D6.3 Policy Brief: The role of debt for fiscal effectiveness during crisis and normal times

Transcription

1 MACFINROBODS FP7-SSH D6.3 Policy Brief: The role of debt for fiscal effectiveness during crisis and normal times Project acronym: MACFINROBODS Project full title: Integrated Macro-Financial Modelling for Robust Policy Design Grant agreement no.: Due-Date: 30 April 2016 Delivery date: 18 April 2016 Lead Beneficiary: UNIPV Dissemination Level: PU Status: submitted Total number of pages: 6 This project has received funding from the European Union s Seventh Framework Programme (FP7) for research, technological development and demonstration under grant agreement number

2 POLICY BRIEF 1 Does the level of public debt affects the fiscal multiplier? When production decreases and the output gap increases, counter-cyclical measures (such as a higher government expenditure) can boost the demand for goods and services. Aggregate supply returns to initial levels in order to match the increased demand and the output gap decreases. In this case, an expansionary fiscal policy has a positive effect over production. The fiscal multiplier measures the change in output relative to the change in government spending, that is, the increase in output induced by the change in government spending. The impact of government expenditure on output is a long-standing question in macroeconomics. For instance, Blanchard and Perotti (2002) show a positive effect of government spending on output and Blanchard and Leigh (2012) find that multipliers have been systematically lower since the start of the Great Recession. Auerbach and Gorodnichenko (2012) show that fiscal policies are considerably more effective in recessions than in expansions. Another key factor, which determines the effectiveness of government stimuli, is the interest rate. Two seminal papers that analyse government spending multipliers in a New Keynesian DSGE framework when the nominal interest rate is bounded at the zero lower bound are those by Woodford (2010) and Christiano, Eichenbaum, and Rebelo (2011). In both papers, the multiplier effect is substantially large when monetary policies are constrained by the zero lower bound on nominal interest rates. The larger the fraction of government spending that occurs when the nominal interest rate is zero, the larger the value of the multiplier. Others authors stress that high levels of government debt could diminish the fiscal multiplier. If the debt to GDP ratio is too high, expansionary fiscal policies could not have more muted impact over output, because agents would expect a spending reversal in a short period of time. For instance, Favero and Giavazzi (2007) show that at some point, an increase in government spending might respond to the level of the public debt since the government intertemporal budget constraint will eventually have to be met. Nickel and Vansteenkiste (2008) exhibit evidence that households in high debt countries tend to increase present savings to overcome a possible higher future expenditure, by showing that the relationship between fiscal deficits and current accounts become statistically insignificant when public debts are relatively high. Deák and Lenarcic (2012) find that expansionary government expenditure boosts output in periods of low debt to GDP ratio but it decreases output as soon as the debt to GDP ratio reaches a certain threshold value. 1 The research leading to these results has received funding from the European Community s Seventh Framework Programme (FP7/ ) under grant agreement Integrated Macro-Financial Modeling for Robust Policy Design (MACFINROBODS, grant no ).

3 The relationship between the GDP and the government expenditure (GE) seems to be non-linear for some countries such as Belgium, Italy, and United Kingdom (UK). Scatter diagrams at figures 1, 2, and 3 (see below) show a positive slope in the data when GDP is drawn in the horizontal axis (x) and GE in the vertical axis (y). This slope is relatively constant until a certain period, from which the government expenditure starts to grow faster than the GDP. Therefore, there is a change in the trend after both variables reach certain levels. Figure 4 shows the debt to GDP ratio time series (Debt) also for Belgium, Italy and UK. By intuition, there is at least one structural break in the behaviour of these variables around the year 2008 and a regime shift after this year. In this graph, a break point divides the data in high and low levels of the debt to GDP ratio for each country, changing the trend. Tariffi (2105) finds that the government debt actually has an impact on the relationship between government expenditure and economic growth. It shows different fiscal multipliers when it is considered multivariate threshold models with both low and high levels of public debt. Non-linear behaviours in sovereign debt to GDP ratio time series determine the relationship between output and government expenditure. A general multivariate threshold autoregressive model is used to explain this nonlinear relationship in all three above-mentioned countries. 2 This model improves robustness adding other fundamentals variables in the right hand side of the equation rather than keeping government expenditure (GE) as the only GDP determinant. In other words, GE remains statistically significant even if the model is augmented with other new variables. Table 1 (see below) shows delays chosen and threshold values for each country and the number of observations in both regimes. Therefore, considering the sum of all significant coefficients of the government expenditure in differences (DGE) lagged 1, 2, 3, and 4 periods, fiscal multipliers in the very short term are 0.2 and -0.2 in Belgium, -0.8 and 0.5 in Italy, and 0.2 and -0.5 in United Kingdom for regimes 1 and 2, respectively. 3 The two main results are the following. First, government expenditure has an impact on the GDP in both regimes in all countries. In Belgium and United Kingdom, an expansionary government expenditure effect is positive with low values of debt to GDP ratio but it is negative when the debt to GDP ratio increases. In Italy, this result is not conclusive. Second, the data shows at least 2 regimes in a non-linear framework if the debt to GDP ratio 2 The model includes not only GE, GDP, and Debt but also trade openness (TO), consumer prices (PI), and interest rate (IR). Variables are included in differences (integrated of order 0) and lagged 1, 2, 3, and 4 periods. The non-linearity is captured by the threshold variable (Debt). Threshold values divide the model in two regimes (good times and bad). There are 4 delays to determine the specification that minimizes the sum of squared residuals since each threshold variable is lagged between 1 and 4 periods in each country. Threshold variables are chosen according to the Akaike selection criteria (AIC) which asymptotically specifies the model with the smallest residual variance as in Tsay (1998). Therefore, the debt to GDP ratio has been selected optimally as a endogenous threshold variable to evaluate non-linearities. The threshold model is also selected through a general to specific methodology after time series have been transformed to I(0). This procedure can be used in the multivariate thresholds autoregressive model because a priori there are not restrictions in the parameters. In this model, the Akaike selection criteria (AIC) only choose the best lags structure (delay) for the threshold variable. For further discussion on general to specific methodologies see Campos, Ericsson, and Hendry (2005). 3 Coefficients are statistically significant at least at 10% in the first regime of each country with the only exception of the value corresponding to the debt to GDP ratio in Italy. Trade openness and the GDP itself (both lagged 1 period) are highly significant in both regimes in Belgium. The debt to GDP ratio is only statistically significant in the first regime of this model. The threshold value is 1.28 and the R2 is equal to 98%. In Italy, debt to GDP ratio is only significant in the regime 2. Apart from the GDP lagged 3 periods, all variables included in this model are significant in both regimes. Government expenditure is not only significant with 2 lags but also with 4. The threshold value is 0.19 and the model explains the data well since the R2 is higher than 70%. In United Kingdom, government expenditure with 3 lags, prices with 4 lags and the GDP with 1, 2 and 3 lags are significant in both regimes. The debt to GDP ratio is statistically significant only in the first period and the threshold value is According to the R2, which is equal to 95%, the threshold model fits the data very well.

4 is used as a threshold variable. However -following Tsay (1998)-, if the debt to GDP ratio is included in the least squares regression, this variable has an impact on the GDP only in the regime 1 in Belgium and United Kingdom but it does in the regime 2 in Italy. These results show that debt to GDP ratio (Debt) is an important financial variable to explain the behaviour of fiscal multipliers. The debt to GDP ratio has been useful to identified estimators in a multivariate threshold autoregressive model and it has been an important tool to observe how the fiscal multiplier changes during good times (low levels of debt to GDP ratio) and bad times (high levels of debt to GDP ratio). Therefore, multivariate threshold models can be employed in order to explain fiscal multipliers considering levels of public debt. Time series related to public finance, (such as government debt to GDP ratio) usually show non-lineal behaviours. The inclusion of this type of variables in models which explain fiscal multipliers requires non-linear methodologies. Expansionary fiscal policies seem to be counterproductive, and even negative, when the level of sovereign debt is taken into account. REFERENCES Auerbach, Alan J., and Gorodnichenko Yuriy (2012), Measuring the Output Responses to Fiscal Policy, American Economic Journal: Economic Policy, vol. 4, n. 2, pp Blanchard, Olivier., and Leigh, Daniel (2012), Are We Underestimating Short-Term Fiscal Multipliers?, in World Economic Outlook of the International Monetary Fund, vol. October, pp Blanchard, Olivier., and Perotti, Roberto (2002), An empirical characterization of the dynamic effects of changes in government spending and taxes output", The Quarterly Journal of Economics, vol. 117, n. 4, pp. 40. Campos, Julia., Ericsson, Neil R., and Hendry, David F. (2005), General-to-specific Modeling: An Overview and Selected Bibliography, International Finance Discussion Papers, n Christiano, Lawrence., Eichenbaum, Martin., and Rebelo, Sergio (2011), When Is the Government Spending Multiplier Large?, Journal of Political Economy, vol. 119, n. 1, pp Deák, Szabolcs., and Lenarcic, Andreja (2012), The Fiscal Multiplier and the State of Public Finances, mimeo. Favero, Carlo., and Giavazzi, Francesco (2007), Debt and the effects of fiscal policy, NBER Working Papers, n Nickel, Christiane., and Vansteenkiste, Isabel (2008), Fiscal policies, the current account and Ricardian equivalence, ECB Working papers, n Tariffi, Leonardo A., (2015), A threshold multivariate model to explain fiscal multiplier7s with government debt, mimeo, University of Pavia. Tsay, Ruey S. (1998), Testing and Modeling Multivariate Threshold Models, Journal of the American Statistical Association, vol. 93, n. 443, pp Woodford, Michael (2010), Simple Analytics of the Government Expenditure Multiplier, NBER Working Paper, n

5 1996Q1 1997Q3 1999Q1 2000Q3 2002Q1 2003Q3 2005Q1 2006Q3 2008Q1 2009Q3 2011Q1 2012Q3 2014Q1 Debt to GDP ratio Gross Domestic Product Gross Domestic Product Gross Domestic Product FUGURES AND TABLES Figure 1: Belgium. Government Expenditure w ith respect to Gross Domestic Product (millions of euros at constant prices) Government Expenditure Figure 2: Italy. Government Expenditure w ith respect to Gross Domestic Product (millions of euros at constant prices) Government Expenditure Figure 3: UK. Government Expenditure w ith respect to Gross Domestic Product (millions of euros at constant prices) Government Expenditure Figure 4: Debt to GDP ratio Belgium Italy United Kingdom

6 Table 1: Multivariate Threshold Model for the GDP Least Squares Regression Belgium Italy United Kingdom Delay chosen: D2DEBT(-2) Delay chosen: D2DEBT(-2) Delay chosen: D2DEBT(-4) Threshold value: Threshold value: Threshold value: Period: 1997Q2-2015Q1 Period: 1997Q2 2015Q1 Period: 1997Q3 2015Q1 Regime 1: 32 obs Regime 1: 43 obs Regime 1: 46 obs Variable Coefficient t: Prob. Variable Coefficient t: Prob. Variable Coefficient t: Prob. C C C DGE(-3) DGE(-2) DGE(-3) DTO(-1) DGE(-4) DPI(-1) DTO(-4) DTO(-3) DPI(-3) DPI(-4) D2PI(-1) DPI(-4) DIR(-1) DGDP(-1) D2GDP(-1) DIR(-4) DGDP(-3) D2GDP(-2) DGDP(-1) D2DEBT(-1) D2GDP(-3) D2DEBT(-1) D2DEBT(-1) Regime 2: 40 obs Regime 2: 29 obs Regime 2: 25 obs Variable Coefficient t: Prob. Variable Coefficient t: Prob. Variable Coefficient t: Prob. C C C DGE(-3) DGE(-2) DGE(-3) DTO(-1) DGE(-4) DPI(-1) DTO(-4) DTO(-3) DPI(-3) DPI(-4) D2PI(-1) DPI(-4) DIR(-1) DGDP(-1) D2GDP(-1) DIR(-4) DGDP(-3) D2GDP(-2) DGDP(-1) D2DEBT(-1) D2GDP(-3) D2DEBT(-1) D2DEBT(-1) R-squared: R-squared: R-squared: Note: Time series have been transformed to I(0) Optimal delays are 2, 2, and 4 for Belgium, Italy, and UK, respectively. D means 1st differences, D2 is 2nd differences, C is the constant, and the negative number in parenthesis is the lag of the variable. Columns 2 and 3 in each country present estimations and p-values associated to the t statistics.