Analyse de l'impact de la dette publique sur la croissance économique au Burundi (1980-2024J

Abstract

This work is entitled ~Î Analysî:s of the impact of public debt on economic growth in Burundi (1980-2024)>>. Based on macroeconomic indicator data incorporated iuta the analytical models, the results of the ADF and pp tests showed that the variables are integrated of mixed order 1(1) and 1(0), \Vith no variable integrated of order 2; this validatcd the use of the ARDL mode!. Using the ARDL mode! selectcd according ta the Hannan-Quinn critcrion, the adjusted R2 = 0.823 ( 82.3%) indicatcs that the model explains virtually aIl ofthe variance in cconomic growth. The F-statistic is highly significant (prob = 0.00024), lllcaning that the model is globally significant. The bounds test yiclded an F-statistic of 12.53, weil above the upper bound 1(1) = 3.61 at the 5% levcl or 1(1) = 4.43 at the 1% levcl; this confirms the existence of a longrun relationship. AccordinglYl estimation of the error correction mode! (ECM) produced a fecdback coefficient (CointEq(-l)) of -1.94, highly significant (p - value ~ 0.0000); this showed that nearly 194% of the previous year's disequilibrium \Vere corrected in the following year. ln the short run, public debt had a significantly negative impact (coefficient of -48.32 for the first difference and p - value = 0.0000), but generates a recovery after 1 ycar (a rebound effect) before a renewed negative effect. In the long run, the impact of public debt remains negative but is not statistically significant (coefficient of -0.219 for the first difference, p = 0.8698), indicating that the harmful effect fades over the period. Model validation tests confirm the absence of major biases, and the CUSUM and CUSUM of squares stability tests attest to the robustness of the ARDL-ECM model. Threshold analysis using the Hansen model identified a breakpoint around 101.6% of the public debt to-GDP ratio for economic growth, but bclow this threshold, the effect is positive (coefficient = 9.11) and significant (p-value = 0.031). Above this threshold, the effect remains positive (coefficient = 6.26) but is not significant (p-value = 0.064). As for the threshold analysis using the quadratic model, a thrcshold of 128.42% \Vas round. This indicatcs that the positive effect changes and tums negative. This confirms the non-linearity of public debt. The results of the ARDL-ECM model estimation confirm the relationship between public debt and economic growth in Burundi. The thresholds round after estimating the threshold models attcst ta the non-linearity of public debt. \·Vith these results, wc mainly recommend prudent management of public debt and strengthcning the mobilization of tax revenues ta crcate sustainable fiscal spacc.