MEFMI Macroeconomic & Financial Management Institute of Eastern and Southern Africa

MEFMI Macroeconomic & Financial Management Institute of Eastern and Southern Africa NATIONAL DEBT AND INFLATION: EVIDENCE OF THE FISCAL THEORY OF THE PRICE LEVEL FROM KENYA A Technical Paper Submitted in Partial Fulfillment of Requirements for Accreditation to MEFMI Written and submitted by Sheila Kaminchia, Assistant Researcher II, Research and Policy Analysis Department, Central Bank of Kenya, P. O. Box 60000-00200, Nairobi, Kenya. Telephone: +254 20 2863245 Email: Kaminchias@centralbank.go.ke May 2014 Disclaimer: The views expressed in this paper are the authors' alone, and in no way reflect those of the Central Bank of Kenya.

Abstract This study looks for evidence of the fiscal theory of the price level in the data for Kenya. The fiscal theory of the price level envisages a situation where fiscal policy interferes with the effectiveness of monetary policy in controlling inflation primarily by occasioning an increase in private wealth. The study finds evidence that monetary policy in Kenya is active in determining inflation only one year after an initial policy response to inflation. Thereafter, fiscal policy becomes more significant in determining inflation. The study also finds that interest rates have a more significant effect on public debt than does reserve money. A monetary policy framework where the interest rate is a policy instrument would, therefore, be more appropriate in managing Kenya s inflation over a medium-term horizon. i

1. Introduction Kenya s development plan which is called the Vision 2030 requires significant fiscal space for implementation of the medium-term plans. The resource framework of Vision 2030 initially allocates about 60 percent of total available resources to infrastructure projects. The development plan is arguably ambitious on government revenue targets. The tax space in Kenya is narrow and this raises pressure to accumulate public debt, which fortunately is comparably low but the growth of which is significantly higher. It is the growth in the domestic debt and not the stock of domestic debt alone that has implications for inflation. Fiscal discipline is highlighted in the Vision 2030 documents and is defined in relation to net present value of debt. The role of monetary policy in the Vision 2030 is to stabilize medium-term inflation. Fiscal policy has implications for the success of monetary policy, the more important one being the central bank s ability to control inflation when the fiscal deficit (and the real value of debt) is high. The standard practice in the event of inflation rising beyond the policy target is for the central bank to raise nominal interest rates. It, however, becomes less effective for monetary policy to influence inflation when the level of public debt is rising. In such a case, raising nominal interest rates would also increase nominal public debt through higher interest payments. The outcome of inflation from an increase in the policy interest rate, however, depends on the thinking of economic agents. According to the monetary theory of price determination, the central bank is expected to increase its policy rate if current inflation rises above the predetermined target. The increase in interest rates will reduce inflation but will increase the public debt. The government is expected to increase taxes or reduce spending so that the value of debt does not increase above the present value of primary surpluses, thereby leaving the real value of debt unchanged (Leeper and Walker, 2011; Baldini and Ribeiro, 2008). Therefore, an increase in government debt will lead the private sector to expect a proportionate increase in taxes or a fall in fiscal spending in future. The private sector will then not adjust their spending plans leaving aggregate demand unchanged. Since there will be no wealth effects arising from a change in nominal debt, then fiscal policy will have no effect on inflation. In this case, monetary policy is active while fiscal policy is passive in determining inflation (Eusepi and Preston, 2011; Leeper and Walker, 2011; Sims, 2011). This proposition, however, breaks down when the private sector only considers the nominal value of public debt. 1

According to the fiscal theory of price determination, there is no private sector expectation that future taxes or public spending will be fully adjusted in view of changing public debt. In the event that the private agents observe an increase in the nominal value of public debt and expect the government not to raise taxes in future to finance the higher debt, then the private agents will necessarily perceive an increase in their personal wealth given prevailing steady prices. This perception leads to an increase in private consumption spending, which will eventually raise domestic prices (Eusepi and Preston, 2011; Leeper and Walker, 2011; Sims, 2011). Even if the central bank was to raise interest rates, inflation will not decline but will rise instead. This is because a higher interest rate will add to the interest expense of the government and fuel a further increase in the nominal value of public debt and add to public perception of increased real wealth (Sims, 2011; Leeper and Walker, 2011). Ensuing inflation will, however, reduce real wealth and consumption will fall accordingly. The central bank may also decide not to fully adjust nominal interest rates in order to reduce inflation to the target level with minimal effect on the nominal value of public debt. This is a case of monetary policy being passive and fiscal policy being active in determining inflation (Eusepi and Preston, 2011). In the period 2000-2011, monetary policy in Kenya held down interest rates while fiscal policy stimulated the economy. The government s indebtedness increased as the primary balance largely remained in deficit (Table 1). Interest payments on domestic debt also eventually increased despite lower interest rates. The performance of the Kenyan economy peaked in the 2006-2008 period while inflation remained marginally higher after 2002. Table 1: Selected Indicators on Domestic Debt for Kenya Period average 2000-2002 2003-2005 2006-2008 2009-2011 Real GDP growth 1.6 4.6 5.0 4.1 Nominal domestic debt as percent of GDP 22.3 24.5 23.1 27.8 Domestic interest payments as percent of GDP 2.4 2.0 2.1 2.4 Primary surplus as percent of domestic debt -1.6 1.6-2.5-0.8 91 treasury bill rate 11.2 5.0 7.1 6.6 Annual average CPI inflation 5.9 10.5 9.0 9.1 Sources: Central Bank of Kenya, Kenya National Bureau of Statistics A more relevant issue for the case of Kenya is the role of fiscal policy in promoting the supply side and where it fails, inflation. This will affect the credibility of monetary policy. The purpose of this study is 2

to understand the interaction between monetary policy and fiscal policy in Kenya, and the implications for inflation management. 2. Literature review The fiscal theory of the price level posits that monetizing fiscal deficits is not the only way to generate inflation. The outturn of inflation also depends on the manner in which public debt is financed. Central bank financing of the government will more likely be inflationary if at the same time the government s debt is fully funded. Government operations may, on the other hand, add to inflationary pressures if the operations result in underfunding of public debt. In such a case, the public is left to guess where the money to pay off the debt will come from. Should private agents expect the government to increase taxes in future to raise the required funds, then private agents will not expect that their wealth will increase with the government s indebtedness. Otherwise, private agents will adjust their spending upwards (downwards) in response to an increase (a decrease) in the nominal value of government debt, and in so doing directly affect aggregate demand (Leeper and Walker, 2011). Alfonso (2002) empirically assesses the fiscal theory of the price level using panel data for 15 countries of the European Union for the period 1970-2001 and finds no evidence to support the theory. The study finds that the EU-15 governments tend to increase the primary budget surplus when the debt-to-gdp ratio rises, which is evidence of monetary dominance (or Ricardian regime). The method used in this study is to regress (i) primary surplus as percent of GDP on debt-to-gdp ratio at first lag and test that the coefficient of debt-gdp ratio, θ, is equal to zero (evidence of fiscal dominance or non-ricardian regime) or greater than zero (evidence of monetary dominance or Ricardian regime). The exercise found θ to be significant and positive. (ii) debt-to-gdp ratio on the primary surplus and tests the hypothesis that the coefficient of the primary surplus, ٧, is less than zero (evidence of monetary dominance) or equal/greater than zero (evidence of fiscal dominance) and finds ٧ to be negative and significant. (iii) average annual change of the price deflator of private final consumption expenditure on its first lag and the first lag of the first difference of total public receipts as a percent of GDP and finds its coefficient to be positive but insignificant indicating monetary dominance. 3

Following Canzoneri, Cumby and Diba (2001), Javid, Arif and Sattar (2008) estimate VARs using annual data for Pakistan for the period 1971-2007 on (i) primary surpluses, public liabilities, debt, inflation, reserve money, seigniorage and output gap, the results from which show a negative response of public liabilities to an increase in the primary surplus indicating monetary dominance. (ii) surplus to GDP, natural logarithm of public liabilities and natural logarithm of GDP, the results from which show a negative response of nominal income and public liabilities to an increase in the surplus thus supporting the evidence of monetary dominance found earlier. (iii) primary surplus, public liabilities and discount rate, the results from which show that liabilities to GDP respond negatively and the discount rate respond positively to an increase in the primary surplus, again supporting monetary dominance. (iv) domestic debt growth, reserve money, real output gap and inflation, the results of which show that growth in domestic debt has more influence on inflation than growth in reserve money, supporting the fiscal theory of the price level. The authors conclude that nominal public debt or money growth matter for price stability in Pakistan. Canzoneri, Cumby and Diba (2001) followed this method in an earlier work and found evidence in support of the Ricardian regime for the case of the United States of America. Also following Canzoneri, Cumby and Diba (2001), VAR analysis is used by Baldini and Ribeiro (2008) to investigate monetary response to inflation shocks and the effect of wealth on inflation in 22 countries in Sub-Saharan Africa. The results from the cointegration exercise show that out of 22 countries considered, 4 including Burundi and Tanzania had fiscal dominant regimes, 5 others including Kenya and Rwanda had monetary dominant regimes, while no regime could be identified for the remaining countries, including Uganda. In other results, domestic debt growth explained variability in inflation in 5 countries including Burundi, money growth explained variability in inflation in 6 counties including Kenya and Tanzania, and both money and debt growth explained variability in inflation in 5 countries including Uganda. VAR results confirmed evidence of fiscal dominance in Burundi and Tanzania. Canzoneri et al. (2004 pp. 21) show that within the European Monetary Union (EMU), the rates of inflation for member states is not perfectly correlated with the aggregate inflation rate implying asymmetric impact of the common monetary policy. The study shows that productivity shocks have 4

significantly more influence on inflation differentials across EMU members than differences in fiscal policy. Raising government spending adds to output (through investment rather than consumption) and inflation. This is attributed to monetary policy letting real interest rates to fall through weaker response to rising inflation for small countries. This finding implies that an increase in government spending encourages savings rather than consumption. Higher interest rates and lower consumption tax receipts then add to the fiscal deficit thus constraining fiscal policy. In other dynamics, Cochrane (2001) considers the maturity structure of public debt and argues that instability in future surpluses can be stabilized using long-term debt, while Bauducco and Caprioli (2011) demonstrate that the ability of governments to share risks through external borrowing dampens volatility in tax rates and domestic debt. The contrary is characteristic of emerging economies. 3. Theory and Econometric Method Fiscal dominance in price determination arises when an upward trend in the fiscal deficit is not backed by an upward trend in real taxes. A rise in price level is, therefore, necessary to stop private economic agents from spending out of perceived increase in wealth. In a monetary dominant regime, innovations in interest rates should stabilize prices. So while monetary policy is adjusting interest rates to stabilize inflation, fiscal policy is helping by adjusting tax rates to stabilize real value of debt (a case of fiscal policy accommodating monetary policy). The central bank can, therefore, create inflation depending on the response of the fiscal authorities. If fiscal policy is not accommodative, then raising the monetary policy interest rate will add to inflation. If fiscal policy is accommodative, then raising the monetary policy interest rate will reduce inflation. This study will assess the co-movement in nominal interest rates, primary surpluses and inflation for Kenya. The foregoing analysis seeks evidence of a strong positive relationship between real taxes and public debt and a weak positive relationship between nominal interest rates and inflation (a case of fiscal policy accommodating monetary policy). In Figure 1 below, real taxes have been trending upwards with real public debt, whereas in Figure 2, an increase in inflation did not always induce an increase in the government s borrowing rate 1. 1 There is no long enough series on the monetary policy rate, which is the central bank rate. The 91- day Treasury bill rate is, therefore, used instead since there is a close correspondence between the monetary policy rate and the Treasury bill rate. 5

Jun 01 Dec 01 Jun 02 Dec 02 Jun 03 Dec 03 Jun 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Dec 08 Jun 09 Dec 09 Jun 10 Dec 10 Jun 11 percent Jun 01 Dec 01 Jun 02 Dec 02 Jun 03 Dec 03 Jun 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Dec 08 Jun 09 Dec 09 Jun 10 Dec 10 Jun 11 million Kenya shillings million Kenya shillings 5,200 5,000 4,800 4,600 4,400 4,200 4,000 3,800 3,600 3,400 3,200 3,000 2,800 Tax revenue -left scale Public debt -right scale 2,800 2,600 2,400 2,200 2,000 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0-200 Figure 1: Real tax revenue and real public debt 20 18 16 14 12 91-Day Tbill 12-month inflation 10 8 6 4 2 0 Figure 2: Inflation and Nominal Treasury bill interest rate 6

The analytical framework used in this study draws from Canzoneri, Cumby and Diba (2001). The choice is motivated by the method s focus on public debt financing. In a comprehensive system, household, central bank and government behavior would be modeled. In such a system, households maximize their utility subject to available resources characterized by endowment incomes and transfer payments from the government. Monetary policy enters the consumers budget constraint through an interest rate that is dependent on a change in the value of consumption (captured as a Taylor rule), while fiscal policy enters directly through the government s budget constraint (see Leeper and Walker, 2011; Sims, 2011; Sims, 1994). The method in Canzoneri, Cumby and Diba (2001) pays attention to the government s budget constraint, which they define in nominal terms as ( ) ( ) in which the government debt at the beginning of period j (B j ) must be funded through (i) amortization by reducing the primary surplus (Tj Gj) during period j (where T j is tax revenue and G j is government expenditure including interest payments), (ii) monetization by increasing the stock of base money (M j ) during period j, and (iii) additional borrowing at the prevailing the interest rate (i j ). Canzoneri, Cumby and Diba (2001) then re-write the budget constraint (1) in terms of total government liabilities (M + B) and divide the variables by nominal GDP. The new expression equates total government liabilities to GDP ratio (w j ) to the primary surplus to GDP ratio (s j ) that includes transfers from the central bank and next period s government liabilities to GDP ratio discounted by the ratio of real GDP growth to real interest rates as follows Equation 2 is then iterated forward to obtain the government s present value constraint as ( ) ( ) This expression says that if the government does not adjust it primary surplus with reference to its total liabilities, then the discount factor, α, will be endogenous in the constraint. This means that nominal income will also be determined within the constraint. Otherwise, nominal income will be determined outside the constraint, that is, either by central bank or household behaviour. The aim now is to identify a link between public debt (proxied by domestic debt to GDP ratio) and real taxes (proxied by the primary surplus to GDP ratio). This is done using a vector autoregressive model (VAR) that is set up as follows 7

4. Estimation results The vector autoregressive model (VAR) described by Equation 4 is applied to quarterly data for Kenya for the period 2000-2011. The VAR is estimated using data on net surplus (SURPLUS), which is total revenue minus total expenditure (including interest payments); and Liabilities (LIABILITIES), which is nominal domestic debt plus reserve money. The nominal net surplus and liabilities variables are both divided by the Kenya consumer price index (CPI) and then divided by real GDP to obtain real values as a ratio of GDP. Using the Augmented Dickey Fuller test, the real variables are determined to be stationary at the 5% level of significance. The results from the unit root tests are presented in Table 1 below. Table 1: Unit Root Test Results SURPLUS (s t ) LIABILITIES (w t ) Level -5.15-3.14 First difference -3.31 Critical values are -3.59 for 1% level; -2.93 for 5% level Given that the two variables, that is, SURPLUS and LIABILITIES, are stationary variables, they are used to estimate a vector autoregression (VAR) model in their levels form. Impulse response functions are derived from the estimated VARs. In order to obtain consistent results, one VAR is estimated using the ordering with SURPLUS first and the other VAR is estimated using the ordering with the first lag od LIABILITIES first. The first VAR allows SURPLUS to have a contemporaneous effect on LIABILITIES such that the non-ricardian regime holds and the second VAR does not allow SURPLUS to have a contemporaneous effect on LIABILITIES. The impulse response functions estimated from the two VARs are presented in Figure 3 below. Figure 3 shows that the response of the first lag of LIABILITIES to SURPLUS tends to be negative in both VARs. The results from the first 8

VAR further indicate that the negative response of public liabilities to an increase in the surplus is significant only for up to 4 quarters. This implies monetary dominance lasts in this system for up to one year. Response to Cholesky One S.D. Innovations ± 2 S.E..0008 Response of SURPLUS to SURPLUS.0015 Response of LIABILITIES(-1) to SURPLUS.0006.0004.0002.0000.0010.0005.0000 -.0005 -.0010 -.0002 -.0015 Response to Cholesky One S.D. Innovations ± 2 S.E..0008 Response of SURPLUS to SURPLUS.0015 Response of LIABILITIES(-1) to SURPLUS.0004.0010.0005.0000.0000 -.0004 -.0005 -.0010 Figure 3: VAR in Surplus and Liabilities The top panel ordering is Surplus Liabilities The bottom panel ordering Liabilities Surplus In order to obtain another view of the impact of fiscal variables on inflation, another VAR is estimated using data on consumer price index (CPI), ratio of domestic debt to GDP (debt), ratio of interest expense to tax revenue (interest expense), 91-day treasury bill (ST rate), ratio of the primary deficit to the market value of domestic public debt (primary deficit), real GDP and ratio of reserve money to CPI (reserve money). Results from the unit root tests conducted on these variables are presented in Table 2 below. The results show that CPI, interest expense, ST rate, real GDP and reserve money are integrated of order 1 while debt and primary deficit are stationary variables. CPI, interest expense, ST rate, real GDP and reserve money are, therefore, modeled in their first difference form while debt and primary deficit are modeled in their levels form. 9

responses Table 2: Unit Root Test Results CPI Debt Interest expense ST rate Primary deficit Real GDP Reserve money Level 1.70-3.35-2.27-2.79-5.78 0.80-0.77 First difference -3.78-15.13-3.93-3.14-6.66 1% level -3.59; 5% level -2.93; 10% level -2.60 The impulse response functions estimated from this VAR are presented in Figure 4 below. Figure 4 shows that an increase in reserve money, interest rates and stock of debt have delayed positive influence on inflation of up to 7 quarters; an increase in the primary deficit raises inflation in the first two quarters but subsequently lowers inflation; and there is no consistent impact of real GDP on inflation. Interestingly, an increase in the component of interest expense has largely negative impact on inflation. 0.60 0.40 0.20 0.00-0.20-0.40-0.60 period real GDP interest rate reserve money primary deficit Figure 4: Responses of Inflation to Cholesky One S.D. Innovations Cholesky Ordering: DREALDGP DCPI DINTRATE DREALRM PRIMDEFICIT DEBT DINTEXPENSE debt interest expense A review of the impact on inflation of a shock to the interest rate shows that a positive shock to interest rates initially raises reserve money and the interest expenses of the Government but initially reduces the stock of domestic public debt (Figure 5). Monetary expansion on the other hand seems to have less significant influence on fiscal variables than interest rates (Figure 6). This is evidence that fiscal side is important in the monetary transmission mechanism in Kenya particularly where interest rates interact with the stock of public debt. 10

Response to Cholesky One S.D. Innovations ± 2 S.E. Response of DREALRM to DINTRATE Response of PRIMDEFICIT to DINTRATE 100 6 50 4 0 2 0-50 -2-100 -4-150 -6.0004 Response of DEBT to DINTRATE.03.02 Response of DINTEXPENSE to DINTRATE.0000 -.0004.01.00 -.01 -.02 -.0008 -.03 Figure 5: Responses to Interest Rate Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of PRIMDEFICIT to DREALRM Response of DEBT to DREALRM Response of DINTEXPENSE to DREALRM 4.0006.03 2 0-2.0004.0002.0000 -.0002.02.01.00 -.01-4 -.0004 -.02 Figure 6: Responses to Reserve Money Shock 11

5. Conclusion and Policy Recommendations The present study is set against a background of a planned expansion in government spending to lay the foundations for enhanced growth in Kenya. This may pose a challenge for the management of inflation in Kenya if the expansionary fiscal stance does not accommodate monetary policy by raising domestic taxes. The aim of raising taxes is to steer public expectations away from rising wealth and thus contain inflationary private consumption expenditure as the fiscal debt rises. The role of interest rates in influencing inflation becomes ambiguous when interest rates significantly influence private wealth. Adjusting interest rates will have a desired influence on inflation if the monetary theory of price determination holds (monetary dominance or Ricardian regime). In this case, domestic taxes are raised proportionately with rising government debt so that private wealth and therefore private spending remains unchanged. With no wealth effects from rising nominal debt, fiscal policy will not interfere with the central bank s ability to control inflation using the policy interest rate. The results of this study confirm the findings of Baldini and Ribeiro (2008) that the Ricardian regime dominates in Kenya. The present study, however, finds that monetary dominance is only significant for up to one year. The study further finds that raising interest rates has a positive effect on inflation of up to 7 quarters (perhaps through increased borrowing costs) and has an immediate positive effect on reserve money and interest expense of government. The study also finds that the stock of public debt has a positive influence on inflation but this effects turns up one year after the initial shock. Collectively, these results suggest that constraining reserve money may not yield the desired level of inflation beyond one year after a policy response to inflation because of the delayed positive effect of a rise in public debt on inflation. Further, although interest rates and reserve money are found to have similar effects on public debt, the effect of interest rates is found to be more significant. The finding that monetary policy can affect inflation through fiscal variables suggests that fixing short term interest rates may be more effective in managing inflation in Kenya over a medium-term horizon. It would, therefore, be more appropriate for Kenya to choose a monetary policy framework where the interest rate is a policy instrument. 12

References Afonso, António. (2002), Disturbing the Fiscal Theory of the Price Level: Can it fit the EU-15?, Department of Economics, ISEG-UTL (Lisbon), Working Paper 1/2002/DE/CISEP. Baldini, Alfredo and Marcos Poplawski Ribeiro (2008), Fiscal and Monetary Anchors for Price Stability: Evidence from Sub-Saharan Africa, International Monetary Fund, Working Paper 08/121. Bauducco Sofia and Francesco Caprioli (2011), Optimal Fiscal Policy in a Small Open Economy with Limited Commitment, Central Bank of Chile, Working Paper No 644 (September). Canzoneri, Matthew B., Robert E. Cumby and Behzad T. Diba (2001), Is the Price Level Determined by the Needs of Fiscal Solvency?, American Economic Review, Vol. 91, No. 5 (December), pp. 1221-1238. Canzoneri, Matthew B., Robert E. Cumby, Behzad T. Diba, Gauti B. Eggertsson, Carlo A. Favero (2004), How Do Monetary and Fiscal Policy Interact in the European Monetary Union? NBER International Seminar on Macroeconomics, pp. 241-325. Cochrane, John H. (2001), Long-Term Debt and Optimal Policy in the Fiscal Theory of the Price Level, Econometrica, Vol. 69, No. 1 (January), pp. 69-116. Eusepi, Stefano and Bruce Preston (2011), Learning the Fiscal Theory of the Price Level: Some Consequences of Debt Management Policy, Federal Reserve Bank of New York Staff Report No. 515, September. Javid, Attiya Yasmin, Ummaima Arif and Abdul Sattar (2008), Testing the Fiscal Theory of Price Level in Case of Pakistan, Pakistan Development Review, 47: 4 Part II (Winter), pp. 763 778. Leeper, Eric M. and Todd B. Walker (2011), Perceptions and Misperceptions of Fiscal Inflation, Bank for International Settlements Working Paper No. 364, November. Sims, Christopher A. (1994), A Simple Model for Study of the Determination of the Price Level and the Interaction of Monetary and Fiscal Policy, EconomicTheory4, pp. 381 399. Sims, Christopher A. (2011), Stepping on a Rake: The Role of Fiscal Policy in the Inflation of the1970s, European Economic Review, 55 (1), pp. 48 56. Thugge, Kamau, Peter S. Heller and Jane Kiringai. (2011), Fiscal Policy in Kenya: Looking Toward the Medium-to Long-Term in Kenya: Policies for Prosperity, 83. 13