關於外債和經濟增長的論文

Abstract

本文探討了政府負債與經濟成長之間的長期關係。為了研究兩者之關係，我們考慮了經濟成長的多元要素之廣泛性，以及專業文獻呈現之錯綜複雜的觀點，比如其中的逆向因果關係和內因性，因此，我們將從數十年間之跨國與國內兩層面進行探討。\n我們利用10個迴歸模型對103個國家的縱橫資料進行分析以及對美國進行時間序列分析。我們在跨國分析的部分，包含運用了追蹤資料分量迴歸分析、非動態縱橫資料門檻迴歸分析以及分流樣本門檻迴歸分析以呈現舉債對所有國家之經濟成長在統計上的負面影響。同時，在國內分析的部分，我們利用了分量迴歸分析、滯後變數模型分析和轉折點門檻迴歸模型進行分析。\n整體而言，我們的實證結果揭示了舉債對長期經濟成長之有害關聯。在跨國研究的部分中隱含，舉債因該經濟體之經濟成長率，對經濟成長有著錯綜複雜的影響，而當該國舉債佔GDP比率超過92.15％時將產生不良影響，超過該門檻之影響程度則取決於該國體制特徵。\n首先，我們從實證去觀察舉債與經濟成長之長期關係，並著重於舉債在不同經濟成長區段之影響。事實上，這個概念架構是使人們能夠不需運用變數下，理解政府舉債額度如何影響該國的人均國內生產毛額之成長。\n我們運用模型1，固定效用分析之分量迴歸模型，其中包含兩種不同制度變數，即民主程度與制度品質。首先，經濟成長受到舉債影響的程度因不同經濟成長率而異。再者，其顯著性也因其百分位數有所不同。我們證實了舉債在1%的顯著水準下在第25百分位數時，對經濟成長造成限制。\n而後，我們運用模型2，靜態追蹤資料中的門檻效應，來測試舉債水準對人均 GDP 增長的非線性影響，以識別在兩個變數在不同狀態中所呈現之不對稱關係。我們發現在有制度變數的模型 2（9.26、48.9、125.22）中，舉債和人均 GDP 之間存在三個門檻效果，而舉債在其他狀態中均限制了人均 GDP之成長。\n接著，我們使用模型3，制度品質的樣本分流與門檻評估，以及模型4，對民主程度樣本分流與門檻評估衡量舉債，以呈現制度之異質性與舉債成長之關聯。在對制度面之評估結果中得出，對於舉債規模佔國內生產毛額高於92.15%以上的國家，制度對於舉債之影響非常關鍵。在兩模型中，制度品質與民主程度低落的國家承受了舉債負擔帶來的負面影響，但在制度完善的國家，其制度對舉債增長的影響尚無定論。\n在國內研究的部分，我們的分量迴歸分析結果指出，舉債對經濟成長的影響，在所有不同成長百分比率並不一致；債務成長與人均國內生產毛額呈現共整合，且另外在短期和長期具有變換的趨勢。此外，我們觀察到舉債因其成長率對經濟成長有著不對等的影響，當門檻值為 2.5%時，低於門檻值對經濟成長有正向影響，當負債成長率高於 2.5%時，舉債將造成經濟成長的限制。\n隨後，在模型5，非線性自我迴歸分配遞延模型區間測試，舉債之變化對國內生產毛額變化之影響被重新檢視。相同地，非線性自我迴歸分配遞延模型區間測試法執行之共整合檢定證實了共整合的存在。此外，其短期和長期影響在兩個回歸模型中具有顯著性差異並趨於收斂。建模結果也顯示，經濟成長在短期內會受到舉債的刺激，但在長期會受其限制。總而言之，短期的影響是正面的；但在長期下將慢慢收斂，對人均GDP成長造成37.3%的反效果。\n接著，我們使用模型6，時間序列分量迴歸分析法來估算舉債對整體成長率分佈之影響。結果顯示，負債成長對人均 GDP 在各項條件百分位數上有著各種影響。我們的模型顯示，低百分位數和高百分位數在所分析的變數之間產生了正向的交互關係。舉例來說，在第 90 個百分位數的正面影響在10% 的顯著水準時具有顯著性差異。同時，在第 30 到 60 個百分位數之範圍，舉債對赤字率的成長具有負面的非統計性結果。\n最後，本論文運用 Fong、Huang、Gilbert 和 Permar（2017 年）發表的轉折點門檻迴歸分析，探討政府舉債之成長對人均 GDP成長之非線性影響。負債成長具有門檻區隔效應，在模型 7轉折點門檻迴歸分析：門檻區隔效應， 有2.5%之機率轉換狀態。在舉債低成長時期，舉債可以刺激經濟成長；然而，在高成長時期，舉債將造成負面影響降低經濟成長。模型8、9和10則證實，舉債不具有門檻轉折、連續或區隔效應。

This thesis explores the long-term relationship between government debt and economic growth. To examine this relationship, we used a cross-country and within-country dimension over several decades that considers a comprehensive variety of economic growth factors and estimation complications exposed in the specialized literature, such as the reverse causality and endogeneity.\nWe use ten models to analyze panel data of 103 countries and time series analysis of the United States of America. Our cross-country analysis involves using panel quantile regression, threshold regression for non-dynamic panels, and threshold regression with sample splitting to show the negative statistical impact of debt on economic growth in all countries. Meanwhile, we use quantile regression, autoregressive distributed lag model, and the threshold change-point regression model for within-analysis.\nGenerally speaking, our empirical results expose a harmful association between debt and long-term economic growth. The cross-country study suggests that debt has a heterogenous impact on economic growth depending on the economy`s growth rate, debt has adverse effects above a 92.15% GDP ratio, and the impact above the threshold will depend on the institutional characteristics such as institutional quality or democratic qualities of the countries.\nFirst, we empirically investigate the long-run association of debt on economic growth, considering its impact on different growth ranges. In effect, this framework enables one to comprehend how the governmental amount of debt influences a country’s gross domestic product per capita growth beyond the usage of the means of such variables.\nWe applied Model 1 Quantile Regression with Fixed Effects analysis which includes two institutional variables, namely democratic quality and institutional quality. Firstly, the impact magnitude of debt on economic growth differently depending on the rate of its economic growth. Secondly, its significance varies, too, depending on the percentile. We confirm that debt constrains economic growth at the 25th percentile with a 1% significance level.\nLater, we employ Model 2 Threshold effects in non-dynamic panels to test the non-linear effects of debt levels on GDP per capita growth to identify the different regimes exposing an asymmetrical relationship between both variables. We find a three threshold effect between debt and GDP per capita in Model 2 (9.26, 48.9, 125.22) with institutional variables. Debt constrains GDP per capita growth in all regimes except in the third regime.\nNext, we evaluate the debt in a Model 3 Sample Splitting and Threshold Estimation for Institutional Quality and Model 4 Sample Splitting and Threshold Estimation for Democratic Quality to address the heterogeneity of institutions in the debt growth link. The institution`s estimation results reveal that the impact is of great significance in determining the effect of debt in countries that have debt magnitudes measured to the gross domestic products above 92.15%. In both models, countries with the low institutional quality and democratic quality suffer from the negative impact of debt overhang, but in countries with strong institutions, the effect of debt on growth is inconclusive.\nRegarding the within-country study, our quantile regression results indicate that the impact on growth is not uniform across the different growth percentiles; debt growth is cointegrated with GDP per capita and has other effects in the short term and long term with a tendency of conversion. Additionally, we observed debt having an unequal impact on growth depending on the debt growth rate, with a threshold point at 2.5%, where below the threshold point, the effects on growth are positive, and when the debt growth rate is higher than 2.5% debt constrains economic growth.\nLater, the gross domestic product changes caused by changes in debt are revisited in the short and long term in Model 5 Autoregressive-Distributed Lag Bounds Testing Approach to Cointegration. Again, the autoregressive distributed lag model bound test of cointegration confirms the existence of cointegration. Moreover, the short-term and long-term effects are statistically significant, which tends to converge in both regression models. The modeling results show that the economy per capita is stimulated by debt in the short term but limited by debt in the long term. Consequently, short-term effects are positive; but slowly converge to long-term adverse effects on growth at 37.3%.\nNext, we used the Model 6 Time Series Quantile Regression approach to estimate the effects of debt on the entire growth rate distribution. Results suggest that debt growth effects on GDP per capita are heterogenous across conditional percentiles of GDP per capita growth. Our models reveal that the low and upper percentiles create positive interactions among the variables analyzed. For example, the positive impact on the 90th percentile was statistical significance at a 10% significance level. Meanwhile, around the 30 to 60th percentile, Debt-to-GDP growth has a negative non-statistical outcome.\nFinally, this dissertation examines the non-linear impact of Public Debt Growth on GDP per capita growth using the threshold change-point regression developed by Fong, Huang, Gilbert, and Permar (2017). Debt growth has a threshold segmented effect with a regime switch at 2.5 percent in Model 7 Change point threshold regression: Threshold Segmented Effects. In a low debt growth regime, debt can stimulate economic growth; however, in a high debt growth regime, debt has an adverse impact reducing economic growth. Models 8, 9, and 10 confirm that debt does not have the threshold hinge, step, or stegments effects.