential influencers of how public debt affects the macroeconomic performance indicators or how the macroeconomic indicators stimulate a rise in public debt stock. Hence, the government must implement debt management policies that peg the fiscal vulnerability ratios at a favorable level to achieve debt sustainability to stimulate economic recovery and performance.
This study is the first to investigate the four leading indicators of macroeconomic performance and public debt, including the interaction effect of fiscal vulnerability. However, it focused on only Ghana as a developing country. Due to potential differences in macroeconomic, political and socio-cultural conditions among developing countries, the relevance of our findings is limited to developing countries with similar conditions. Therefore, a future study should examine these indicators in other jurisdictions or expand the scope to include more developing countries in West Africa or Sub-Saharan Africa to see if the same or different results will be achieved.
The supplementary material for this article can be found online.
