Treasury Auction Interest Rates and
Economic Performance in Turkey
Department of Economics
Phone: + 90 312 290 1643
Fax: + 90 312 266 5140
February 4, 2003
∗I would like to thank Kenneth Kuttner for providing the computer program for the Blanchard and Perotti
routine and Anita Akka¸ s, Michael Claxton, Patrick Conway, Tarık Kara, George Karras, Kamuran Malatyalı, Zafer
Mustafao˘ glu, Bilin Neyaptı, Michael Salemi, Faruk Sel¸ cuk and Hakan Hakkı Yılmaz for their helpful comments. I
am particularly grateful to Richard Froyen for his criticism and support throughout this study.
Treasury Auction Interest Rates and
Economic Performance in Turkey
This paper proposes the idea that innovations in treasury auction interest rates are informa-
tive for output and prices in Turkey. The reason for the focus on this relationship is that treasury
auction interest rates measure the state of fiscal policy better than taxes, government spending
or both. We show that a positive innovation in treasury auction interest rates behaves in the
same way as the expansionary fiscal policy does. Importantly, this paper also demonstrates that
inflation in Turkey is a fiscal rather than a monetary phenomenon.
Key words: Fiscal Policy, Interest Rates and Business Cycles.
JEL codes: E62, E32 and H30.
Does fiscal policy affect the economy? If so, how can we measure this effect when government
spending, taxes and deficits do not represent true government purchases, receipts or deficits? This
study provides new empirical evidence on both of these questions. While the second question may
not be relevant for all economies, especially where government does not try to hide its spending or
avoid accountability, the first question is valid for all economies.
In order to account for the effects of fiscal policy on economic performance within a dynamic
framework, most studies use the U.S. data on government spending, taxes or deficit to measure
government fiscal stance (for example, Blanchard and Perotti, 2003; Burnside, Eichenbaum and
Fisher, 1999; Eichenbaum and Fisher, 1998; Fatas and Mihov, 2000; and Fatas and Mihov, 2001).
However, these variables do not measure the fiscal stance for all countries. Polackova (1998) argues
that governments’ hidden financial commitments and contingent liabilities – not given explicit
budget allocation nor even officially recognized – constitute a major cause of concern for fiscal and
macroeconomic instability in a considerable number of countries. Alesina, Hausmann, Hommes and
Stein (1999) argue that a less transparent fiscal system tends to produce more surprise liabilities
which destabilize these economies. Furthermore, Easterly (1999) notes that governments seem to
have an incentive to hide liabilities in off-budget items. He argues that even some of the EURO
countries like Austria, France, Germany and Italy adopted economic policies, one way or another,
to hide their liabilities in order to meet the Maastricht criteria. Even if this concern exists, to
the best of our knowledge there is no study to assess effects of governments’ total fiscal stance on
economic performance. The purpose of this paper is to use an alternative measure of the fiscal
stance and assess the effects of fiscal policy on economic performance, when the budget does not
show all the government’s spending and revenues.
Turkey provides an unique laboratory environment to assess the effects of economic policies on
its economic performance. First, Turkey has had a high and volatile inflation rate since the mid-
1970s without running into hyperinflation due to low price indexation. High inflation magnifies
relationships among various macroeconomic variables. Moreover, the low level of indexation allows
us to capture the effect of economic policy on prices and output relatively easily compared to
other developing countries where regulations like price freezes or indexation have been heavily
used. Second, Turkey has highly liberalized and efficient financial markets. Treasury bill and bond
markets are deep and efficient. These make interest rates for Treasury paper representative of public
perceptions of future economic developments and a good measure of illiquidity in the markets.
As is the case in many countries, government spending/revenue figures do not actually represent
the stance of the fiscal policy in Turkey.In order to assess the fiscal policy, this paper uses
the innovations in the spread between treasury auction interest rates and interbank interest rates
calculated from a Vector Autoregressive (VAR) specification as a proxy for the stance of the fiscal
policy. This paper is organized as follows: Section two elaborates the available fiscal stance measures
and specifies the VAR model that is used in this paper. Section three introduces the data used.
Section four presents empirical evidence that the spread between the treasury auction interest
rate and the Central Bank’s interbank rate provide information about the future movements of
macroeconomic variables by considering two different criteria. Section five presents the empirical
evidence on the dynamic response of the economy to the innovation in auction interest rates.
Caveats are discussed in section six and the last section offers some conclusions.
2 Fiscal Policy Measures
The government spending figure from the Turkish Consolidated Budget is not a good measure of the
stance of fiscal policy. The consolidated budget does not include losses of state-owned enterprises,
subsidies to social security system, duty losses of the publicly-owned banks, contribution to rolling-
fund institutions, budgets of local governments and default payments on guaranteed investment
project credits by the Treasury (see, Report Drawn up by Special ad hoc Committee on Restructuring
of Public Fiscal Management and Fiscal Transparency for a detailed review). The consolidated
budget constitutes around 60-80% of the public sector borrowing requirement (PSBR) and this
ratio changes from year to year. Therefore, a stable relationship between the figures from the
consolidated budget and PSBR does not exist. Moreover, PSBR figures are only available on an
annual basis. Even if these figures were available in a higher frequency, transferring funds from the
Treasury to finance state-owned banks (as in 1998 and 2001) would introduce large changes in the
PSBR even if these losses were realized in previous years. Hence, an alternative measure of the
stance of fiscal policy is in order.
If we want to measure the stance of the fiscal policy, there are two methods that could be
followed. First, one may define a new series by going backwards and incorporating excluded items.
Even if defining a new variable is quite costly, this still would not guarantee that the newly created
variable would continue to measure the government’s fiscal stance in the future.Second, one
may use a proxy of its spending or deficit. If we can find a variable whose innovation could be
interpreted as “policy shock” where the systematic part of the variable could be captured by lagged
economic variables, and if we can assume that these measurable policy shocks are independent of
contemporaneous economic disturbances, then the reduced-form responses of the specification can
be used to observe the policy shocks and these measures assess the fiscal aggregate effect of a policy
Let’s assume that economy can be represented by the following specification
Yt= B0Yt+ B1Yt−1+ C0Pt+ C1Pt−1+ ut
Pt= D0Yt+ D1Yt−1+ GPt−1+ vt
Where Ytvector includes a set of non-policy variables, Ptvector includes a set of policy variables,
ut and vt are orthogonalized residuals. Note that this system is not identified. If, on the other
hand, we assume that D0 = 0 – that is, there is no feedback from the sate of the economy to
the current policy setting contemporaneously – the specification can be converted into a standard
Vector Autoregressive (VAR, hereafter) specification by substituting equation 2 for Ptin equation
1 as can bee seen in equation 4 below.1
Pt= D1Yt−1+ GPt−1+ vt
Yt= (1 − B0)−1[(B1+ C0D1)Yt−1+ (C0G + C1)Pt−1+ ut+ C0vt](4)
Here, we can identify the effect of policy innovation on the non-policy variables with the impulse
response functions of Y to past changes in v by using the VAR specification as stated in equations
3 and 4 where Ptprecedes Yt.
Regarding what should be included in Ptas a representative of the government’s fiscal stance,
up until now, changes in government spending, taxes or deficit have been taken as the indicator of
the fiscal policy. This paper proposes using the spread between the Treasury auction interest rates
and the overnight interbank interest rates as a measure of fiscal policy. The cost of government’s
borrowing relative to borrowing/lending rate of banks among themselves increases as government
1One could also assume C0 = 0 for the identification – that is, there is no feedback from the policy setting to the
state of the economy contemporaneously. This specification will also be considered later in the paper.
implements its expansionary fiscal policies regardless if these spending (and revenues) are recorded
on on-budget items or on off-budget items. The first reason for the Treasury’s increased borrow-
ing cost will be that less liquidity will be available in the markets as government borrows. The
second reason is that even if government spending is not recorded in the consolidated budget and
performed through forcing the state owned institutes to record looses (where these losses are ul-
timately financed by the Treasury), the public will foresee these future liabilities and will be less
willing to finance the government’s deficit. Thus the Treasury auction interest rates will increase
relative to central bank’s overnight interbank interest rates. Therefore, observing the spread be-
tween the Treasury’s auction interest rate and the interbank interest rate provides an additional
channel to assess the stance of the fiscal policy. In this paper, we propose the idea that the spread
between the treasury auction interest rate and the interbank rate can be used as an indicator of the
government’s fiscal stance. If this is the case, then the dynamic response of the economic variables
to the innovation in the treasury auction interest rate and interbank rate spread will measure the
effect of economic variables on fiscal policy. The empirical evidence presented here suggests that
expansionary fiscal policy is associated with higher income and prices. These are parallel to the
predictions of standard IS/LM models as well as to the provided empirical evidence on fiscal policy
(see for example, Blanchard and Perotti, 2003; Burnside et al., 1998; Edelberg, Eichenbaum and
Fisher, 1998; Fatas and Mihov, 2000 and Fatas and Mihov, 2001). However, the empirical evidence
presented here suggests that the domestic currency depreciates with an expansionary fiscal policy.
This is not what the Mundell-Fleming model suggests under perfect capital mobility; however, this
could be due to higher default risk premium with increased deficit (see: McMillin and Koray, 1990
and Koray and Chan, 19991). In addition, expansionary fiscal policy seems to be associated with
loose monetary policy. The state of the monetary policy could be explained by its low independence
or, as stated at various times by the governors, by the Central Bank of the Republic of Turkey’s
concern with the stability of the financial markets rather than with inflation.
The treasury auction interest rate is the weighted average of each auction’s interest rate for the
corresponding month.2As income measures, three variables are used: industrial production (ip),
2The Treasury opens auctions for various maturities each month. Here we disregarded these different maturities
for each auction when we calculated the treasury auction interest rate for each month . Therefore, the auction
interest rate that we used here is a mixture of several “forward rates”, which is implicit in the term structure of
interest rates with different maturities. In other words, the auction interest rate variable is a pooled time series of
capacity utilization rate of the private sector (cu), and the number of building permits given by
the local authorities (building). Prices (p) are the wholesale price indices. The exchange rate
(exchange) is the official exchange basket that the Central Bank of the Republic of Turkey has
been following in its operations: 1 USD + 1.5 DM. The depreciation rate is the logarithmic first
difference of the official exchange basket, and the interbank interest rate is the Central Bank’s
overnight interbank interest rate . Money (m) is M1 plus Repo volume. All data are available from
the data delivery system of the Central Bank of the Republic of Turkey (CBRT).3
In this paper, the spread between the Central Bank’s interbank interest rate and the depreciation
rate of the basket (Monetary Spread hereafter) is used as an indicator of the monetary policy where
the Monetary Spread is defined as Monetary Spreadt=
∗ 100, as suggested
by Berument (2001). The selection of the interbank interest rate and depreciation rate as a spread
measure is not arbitrary. These two variables, simultaneously or separately, are used by the Central
Bank of the Republic of Turkey to implement its economic policies. Central bankers of developing
and small open economies also: (1) may be concerned with the currency substitution problem,
and (2) may have an incentive to monitor their foreign exchange reserves in addition to ones that
concern by central bankers of developed economies. On currency substitution, the public may
have an incentive to avoid using domestic currency, preferring foreign currency to guard themselves
against inflation. Agents prefer to hold more of their wealth in foreign currency rather than in
domestic currency if the interest rate is lower or the depreciation of the domestic currency is
higher. On the level of foreign exchange reserves, the central bank has an incentive to monitor
its foreign exchange reserves, eliminating either the risk of speculative attack or the balance of
payment crisis. Reserves also increase as the domestic interest rate increases (due to either capital
inflow or the decreasing foreign exchange demand of domestic residents) and decrease as the return
on the foreign exchange increases. In order to account for these, Berument (2001) proposes a new
measure to monitor monetary policy. In particular, he argues that the extend to which interbank
interest rate exceeds the depreciation rate of a local currency (Monetary Spread) can be used as
an indicator of the stance of a central bank’s monetary policy. Using the Monetary Spread as an
forward rates with different maturities. Calvo and Guidotti (1992) and Missale and Blanchard (1994) argue that
there is a relationship between interest rate and maturities. Empirical evidence from Turkey (not reported here)
suggests that this relationship exists. Therefore, the “variable-maturity” auction interest rate variable that we used
is an monotonic transformation of the “constant-maturity” auction interest rates that one might use to measure the
stance of the fiscal policy, and thus, we can use the (variable-maturity) auction interest rates as an indicator of fiscal
policy as suggested by the envelope theorem.
indicator of a central bank’s monetary policy does not mean that the central bank controls both
of these instruments simultaneously, but rather the central bank may control one of the two and
merely watch the other. However, even in this case, the Monetary Spread might be used as an
indicator of monetary policy. This measure is also robust in the case of central bank’s switching
between pure exchange rate targeting and interest rate targeting. Here, the central bank may cut
the liquidity provided to the public by raising interest rates at a given level of depreciation, or
it may keep domestic interest rates stable and buy domestic currency from the public by selling
foreign currency at a lower rate. However, following Bernanke and Blinder (1992), we also used
the short term interest rates (Central Banks’s interbank interest rate) to identify the monetary
policy. We observed the price, exchange rate and liquidity puzzles Kim and Roubini (2000) discuss.
However, none of these puzzles was present when Monetary Spreadtwas used to identify monetary
Bernanke (1990) argues that interest rates as well as interest rate spreads can be used as
predictive variables for the economy. Stock and Watson (1989) showed that the predictive power
of the spread between the commercial paper rate and the T-bill rate and the spread between the 10
year T-bond rate and the 1-year T-bond out-performed the other variables considered. Bernanke
(1990) also argues: “The best single predictor among interest rate variables has been found to be
the spread between the commercial paper rate and the treasury bill rate.” He claims that the
reason for the high predictive power is that this spread captures the default risk perceived by the
market and assesses the likelihood of recession.
Similarly, we use the spread between the treasury auction interest rate and the overnight interest
rate as a measure of the ease with which the Treasury can borrow from the public. If the treasury
auction interest rate increases relative to the overnight interbank interest rate, we take this to
indicate that the public is more hesitant to finance the government’s deficit relative to how much
banks like to finance each other due to the high level of government debt. Or it could mean that
the public foresees the interbank rate as an indicator of inflation and the higher spread between
the auction and overnight rates may suggest a measure of higher real interest rates. Here, in order
to measure fiscal policy, the auction interest rate is measured relative to the previous month’s
One of the former vice-governors of the CBRT also argues that the public takes the interbank
4The reason that we deflated the auction interest rate with the last month’s rather than the current month’s
interbank rate was to avoid simultaneity that could raise the issue that the central bank may set the interbank rate.
interest rate as a benchmark when they make their offers to the Treasury for T-bills (or bonds).5
Hence, we took the spread between the treasury auction interest rate and the interbank rate as
an indicator of fiscal policy. We also try different spread definitions, such as spread with lagged
inflation and spread with lagged depreciations. The basic results of the paper were robust.
The estimates use data from 1986:10 to 2000:10. The selection of the beginning of the sample
is dictated by the availability of the data and we end the sample in 2000:10 to avoid the series of
financial crises and stress periods that started on November 19, 2000 and continued on February
22, 2001, July 7, 2001 and September 11, 2001. All these variables enter into the VAR analysis
in logarithmic form except the Auction and the Monetary Spread. When the regression analysis
is performed, twelve monthly dummies are included to account for the seasonality, three dummies
for the 1994 financial crises (1994:03, 1994:04 and 1994:05), one dummy for the month that the
Treasury did not go to auction (1999:12), one for the period for which repo data is not available
(prior to 1995:11), and last one when the new money definition is introduced (after 1990:01).
4 The Treasury Auction Interest Rate and Other Fiscal Measures
In this section, following the method of Bernanke and Blinder (1992), we present two measures
of the information content of the treasury auction interest rate compared to more standard fiscal
measures: taxes and government spending. In the first sub-section, the predictive power of Auction
interest rates will be compared with the other fiscal variables. The second sub-section deals with
the relationship between the innovations in Auction and the general perception of the state of the
4.1The Information Content of The Treasury Auction Interest Rate
In this sub-section we assess the predictive power of the Auction interest rate. In particular, we
perform a battery of Granger causality tests as reported in Table 1. In each row, we use three
measures of real economic activity and report the marginal significance level of the logarithm of the
lagged value of tax revenue, government spending, Auction, income, prices, exchange, Monetary
spread and Money with one lagged order6.7Here the predictive power of Auction is higher than
5Kumcu (June 8, 2001).
6The lag order is determined to be one by the Bayesian Information Criteria.
7When the VAR equation is specified, variables appear in several forms in the specification: (i) exchange rate
appears twice as the exchange rate itself and in Monetary Spread where the Monetary Spread is the difference
between the interbank interest rate and the logarithmic first difference of the exchange rate and (ii) auction is
the predictive power of government spending and tax revenue for all three measures of income (the
industrial production, the capacity utilization rate of the private sector, and the number of building
permits given by the local authorities). Panel B reports the marginal significance level for prices.
The predictive power of Auction is higher when industrial production is taken as a measure of
income, but not for building or capacity utilization. This gives a strong indication that Auction
has predictive power on income and prices.8
Table 1: Marginal Significance Levels on the Behavior of Income and Prices.
Panel A: On Income
Panel B: On Prices
The Granger Causality test has drawbacks because the right hand side variables are not or-
thogonalized. In order to account for that, we used variance decompositions of forecast variables.
This method also has its drawbacks in that the results of the estimates depend on the ordering of
the explanatory variables, the forecast horizon and low level of statistical significance (see: Runkle,
1987). Therefore, Table 2 reports the variance decomposition with 6, 12, 18, 24 and 36 month hori-
zons by using three income measures in panels A, B and C. The variables are ordered as logarithm
of tax revenue, logarithm of government spending, Auction, Income, prices, exchange, Monetray
Spread and Money. We also tried an alternative ordering method but the results were robust to
this change. For almost all income measures and time horizons, Auction explains the variability
the difference between the treasury auction interest rate and the lag value of the interbank interest rate, which is
also present in the lagged values of Monetary Spread. These could be considered as problems. Here we impose
these differences as constraints in order to identify the monetary and fiscal policies and we treat these variables as
separate variables. Entering different rates along with their components and their differences along with their levels
is common in the literature (see for example Bernanke, 1983; Bernanke, 1990; Bernanke and Blinder, 1992; Friedman
and Kuttner, 1992 and Strongin, 1995).
8The levels of government spending and revenue may affect the economy separately.Therefore, we entered
government spending and tax revenue figures separately. However, we also try various measures of deficit. We
could not take the logarithm of the deficit (since it could were negative values), so we used other measures like
government spending/tax revenue, government spending-tax revenue over interpolated monthly GDP figure and real
deficit. Under all these alternative definitions, the predictive power of Auction was above the deficit measures.
Table 2: Variance Decomposition of Forecasted Income∗
RevenueSpendingAuction IncomePrices Exchange M. SpreadMoney
Panel A: Industrial Production as an Income measure
(2.58) (2.16) (4.29)
4.88 4.11 9.81
(2.50) (2.10) (4.29)
Panel B: Capacity Utilization as an Income measure
Panel C: Building as an Income measure
∗ The numbers in parentheses are standard error values.
of income better than revenue or spending. Table 3 repeats the same analysis for prices. A higher
percentage of the variability of prices is explained by Auction than by government spending and
taxes.9In sum, Auction has higher predictive power for both income and prices than conventional
4.2 Public Perception and Fiscal Policy
In order to assess whether Auction could be used as a measure of fiscal policy. We try to assess
whether innovations in Auction capture the perception of public on the behavior of the fiscal
policy. Figure 1 plots the cumulative sum of Auction innovation: vt. The downward movement of
9Blanchard and Perotti (2003) assess the effect of shocks in government spending and taxes by using a structural
VAR approach that accounts for the institutional structure of the tax and transfer systems. The timing of tax
collection and spending activities are used to identify the automatic response of taxes and spending activities, and
by implication, to infer deviations from the automatic response. We also used their methodology but the basic result
in this sub-section was unchanged.
Table 3: Variance Decomposition of Forecasted Prices∗
Panel A: Industrial Production as an Income measure
Revenue SpendingAuction Income PricesExchange M. Spread Money
Panel B: Capacity Utilization as an Income measure
Panel C: Building as an Income measure
∗ The numbers in parentheses are standard error values.
the variable indicates the fiscal tightness and upward movements are indication of fiscal looseness.
Figure 1 suggests that periods between 1990:01 and 1993:04, between 1994:06 and 1996:06, and
between 1999:06 and 2000:10 have tight fiscal policy.
[Insert Figure 1 here]
The first fiscal tightness is observed in 1990:01, when for the first time in its history, on January
16, 1990, the Central Bank of the Republic of Turkey announced its monetary program for the
year. Since the government knew about it in advance, the announcement of the monetary policy
can be assumed to be a part of a substantial stabilization program. Here we might be capturing
this announced fiscal tightness. To be specific, the program was intended to limit Central Bank
credit to the public sector and the real credit extended to the public and private sectors together
declined by the end of year.
A loose state of fiscal policy can be observed after 1993:05. This date is associated with the
death of president Turgut¨Ozal on April 17, 1993 and the election of Prime Minister S¨ uleyman
Demirel by the Parliament as the president. After the self-inflicted crisis on April 4, 1994, it seems
to have taken two months to tighten fiscal policy. In 1994:06, Turkey signed a stand-by-agreement
with the IMF. The ensuing tight fiscal policy is suggested by Figure 1. With the establishment
of the 54thgovernment by the populist leader Necmettin Erbakan in April 1996, an expansionary
fiscal policy is observed until 1999:06, when the government was ready to sign another stand-by
agreement with the IMF. It is interesting to note that this tight fiscal policy continued to be
implemented even after the August 17, 1999 earthquake in which around 18,000 lives were lost.10
5 Empirical Evidence
Before starting to discuss what the empirical evidence reveals regarding how the expansionary fiscal
policy affects economic performance, what the economic theory suggests should be discussed. Next,
the empirical evidence will be discussed. The third sub-section is on robustness statistics and the
last sub-section provides empirical evidence on variance decomposition of the forecasted variables.
5.1 General Description of the Likely Effects of a Fiscal Expansion
Ricardian Equivalence suggests that as long as taxes are lump-sum, agents have infinite time
horizons, future tax burdens are certain, capital markets perfect and expectations are rational,
then there is no difference between deficits and taxes to finance spending and there is no effect on
consumption, hence on income.
Even if inflation is a monetary phenomenon, governments print money to finance deficits. Hence,
fiscal expansion should be inflationary (see: Fisher and Easterly, 1990). If the interest rate is
flexible, increasing the government deficit will increase interest rates, while decreasing investment
and consumption. Unless crowding out is complete, output increases. Therefore, economic theory
suggests that expansionary fiscal policy is associated with non-decreasing output, non-decreasing
prices and loose monetary policy. However, the effect of fiscal policy on money is uncertain. Even if
the money supply increases, money demand decreases due to lower consumption and higher interest
rates. Hence, the effect on money is not well defined.
10It is often argued that the government used the earthquake as an excuse to tighten fiscal policy. The government
raised 3,743 trillion TL as additional revenue to help earthquake victims but spent only 2,098 trillion TL for this
purpose within two years of the earthquake (Dunya, 2001).
Furthermore, the Mundell-Fleming model suggests that expansionary fiscal policy increases
interest rates and this appreciates the domestic currency due to an (un)covered interest rate parity
condition. However, a higher deficit means higher interest rates and this may mean a higher default
risk, and this may depreciate the domestic currency (see, McMillin and Koray, 1990; Koray and
Chan, 1991 and World Economic Outlook, 1995).
5.2Implication of the Specification
In this sub-section, the economic effect of a change in fiscal policy will be studied. Figure 2 displays
the estimated impulse responses to an expansionary fiscal policy, depicted as a positive innovation
in Auction, for the six macroeconomic variables considered. The impulse responses are reported
for a 12 month horizon. In order to calculate the confidence bands, the Bootstrap method of 500
draws is used. The middle line shows the median of the draws and the other two lines show the
confidence interval at the 80% confidence bands.11
[Insert Figure 2 here]
We will interpret the impulse responses of our baseline model where the order and the list of the
variables are Auction, Income, prices, exchange, Monetary Spread and Money. First, the shock to
fiscal policy persist for six months. The persistence of fiscal policy is something expected. Output
tends to increase for four months and then return to its initial level. The increase in output leads
to a decline in the deficit through higher tax revenue. Parallel to that, the Auction variable tends
to decrease. The evidence on the behavior of output and Auction are parallel to the findings of
Eichenbaum and Fisher (1998), Edelberg, Eichenbaum and Fisher (1998) Fatas and Mihov (2001).
Increase in Auctiontincreases prices permanently in a statistically significant fashion. This result
is also robust with alternative specifications.
Next, we look at the behavior of monetary policy as measured with Monetray Spreadt. It seems
that the Central Bank adopts a loose monetary policy following a loose fiscal policy. This finding
is parallel to Fisher and Easterly (1990), the statements of Central Bank Governors in the past
(see, Berument, 2001) and also consistent with low degree of independence of the bank from the
government (see, Berument and Neyapti, 1999). However, this is not statistically significant. The
money aggregate falls with expansionary fiscal policy. This might be due to higher interest rates
11Calculated impulse responses and their confidence bands for all six variables are available from the author upon
or lower consumption and investment if the expansionary policy results from higher government
Concerning the behavior of the exchange rate, the effect of expansionary fiscal policy depends
on the slope of the BP curve within the Mundell-Fleming model. If the slope of the BP curve
is higher than the slope of the LM curve, expansionary fiscal policy would increase output but
depreciate the local currency. On the other hand, if the slope of the BP curve is less than the slope
of the LM curve (as it would be under perfect capital mobility), then expansionary fiscal policy
appreciates the currency. Agenor, McDermott and Ucer (1997) show that expansionary fiscal policy
appreciates the temporary component of the real exchange rate for Turkey.12On the other hand,
Obstfeld (1991) argues that if the government cannot commit itself to a low fiscal deficit, it may
depreciate the local currency. Giorgianni (1997) also argues that in countries with large fiscal
deficits, expansionary fiscal policies increase the likelihood of debt consolidation and depreciate the
domestic currency through its risk premium. He provides empirical evidence for this with Italian
data. Moreover, Celasun, Denizer and He (1999) also note this depreciation effect for Turkey.13
One thing that needs to be highlighted here is that the effect of fiscal policy is observed faster in
Turkey compared to the U.S. An expansionary fiscal policy increases output almost instantly and
the output level returns to its initial level within six months. On the other hand, studies on the
U.S. (such as Burnside et al., 1999) find that the effect of fiscal policy is observed after 3 quarters
and lasts for 12 quarters. The relative size of the government to GDP is small in the US compare to
Turkey. Furthermore, fiscal policy affects the output through relative few transmission channels in
Turkey compare to the US. Therefore, observing the effect of fiscal policy during a short period of
time is a reasonable. Moreover, the effect of fiscal policy on the economy does not last long possibly
due to institutions and contract flexibility. It is common that wages are settled for 6 months. Lately,
wages of civil servants are adjusted monthly with the realized inflation. This makes the market
clear faster and the effect of policy does not last long. Lastly, Agenor and Montiel (1999) discuss
two reasons for the inertia. One is that there is implicit or explicit backward looking indexation in
nominal variables and the other is the lack of credibility in any program the government adopts.
Turkey does not have either of these.
12We could not replicate their findings on exchange rate possibly because we use different specifications than theirs
and the real exchange rate variable that they used is calculated by IMF is not available publicly and we could not
use that variable.
13One may look at World Economic Outlook (1995: 79-81) for the mixed results of the effect of fiscal policy on
One item that affects Auction interest rates in Turkey is the foreign capital inflow. Although
Celasun et al. (1999) find that the capital inflow is associated with neither public consumption
nor public investment, both Celasun et al. (1999) and Alper (2002) argue that the capital inflow
is associated with the performance of the private sector (private consumption, private investment
and total output). Hence, the capital inflow setting is entered into the VAR as the first variable.14
Figure 3 suggests that the results on the effect of Auction innovation on all six variables that are
used in the benchmark specification are parallel to Figure 2. However, capital inflow tends to
decrease for 2 months. This is on a parallel with the increased risk premium hypothesis discussed
in the previous section. After 1997, there has been a decrease in capital inflow (see: Celasun et
al., 1999), which accounts for why the sample ends in 1997:12. The results were parallel to the
benchmark specification again.15
Next, lag length was increased up to 6 lags. The level of significance decreases in the impulse
response analysis for the effect of Auction on other variables, but the results were similar.
It is possible that there is a structural change in the model.Celasun et al. (1999) argue
that after 1997 there was a change in the capital inflow and that borrowing by the Treasury was
more difficult than in the pre-1997 period. Therefore, the sample is ended in 1998:08 with the
Russian financial crisis and the analysis is performed again. Turkey also experienced a self-inflicted
financial crisis in April of 1994. As a second stability test we start to sample from 1995:01. When
the impulse responses are calculated for the post- 1995 period, the results are parallel to the
benchmark specification for the behavior of output, prices, the exchange rate and money. However,
when we start to sample after 1994 we observe that expansionary fiscal policy is associated with
tight monetary policy. When the sample includes pre-1994 and excludes post-1997, we find that
expansionary fiscal policy is associated with loose monetary policy.
Different definitions of Auction are also used to measure the fiscal policy. The auction interest
rate is first denominated with lagged inflation:
∗ 100, then denominated with
lagged value of depreciation:
∗100. Results using each of these measures are
parallel to the benchmark specification; but when the auction interest rate is denominated with
depreciation, there is an increase in money variable rather than a decrease when we introduce one
standard deviation shock to Auctiont. This may mean that when the auction interest rate relative
to the foreign exchange rate increases, agents switch from foreign currency to local currency, and
14For normalization, the capital inflow figure is divided by the lag value of imports.
15Other impulse responses are available from the author upon request.
this increases the money variable.
One of the crucial items in the determination of the auction interest rate is the deepness of the
financial market. If financial markets are not deep, then expansionary fiscal policy would increase
the interest rate more than if the financial markets are deep. In order to capture changing financial
deepness, we include the time trend, the logarithm of time trend, the square of the time trend and
the lag value of financial deepness measured as a Real M2Y-Industrial product ratio. When all
these four variables are included individually in the VAR specification as additional regressors, the
basic inferences of the benchmark model are found to be robust.
Agents may like to see the response of the monetary authority to the state of the economy before
they make their offer to the Treasury on an auction. In order to incorporate this, Auction was
entered after Monetray Spread. Impulse responses are also parallel to our benchmark specification.
Observing similar results with two different orderings further strengthen the conclusion of the paper.
The effect of fiscal policy on income can also be estimated by using other income measures.
Here we used the logarithm of the number of building permits given by the local authorities and
the private sector capacity utilization rate. The impulse responses gathered from these are also
parallel to the benchmark specifications.
5.4 Evidence from Variance Decompositions of Forecasted Variables
Impulse response functions reported here assess the dynamic effect of fiscal shocks. In order to
understand how fiscal policy shocks contribute to the volatility of various economic agents, it is
necessary to use forecast error variance decomposition analysis. There are two reasons why this
question is important. First, this question helps to assess whether fiscal policy shocks have been an
important independent source of impulses to the business cycles. Second, this aids the identification
strategy, which assumes that fiscal policy is mostly exogenous.
[Insert Table 4 here]
Table 4 reports the percentage of variation in the first 6, 12, 18, 24 and 36 step-ahead forecast
error variance for all six variables in each panel where the industrial production is used as a income
measure. The first thing that strikes attention is that the biggest variation of Auction is explained
by Auction. This supports the proposition that, as identified, Auction is an exogenous variable
and this supports the identification strategy adopted in this paper.
Second, at least 10% of the variation in output is explained by Auction and this is statistically
significant at the 5% level. At most, 0.32% of the variation is explained by monetary policy as
identified with Monetary Spread, this suggests that the Auction variable is not something that
policy makers can ignore in Turkey. The low explanatory power of the variation of output by the
variation of monetary policy is parallel to Berument (2001), Kim (1999) and Kim and Roubini
(2000). Except for Auction and income itself, none of the other four variables can explain the
variation of income in a statistically significant fashion. The highest percentage of the variation of
prices is explained by price itself. The variation in the exchange rate also explains the variation
in prices. Importantly Auction explains the variation in prices after step 12. This supports the
proposition that inflation is a fiscal phenomena in Turkey.
Using innovation in the treasury auction interest rate as a measure of fiscal policy does present
some problems. First, our model implicitly assumes that, as long as they are equal to each other,
increasing government spending or decreasing government receipts have the same effect on the
economy. The Keynesian IS/LM framework claims otherwise with the implied positive balanced
budget multiplier. Moreover, decreasing taxes may have an additional expansionary effect through
the supply side. Lastly, not only the level, but also the composition of the government spending
may have an effect on the economy (see: Ramey and Shapiro, 1998).
This paper argues that the spread between Treasury auction interest rates and Central Bank
interbank rates (Auction) is an indicator of fiscal policy. In order to demonstrate this, we showed
that the auction interest rate can be used as an information variable for output and prices, and
found that the Auction spread has higher information content for the economy than do government
spending and taxes. This might be due to the fact that variables from the consolidated budget
are not good indicators of the government’s fiscal stance in Turkey. These variables may not
measure the fiscal stance as well as the auction interest rate. Finally, this paper shows that when
expansionary fiscal policy is identified with positive innovations in auction interest rates, output
and prices increase and local currency depreciates.
Table 4: Variance Decomposition of Forecast Variables
The numbers in parentheses are standard errors.
Agenor, Pierre-Richard and Peter Montiel, Development Macroeconomics, Second Edition,
Princeton, NJ: Princeton University Press, 1999.
, C. John McDermott, and Murat Ucer, “Fiscal Imbalances, Capital Inflows, and the
Real Exchange Rate: The Case of Turkey,” European Economic Review, 1997, 41, 819–25.
Alesina, Alberto, Ricardo Hausmann, Rudolf Hommes, and Ernesto Stein, “Budget
Institutions and Fiscal Performance in Latin America,” Journal of Development Economics,
1999, 59, 253–273.
Alper, Emre, “Business Cyclles, Excess Volatility and Capital Flows: Evidence from Mexico and
Turkey,” Emerging Market Finance and Trade, 2002, 38(4), 25–58.
Bernanke, Ben, “Nonmonetary Effects of the Financial Crisis in the Propagation of the Great
Depression,” American Economic Review, 1983, 73(3), 257–276.
, “On the Predictive Power of Interest Rates and Interest Rate Spreads,” New England Eco-
nomic Review, Federal Reserve Bank of Boston, 1990, pp. 51–68.
and Alan Blinder, “Federal Funds Rate and the Channels of Monetary Transmission,”
American Economic Review, 1992, 82(4), 901–921.
Berument, Hakan, “Measuring Monetary Policy for a Small Open Economy: Turkey,” paper
presented at the 76th Annual Western Economic Association International Conference 2001.
and Bilin Neyapti, “How Independent is the Central Bank of the Republic of Turkey (in
Turkish),” Iktisat, Isletme ve Finans, 1999, pp. 11–17.
Blanchard, Olivier and Roberto Perotti, “An Empirical Investigation of the Dynamic Effects
of Shocks to Government Spending and Taxes on Output,” Quarterly Journal of Economics,
Burnside, Craig, Martin Eichenbaum, and Jonas D.M. Fisher, “Assessing the Effect of
Fiscal Shock,” mimeo 1999.
Calvo, Guillermo A. and Pablo E. Guidotti, “Optimal Maturity of Nominal Government
Debt: An Infinite-Horizon Model,” International Economic Review, 1992, 33(4), 895–919.
Celasun, Oya, Cevdet Denizer, and Dong He, “Capital Flows, Macroeconomic Management,
and the Financial System: The Turkish Case, 1989-97,” International Monetary Fund Working
Dunya, “One Quadrillion Tax Collected for Earthquake is Unaccounted for (in Turkish),” August
17, 2001, 2001, p. 4.
Easterly, William, “When is Fisal Adjustment an Illusion,” Economic Policy, 1999, 14, 55–86.
Edelberg, Wendy, Martin Eichenbaum, and Jonas D.M. Fisher, “Understanding the Effect
of a Shock to Government Purchases,” NBER Working Paper, 1998, 6737.
Eichenbaum, Martin and Jonas D.M. Fisher, “How does an Increase in Government Pur-
chases Affect the Economy?,” Economic Perspective, 1998, Federal Rezerve Bank of Chicago
Economic Perspectives; 22(3), 3rd Quarter, 29–43.
Fatas, Antonio and Ilian Mihov, “Fiscal Policy and Business Cycles: An Empirical Investiga-
tion,” mimeo 2000.
Evidence,” mimeo 2001.
, “The Effect of Fiscal Policy on Consumption and Employment: Theory and
Fisher, Stanley and William Easterly, “The Economics of Government Budget Constraint,”
World Bank Research Observer, 1990, 5, 127–42.
Friedman, Benjamin M. and Kenneth N. Kuttner, “Money, Income, Prices and Interest
Rates,” American Economic Review, 1992, 82(3), 472–492.
Giorgianni, Lorenzo, “Foreign Exchange Rate Risk Premium: Does Fiscal Policy Matter? Evi-
dence from Italian Data,” Internatioinal Monetary Fund Working Paper 97-39, 1997.
Kim, Soyoung, “Do Monetary Policy Shock Matter in the G-7 Countries? Using Common Iden-
tifying Assumptions About Monetary Policy Across Countries,” Journal of International Eco-
nomics, 1999, 48, 387–412.
and Nouriel Roubini, “Exchange Rate Anomalies in Industrial Countries: A Solution with
a Structural VAR Approach,” Journal of Monetary Economics, 2000, 45, 561–86.
Koray, Faik and Pingfai Chan, “Government Spending and the Exchange Rate,” Applied Eco-
nomics, 1991, 23, 1551–58.
Kumcu, Ercan, “Risk in the Exchange Rate-Interest Rate Tradeoff (in Turksih),” Hurriyet, June
McMillin, W. Douglas and Faik Koray, “Does Government Debt Affect the Exchange Rate?
Empirical Analysis of the U.S.-Canadian Exchange Rate,” Journal of Economics and Business,
1990, 42, 279–88.
Missale, Alessandro and Oliver Jean Blanchard, “The Debt Burden and Debt Maturity,”
American Economic Review, 1994, 84(1), 309–319.
Obstfeld, Maurice, “A Model of Currency Depreciation and the Debt-Inflation Spiral,” Journal
of Economic Dynamics and Control, 1991, 15, 151–177.
Polackova, Hana, “Government Contingent Liabilities: A Hiddden Risk to Fiscal Stability,”
World Bank Policy Research Working Papers 1998.
Ramey, Valerie A. and Matthew D. Shapiro, “Costly Capital Reallocation and the Effects
of Government Spending,” Carnegie Rochester Conference Series on Public Policy, 1998.
Report Drawn up by Special ad hoc Committee on Restructuring of Public Fiscal
Management and Fiscal Transparency, State Planning Organization, Ankara, Turkey:
State Planning Organization, 2000.
Runkle, David E., “Vector Autoregressions and Reality,” Journal of Business and Economic
Statistics, 1987, 5, 437–42.
Stock, James and Mark W. Watson, “New Indexes of Coincident and Leading Indicators,”
in O. Blanchard and S. Fischer, eds., NBER Macroeconomics Annual, Cambridge, MA: MIT
Press, 1989, pp. 351–93.
Strongin, Steven, “The Identification of Monetary Policy Disturbances Explaining the Liquidity
Puzzle,” Journal of Monetary Economics, 1995, 35, 463–497.
World Economic Outlook, International Monetary Fund, Washington, DC: International Mon-
etary Fund, 1995.
Supplement: Treasury Auction Interest Rate and
Economic Performance in Turkey
Department of Economics
Phone: + 90 312 290 1643
Fax: + 90 132 266 5140
This document provides additional figures to Treasury Auction Interest Rate and Economic
Performance for Turkey not reported in the original paper
Please Disregard this page Download full-text
Bernanke (1983), Bernanke (1990), Bernanke and Blinder (1992), Friedman and Kuttner (1992)
and Strongin (1995)
Blanchard and Perotti (2003), Burnside, Eichenbaum and Fisher (1999), Eichenbaum and Fisher
(1998), Fatas and Mihov (2001) and Fatas and Mihov (2000).
Burnside et al. (1999), Edelberg et al. (1998), Fatas and Mihov (2000), Fatas and Mihov (2001) ).
McMillin and Koray (1990) and Koray and Chan (1991) Runkle (1987) Fisher and Easterly (1990)
McMillin and Koray (1990), Koray and Chan (1991) and World Economic Outlook (1995)
Berument and Neyapti (1999) Ramey and Shapiro (1998)
Report Drawn up by Special ad hoc Committee on Restructuring of Public Fiscal Management
and Fiscal Transparency (2000) Dunya (2001)
Blanchard and Perotti (2003),