Interest rate corridor, liquidity management, and the overnight spread.
The aftermath of the recent global crisis is characterized by expansionary monetary policies of advanced economies, which have led to excessive liquidity in global markets and rapid changes in global risk perception. Beginning from the first half of 2009, monetary easing in advanced economies has led to a surge in capital inflows toward emerging economies including Turkey. An outcome of this development was the potential macroeconomic instabilities faced by emerging countries. A major concern has been that unfettered inflows can lead to problems in these countries such as exchange rate appreciation, excessive credit growth, and/or asset price bubbles. Hence, many emerging markets had to amend their monetary policy frameworks and take macroprudential measures in order to address the financial stability challenges posed by volatile capital flows. (1) This article focuses on the experience of Turkey, where the monetary policy framework has been extensively modified in the aftermath of the global financial crisis.
As of the second half of 2010, the Central Bank of the Republic of Turkey (CBRT) has changed the general framework of the inflation targeting regime and developed new policy instruments to support financial stability as a complementary objective to price stability. The new framework largely aims at containing the effects of capital flows on the domestic economy, especially on credit growth and current account deficit, without prejudice to the price stability objective.
In this respect, the CBRT increasingly emphasized credit growth and exchange rate as key channels in monetary policy transmission and stressed the need to have a separate control on these two variables, which in turn requires the use of multiple instruments. (2) A conventional inflation targeting framework with a single instrument might lead to policy trade-offs when large external shocks dominate. For example, raising the policy rate in order to prevent overheating and excessive credit growth during a capital inflow surge might pull even higher capital inflows, leading to overvaluation in domestic currency and a larger buildup of external imbalances. On the other hand, cutting the policy rate to prevent further capital inflows and currency appreciation might feed into a larger boom and have adverse consequences for inflation.
The new policy mix of the CBRT entailed the joint use of the interest rate corridor between overnight borrowing and lending rates, liquidity policy and required reserves in addition to the policy rate. (3) In this article, we focus on the interest rate corridor and liquidity management, and analyze the effect of these policies on the overnight market rate, which is represented by the Borsa Istanbul (BIST) overnight repo interest rate.
The corridor system is employed by many central banks around the world to set interest rates. (4) It refers to the window between a central bank's overnight borrowing and lending rates, where the borrowing rate acts as the floor and the lending rate acts as the ceiling for the values that overnight market rates can take. In a conventional corridor system, overnight market rate is generally kept close to the policy rate. After the global crisis, however, many central banks including the CBRT have abandoned this practice due to the introduction of unconventional policies. (5)
The CBRT has operated an interest rate corridor since 2002. However, the way the corridor system is operated changed considerably in 2010. Until the second half of 2010, the CBRT pursued an overfunding policy which led to a net liquidity surplus in the system. Hence, between early 2002 and mid-2010, the overnight borrowing rate of the CBRT was the effective policy rate with the BIST overnight rate forming very close to the floor of the corridor. In May 2010, the CBRT introduced a new liquidity instrument, 1-week fixed-rate repo auctions. Since then, the CBRT has provided funding mainly through this instrument. The CBRT may adjust the width of the interest rate corridor independently of the policy rate (1-week repo rate) and in an asymmetrical manner, that is, the policy rate does not need to be halfway in between overnight borrowing and lending rates.
In addition to daily and weekly liquidity instruments, the CBRT also used monthly repo auctions in this period. The presence of these liquidity instruments with different maturities allows the CBRT to change its funding terms on a daily basis by adjusting the amount of funding provided through each instrument. Thus, in the new monetary policy framework, the effective policy rate is represented by the weighted average of different rates at which the CBRT funds the system (named as "the CBRT average funding rate") rather than the 1-week repo rate. (6) The average cost of the CBRT funding provided to banks and the overnight market rate can be determined at different values within the corridor according to the intended monetary policy stance. With this framework, the CBRT aims to facilitate a fast and flexible reaction to the volatility in short-term capital movements together with controlling credit growth (CBRT 2012, 2013). (7)
Under this setup, the overnight market rate (represented by the BIST overnight repo interest rate) can exhibit larger fluctuations inside the corridor and can move far away from the CBRT average funding rate (Figure 1).
In fact, in the period following the adoption of the new framework, the spread between the overnight rate and the CBRT average funding rate (overnight spread) has become wider and more volatile compared to the conventional policy episode (Table 1, Figure 2).
[FIGURE 1 OMITTED]
In this article, we analyze the determinants of the overnight spread providing evidence from both the conventional and the new monetary policy episodes in Turkey. We focus on the interest rate corridor and liquidity management policies, and try to find out to what extent the nonstandard monetary policy measures affect the overnight spread. This analysis is important to shed light on the changing nature of the monetary transmission after the adoption of the new policy framework in 2010.
Our article is related to a burgeoning literature that aims to understand the determinants of the overnight spread. Linzert and Schmidt (2011) and Beirne (2012) analyze the determinants of the EONIA spread, that is, the spread between the Euro Overnight Index Average (EONIA) and the main policy rate of the European Central Bank (ECB). Linzert and Schmidt (2011) aim to explain the widening of the EONIA spread from mid-2004 to mid-2006 by linking it to the change in the operational framework of monetary policy at ECB in March 2004. They find an important role for the liquidity deficit as well as the tightness of liquidity conditions in accounting for the widening in the spread. Beirne (2012), on the other hand, examines the determinants of the EONIA spread during the crisis and non-crisis times, relating the spread to the liquidity policies of the ECB in addition to liquidity and credit risk. Similarly, Nautz and Offermanns (2007) examine how the EONIA rate (overnight market rate) adjusts to the policy rate of the ECB and they find a strong role for the change in the operational framework of ECB. Although the spread between policy rate and overnight market rate has risen because of the change in the implementation of the repo auctions, the study argues that this change did not lead to a loss of control over the EONIA. Recently, Bech and Monnet (2013) study the impact of unconventional policies on the overnight interbank market of six developed countries in the aftermath of the financial crisis. They emphasize the role of excess reserves in driving the overnight rate to the floor of the corridor. Overall, empirical studies provide mixed evidence on the determinants of the overnight spread. One common finding is that liquidity policy is an important factor in affecting the spread.
[FIGURE 2 OMITTED]
A related strand of the literature analyzes the effects of unconventional policy measures on various other short-term money market rates and spreads mainly using event-study methodologies. Examples include Krishnamurthy and Vissing-Jorgensen (2011) for the United States, Joyce et al. (2011) for the United Kingdom, and Brunetti et al. (2011) and Szczerbowicz (2011, 2012) for the euro area, who find mixed evidence on the effectiveness of unconventional policies in reducing money market spreads that widened following the global financial crisis.
To the best of our knowledge, this is the first article which investigates the determinants of the overnight spread in Turkey. Moreover, this is one of the few articles in the literature that analyze the effects of nontraditional policy instruments on the overnight spread. (8) Our article contributes to the understanding of the interest rate corridor and liquidity policy in an unconventional setup using the case of Turkey as an example. We use an interval censored regression framework to regress the overnight spread on a range of variables related to the liquidity policy of the CBRT and other policy instruments, the liquidity need of banks, interest rate expectations, and uncertainty. We find that the liquidity policy of the CBRT is an important determinant of the overnight spread in the unconventional monetary policy episode. Hence, we argue that the rise in the short-term money market spread in the new episode is a natural consequence of pursuing a monetary policy with multiple instruments.
[FIGURE 3 OMITTED]
The rest of the article is organized as follows. Section II explains the details of the liquidity management of the CBRT. The explanatory variables and the regression methodology are outlined in Section III. and estimation results are reported in Section IV. Finally, Section V concludes.
II. CBRT's LIQUIDITY MANAGEMENT FRAMEWORK
The daily liquidity management framework of the CBRT is depicted in Figure 3. The 1week repo rate lies between the overnight borrowing and lending rates of the CBRT. In addition to overnight and weekly funding, the CBRT also conducts open market operations (OMO) with 1-month maturity as well as primary dealer (PD) repo. (9)
Figure 4 presents the time frame for the operational steps in the CBRT's liquidity supply auctions and some stylized facts regarding the market structure. The timeline of the CBRT's standard operations and opening/closing times for various markets are also provided in Figure 4.
Each morning on a business day, the amount of liquidity provided to banks through weekly and monthly OMO is determined. The CBRT announces the allotment amount at 10:00 a.m. and carries out the auction afterwards. The CBRT conducts fixed-rate tenders with 1-week maturity on a daily basis (and with 1-month maturity on each Friday), and announces the auction results at around 11:15 a.m. Hence, by 11:15 a.m., financial institutions figure out how much funds they can obtain via the CBRT through 1-week repo (and also via 1-month repo (10)) auctions. Having received this information, the banks that could not cover their liquidity needs through these auctions have to rely on the overnight repo market with a higher cost, that is, BIST repo market and PD repo facility. Benchmark overnight market rates are set at 2.00 p.m. at the overnight repo/reverse repo market at BIST, which is the most active market for overnight funding across banks and other intermediaries. (11) Banks can also use standing facilities of the CBRT at overnight maturity at the upper and lower bounds of the prevailing interest rate corridor between 10:00 a.m. and 4:00 p.m. The Late Overnight Window (LON) facility by which the CBRT fulfills its lender-of-last-resort function operates between 4:00p.m. and 5:00p.m. at more discouraging interest rates for the banks. Finally, for the last day of the reserve maintenance period, in order to enhance the flexibility of the bank's liquidity management, the use of LON facility is extended until 5:15 p.m.
[FIGURE 4 OMITTED]
As part of the new monetary policy framework, the CBRT can induce additional monetary tightening when deemed necessary by lowering the amount of funding provided at the policy rate via 1-week quantity auctions. On these days, the CBRT may conduct 1-week repo auctions via traditional method. (12) Financial institutions cover their liquidity needs at a significantly higher cost through the BIST repo market, the CBRT overnight lending, or the PD repo (see Figure 5, for the time frame on days of additional monetary tightening).
During the days of additional monetary tightening, (13) the CBRT signals the market players about its tight stance by not announcing 1-week quantity auctions at 10:00 a.m. The CBRT may provide 1-week repo funding via the traditional method in which financial institutions bid competitively and intra-day auctions can be held anytime during the trading day. (14) Banks are forced to meet their funding requirements through other sources at higher interest rates within the corridor. PD banks can also obtain repo funding from the CBRT within their limits at a rate that is slightly lower than the upper bound of the corridor. Such different sources of funding provided by the CBRT increase the average cost of funding, and cause money market interest rates to approach close to the upper bound of the corridor. (15)
III. EXPLANATORY VARIABLES AND METHODOLOGY
A. The Determinants of the BIST Spread
Liquidity conditions in the money market determined by the interaction of the liquidity supply and demand factors stand as natural candidates to explain the behavior of the overnight spread. The explanatory variables given below are considered in two subgroups, namely: "liquidity supply and monetary policy" and "liquidity demand and other liquidity conditions." (16) The full list of the explanatory variables and the expected signs of their estimated coefficients are given in Table 2.
[FIGURE 5 OMITTED]
In addition to these explanatory variables, we also include the first and tenth lags of the spread. The tenth lag is used to capture potential seasonality since the reserve maintenance period in Turkey is ten business days.
Liquidity Supply and Monetary Policy. As described above, the CBRT provides liquidity through various instruments. The funds provided through overnight repo transactions constitute an important indicator, representing the overnight liquidity shortage, that is, the deficit that could not be covered in open market repo auctions at the beginning of the day. Hence, as a general indicator of daily aggregate liquidity conditions, we use net liquidity deficit calculated as the sum of CBRT's net overnight funding at BIST repo market and PD repo usage (Figure 6).
This indicator is directly related to the overnight spread. If the CBRT provides the exact amount of the daily liquidity needs of the banking system through 1-week and 1-month repo auctions, the realized net liquidity deficit is zero or negligible, and it can be stated that the CBRT implements a neutral liquidity policy. On the other hand, if the CBRT chooses to provide less liquidity than what is requested by the banks through 1-week and 1-month repo auctions, this situation leads to a positive net liquidity deficit and therefore should increase the BIST spread.
The net liquidity deficit in our case is akin to the liquidity policy variable used in Linzert and Schmidt (2011), which is defined as the difference between actual and benchmark allotments. Because the benchmark allotment is defined as the allotment that ensures a neutral liquidity situation during the period of the main refinancing operation, a higher (smaller) allotment than the benchmark creates relatively abundant (scarce) liquidity conditions and therefore should reduce (increase) the spread.
The existence of daily repo auctions necessitates the CBRT to closely monitor the liquidity needs of the banking system and offset any liquidity imbalances which may arise due to unexpected liquidity shocks. When making the decision of how much to provide in its standard daily funding operations (at 1-week and 1-month maturity), the CBRT takes into account both reserve requirements and forecasts of autonomous factors into account. Each morning on a business day, based on its daily liquidity forecasts, the CBRT decides the composition of its funding. The higher the degree of accuracy in daily liquidity forecasts, the smaller are the effects of other possible nonpolicy elements on the net liquidity deficit variable. In other words, accurate liquidity forecasts suggest that the "net liquidity deficit" can be treated as an exogenous variable. (17)
As banks have the flexibility to postpone their daily liquidity needs over the course of the 2-week maintenance period (except for the last day), one might expect a decline in bank's liquidity demand in response to a hike in the BIST O/N repo rate. However, data and anecdotal evidence suggest that banks in Turkey are reluctant to postpone their liquidity demand when faced with higher O/N interest rates within the maintenance period. That is why they have a highly predictable behavior, and the CBRT can forecast the daily liquidity deficit with high precision.
In fact, looking at the data, we observe a positive correlation between the BIST liquidity deficit and the BIST O/N rate. If the BIST liquidity deficit responded to the BIST O/N rate strongly, the relationship would be negative or ambiguous. In other words, if there was serious reverse contemporaneous feedback from the spread to the deficit, this would create a downward bias on the coefficient of the net liquidity deficit, making our estimates either negative or lower and less significant. However, despite this potential bias, our estimation results in Section IV show that the coefficient of the net liquidity deficit in explaining the BIST spread is large and significant.
In the empirical analysis, net liquidity deficit is represented as a ratio by dividing the deficit to the total Turkish lira (TL) reserve requirements of the banks in the current maintenance period. In other words, this variable measures banks' overnight liquidity deficits relative to their short-term liquidity needs. Higher net liquidity deficit is associated with tight liquidity conditions and is expected to drive the overnight spread upwards.
Another variable on the liquidity supply side is the CBRT weekly and longer-term funding, which is also represented as a ratio of the total TL reserve requirement liabilities of the banking system. This variable is expected to affect the spread negatively as higher weekly and longer-term CBRT funding would ease liquidity conditions. To control for the maturity effect, we construct the variable maturity structure of the CBRT funding. We calculate the ratio of 1-week repo in total funding (at 1-week and longer maturity) provided by the CBRT. An increase in this explanatory variable implies that a higher share of liquidity is provided at shorter maturity. This can be interpreted as tighter liquidity conditions since providing funds at longer maturity reduces uncertainties associated with banks' liquidity management. Hence, we expect to see an increase in the BIST spread whenever this variable goes up. In order to rule out concerns about endogeneity, we use the lag of this variable.
[FIGURE 6 OMITTED]
We also consider CBRT overnight borrowing and lending rates as determinants of the BIST spread. An increase in the overnight lending rate is predicted to create an upward pressure on the overnight rate and thus the spread. (18) When the banking system is in need of liquidity, the funding uncertainty gets higher which makes the upper bound an important cost indicator for the banking system. On the contrary, as the overnight borrowing rate is an important indicator for the short-term capital inflows, the overnight rate is expected to be negatively related to the borrowing rate. A raise in the overnight borrowing rate might bring in short-term capital inflows and ease liquidity conditions in the overnight money market, putting a downward pressure on the overnight rate. For example, on August 5, 2011, the CBRT raised the lower bound by 350 basis points to avoid a decrease in capital inflows due to deepening of the European debt crisis. When the borrowing rate increased, the overnight rate decreased sizably at the same time (see also Figure 1).
Liquidity Demand and Other Liquidity Conditions. Reserve requirements of banks constitute the main source of the liquidity demand of financial institutions. Liquidity demand is also related to expectations and uncertainty. When determining the quantity announced at each repo auction, the CBRT takes into account both reserve requirements and forecasts of autonomous factors such as changes in Treasury accounts and currency in circulation.
To measure banks' liquidity demand due to reserve requirements, we use cumulative average reserve fulfillments since the beginning of the maintenance period calculated as a ratio of the average reserve requirement. We take cumulated averages because the reserve averaging mechanism allows banks to smooth their reserve fulfillments within the maintenance period of 14 days. In the analysis, we include this variable as a ratio of the average reserve requirements to derive a relative figure. This variable is named as the cumulative reserve position. If reserve fulfillment of the banking system has been high on average at a specific day of the maintenance period, the pressure coming from reserve requirements would be low in the following day, driving the spread downward. Thus, we expect the lag of this variable to affect the BIST spread in the opposite direction.
The bidding behavior in open market repo auctions is another important factor in determining the liquidity situation in the system. Therefore, we introduce some variables that may influence the bidding behavior such as bid-to-cover ratio, liquidity uncertainty, interest rate expectations, and banks' heterogeneity in terms of liquidity position.
The bid-to-cover ratio is defined as the ratio of total bid volume to the amount covered by CBRT in the open market repo operations. This variable reflects the amount of liquidity demand by banks which has been met by the CBRT. We expect this variable to affect the BIST spread positively. A high bid-to-cover ratio shows that demand for central bank funding has only been satisfied to a low degree which leads banks to rely more on alternative funding opportunities at higher costs such as the interbank money market. This in turn would drive the spread upwards. To measure the effect of this variable correctly, we have to take into account changes in the way the repo auctions are conducted. Since May 20, 2010, 1-week repo auctions have been conducted via quantity auction method and each institution could bid up to the announced auction amount. In this set-up, all bids are allocated to the full amount if the total amount of bids is less than or equal to the announced auction amount. (19)
As of August 5, 2011, some amendments were made in order to enhance the efficiency of liquidity management and facilitate a more balanced distribution of central bank liquidity across the financial system (CBRT 2011). Accordingly, each institution's bid for the repo auction was limited by 20% of the announced auction amount. This upper limit was increased to 30% to be effective from January 2, 2013. Hence, we introduce interaction dummies for these amendments related to the open market repo auctions (auction amendments dummy).
Funding uncertainty is an important factor for banks' bidding strategies. Whenever daily funding need of banking system is high and there is uncertainty regarding to what extent this need would be covered by the CBRT or the market, banks may settle for higher funding costs. Moreover, increased aggregate liquidity uncertainty may lead banks to bid at higher rates due to the risk of not obtaining the desired liquidity. This would lead overnight rates to increase, and consequently drive the spread upwards. Therefore, we expect to find a positive relation between funding uncertainty and the BIST spread. CBRT announces aggregate required reserve balances and market opening reserves at 10:00 a.m. each morning. (20) Banks take this information into account when deciding how much to bid in OMO tenders because the difference between market opening reserves and required reserve balances at the beginning of the day is a proxy for intra-day cash flows. Therefore, the ratio of the expected intra-day cash flows to the amount of reserve balances at the beginning of the day contains information on funding uncertainty. (21)
Liquidity distribution among banks is another factor that possibly affects the bidding behavior of banks. In the BIST repo-reverse repo market, there are two submarkets: the interbank repo-reverse repo market (IRM) and the general repo-reverse repo market. The IRM was established to facilitate the repo transactions in organized market conditions, without having to meet the reserve requirement of the CBRT. (22) Different from the IRM, in the general repo-reverse repo market, investment funds can also operate in addition to banks. To represent the liquidity distribution among banks, we use the ratio of the volume of overnight repo transactions in IRM to the total volume of overnight transactions in the BIST repo-reserve repo markets. If this ratio is high, then the transaction volume in IRM is high which points to a heterogeneous distribution of funds among banks. In other words, this means that while some banks have liquidity shortage, some others have liquidity surplus. A bank faces higher uncertainty when it has to borrow from other banks rather than borrowing from the CBRT. For instance, a bank in need of funds cannot foresee how much a bank with a liquidity surplus is willing to provide and at what price. Banks with a liquidity surplus may demand high interest rates given the oligopolistic market structure. According to anecdotal information, it is known that whenever the transaction volume in IRM increases, banks have a tendency to bid in CBRT repo auctions although they are not in need of funds. Through this way, they try to make profits by lending to those banks short of liquidity. Thus, we expect that an increase in repo transactions in the IRM will lead the overnight interest rates and BIST spread to increase. In the analysis, we use the lag of this variable.
Interest rate expectations of banks also affect their bidding behavior. Similar to Linzert and Schmidt (2011), we use swap rates as an indicator for policy rate expectations. Accordingly, we use the difference between overnight swap rates and CBRT average funding rate to capture the expectations of the system regarding the changes in short-term interest rates. When banks expect an increase in interest rates within the maintenance period they would tend to bid at higher rates because refinancing would become relatively costly in the future. As a consequence, we expect a positive relationship between the swap spread and the overnight spread.
Dummy Variables. To capture some autonomous factors and stylized facts, we introduce several dummy variables in our regression model. Banks have a tendency to hold large amounts of reserves at the beginning of each reserve maintenance period. This may lead them to overbid in open market repo auctions and rely more on the BIST repo market driving the BIST spread upwards. Hence, we use a first day dummy to capture this effect. Similarly on the last day of a reserve maintenance period, banks may have a different bidding behavior. On the last day of a maintenance period, most of the banks have fulfilled the bulk of their reserve requirements and their demand for liquidity tends to be low. That's why we introduce a last day dummy to take this behavioral phenomenon among banks.
Relying on the observation that banks demand higher liquidity on tax payment days and quarterly balance sheet reporting periods, we also use dummy variables (balance sheet dummy and tax dummy) for these periods in our regression model. We predict these variables to put upward pressure on the BIST spread.
We laid out the potential determinants of the BIST spread in the preceding section. An important point to keep in mind when estimating a regression model for the BIST spread is the fact that the BIST overnight repo rate is bounded by the interest rate corridor. In other words, when the BIST overnight rate is nearby the lower or the upper bound of the corridor, even a sizable change in explanatory variables may not cause a visible change in the BIST overnight repo rate as it has no place to move further. One option is to eliminate the observations when the overnight repo rate hits the boundaries of the interest rate corridor. Exclusion of those observations, however, can discard potentially important information. In order to address the issue, we use a censored regression model. (23) We specify the regression model in terms of the latent level of the BIST spread, s*, that is, the level of the spread that would have been observed if the BIST overnight rate was not bounded by the interest rate corridor:
(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where the latent variable s* is linked to the observed spread, s, by the following condition:
(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
In Equation (2), [[D.sub.t].bar] ([bar.[D.sub.t]) is a binary indicator of censoring from the left (right). To be more precise, the binary indicator [[D.sub.t].bar] ([bar.[D.sub.t]) equals to one when the BIST overnight repo rate is sufficiently close to the lower (upper) bound of the corridor and zero otherwise. In other words, the BIST spread is assumed to be censored when [[D.sub.t].bar] = 1 or ([bar.[D.sub.t]) = 1 but fully observed otherwise.
Variables used in the specification of the latent process for the BIST spread--in the same order as in Equation (1)--are the CBRT's funding at weekly and longer maturity ([cbt.sup.funds.sub.t]), net liquidity deficit ([L.sub.t]), lagged value of the maturity of the CBRT funding ([maturity.sub.t-1]), lagged value of the cumulative reserve position of the banks ([RP.sub.t-1]), change in the bid-to-cover ratio ([DELTA]bid/cover), lagged value of liquidity heterogeneity among banks ([H.sub.t-1]), change in the upper bound of the corridor ([DELTA][lb.sub.t]), change in the lower bound of the corridor ([DELTA][lb.sub.t]), interest rate expectations ([i.sup.E.sub.t]), funding uncertainty ([uncertainty.sub.t]), dummy for the first day of the maintenance period ([D.sup.first day.sub.t]), dummy for the last day of the maintenance period ([D.sup.last day.sub.t]), dummy for the tax payment periods ([D.sup.last day.sub.t]), dummy for the balance sheet periods ([D.sup.balance sheet.sub.]) and an interaction dummy for the auction amendments ([D.sup.auctionamendtments.sub.t). Table 2 presents the explanation for each variable outlined above and the expected signs of their estimated coefficients in affecting the spread.
In econometric terms, Equations (1) and (2) constitute an interval censored regression, a in Equation (1) is a scale parameter that is identified and estimated alongside other parameters of the model. Errors are denoted by e, and are assumed to be normally distributed. The parameters of the model are estimated by maximizing the associated log likelihood function. (24)
IV. EMPIRICAL FINDINGS
Tables 3 and 4 represent the estimation results of the new framework period (May 2010-May 2013). We also present the results of estimated model for the conventional policy episode (November 2007-May 2010,25 Tables 5 and 6).
A. Unconventional Policy Episode
As Table 3 suggests, the first lag of the spread is found to be significant, while the tenth lag is insignificant. Explanatory variables that are directly or closely related to CBRT policies have statistically significant effects on the BIST spread.
The net liquidity deficit has the intended effect on the BIST spread. A higher net liquidity deficit leads the spread to increase as this puts upward pressure on the overnight market rates. Similarly, the CBRT weekly and longer-term funding has a significant impact on the spread. The spread tends to fall when the CBRT funding increases, which indicates an easing in the banks' liquidity conditions. The lagged value of the variable that captures the effect of maturity structure of central bank funding is also found to be significant in explaining the spread. This implies that as the average maturity of the CBRT funding falls, the overnight rate (and the spread) tends to increase.
As expected, a change in the upper bound of the corridor is found to affect the spread in the same direction. In fact, during the period of the analysis, the banking system generally has liquidity deficit. This situation leads the central bank to resort to the upper bound of the corridor as a policy tool more frequently compared to the lower bound. We also observe that the lower bound has affected the spread in the opposite direction as expected (despite with a lower coefficient in magnitude than the upper bound).
Variables that are related to the liquidity demand of the system are found to be statistically significant except for the first lag of average cumulative reserve position of banks. As our predictions suggest, the change in the bid-to-cover ratio has a significant positive effect on the spread. The estimated sign of funding uncertainty is also in line with our expectations, where a rise in uncertainty puts upward pressure on the spread.
The estimation results suggest that short-term interest rate expectations affect the spread in the same direction. The liquidity distribution among banks is found to impact the spread positively. The more heterogeneous the liquidity distribution among banks is, the more the spread tends to move upward.
In sum, all the explanatory variables that we consider, except the average cumulative position of banks, are statistically significant and have the expected sign. Estimating the cumulative reserve position of the system correctly is difficult for the banks because reserve requirement mechanism in Turkey complicates this calculation. (26) That is why banks may not be able to use this variable as an input in their bidding decisions which makes cumulative reserve position of banks insignificant in our analysis.
Finally, all dummy variables except the one for auction amendments are found to be significant in explaining the BIST spread. The bidding behavior of banks change in the first (last) day of the maintenance period and this puts upward (downward) pressure on the spread. Moreover dummy variable for balance sheet reporting period has a positive sign indicating that banks tend to increase their liquid assets around these days. As predicted, the tax-paying days are associated with a rise in the spread as banks' demand for liquidity increases. The dummy variable that captures the effect of the amendment to repo auctions in 2011 is found to be insignificant, which implies that this operational change did not alter the bidding behavior of banks.
Many studies in the literature document that the liquidity conditions are very important in determining the overnight spread. Thus, among other conditions, in order to see the relative strength of the liquidity policy, we carry out a standardized regression. When we standardize the variables, it is possible to compare the coefficients for each variable. The standardized estimation results for the new monetary policy episode are documented in Table 4. These results show that, among the policy variables, net liquidity deficit has the highest coefficient and thus it is the most important determinant of the overnight spread. Maturity structure is also crucial in explaining the deviations in the overnight spread. Among other factors, short-term interest rate expectations have the highest standardized coefficient. Bid/cover ratio and liquidity distribution also have sizable standardized coefficients.
B. Conventional Policy Episode
Most of the variables we use in the analysis above are associated with the new policy framework. Therefore we can only use a limited set of variables when running the regression for the conventional policy episode.
Table 5 presents the estimation results. The first lag of the spread is found to be significant while the tenth lag is insignificant similar to the results for the new policy period. Net liquidity deficit, the CBRT funding, bid-to-cover ratio, dummy variables for balance sheet reporting period, and the first day of a reserve maintenance period continue to have statistically significant effect on the spread in the conventional policy period. Remaining variables lose their significance in this period.
The standardized estimation results for the conventional policy episode are documented in Table 6. The net liquidity deficit has the highest standardized coefficient suggesting that it is the most important variable in explaining the deviations in the BIST spread in this period as well. Also, the CBRT weekly and longer-term funding is found to have an important explanatory power in this period.
Overnight market is the primary link in the interest rate channel of the monetary policy transmission. Since unconventional policies adopted by both developed and emerging market economies after the global crisis have complicated this transmission channel, many studies flourished recently trying to understand how this channel is modified. In this respect, our study tries to shed light on the changing nature of the monetary transmission in Turkey since the adoption of the new policy framework in 2010.
The CBRT has significantly changed its monetary policy strategy as of the second half of 2010 in order to address the financial stability challenges posed by volatile capital flows. Accordingly, giving more weight to the financial stability, the new monetary policy framework was tailored to meet the specific challenges of the new era. Thus, to offer a diverse, flexible and nonstandard policy approach, the CBRT has started using a policy mix of 1-week repo auction rates (policy rate), the interest rate corridor between overnight borrowing and lending rates, liquidity policy, and required reserve system.
Under the new monetary policy framework, the spread between the BIST overnight repo interest rate and the CBRT average funding rate (the overnight spread) has become wider and more volatile compared to the conventional policy episode. This study analyzes the determinants of the overnight spread providing evidence from both the conventional and the new monetary policy episodes in Turkey. Our results show that, among the policy variables, net liquidity deficit is the most important determinant of the overnight spread in the new monetary policy episode. We conclude that the widening in the spread is a natural consequence of a monetary policy framework with multiple instruments and objectives, whereby the overnight market rate can be affected by liquidity policies in addition to the policy rate.
LIQUIDITY SUPPLY AND DEMAND FACTORS
We present the basic supply and demand factors of liquidity by focusing on a simplified version of CBRT's balance sheet. Table A1 provides a stylized version of the CBRT's balance sheet, and it contains all relevant information to understand CBRT's current liquidity management without presenting any numerical figures.
In addition to CBRT's standard refinancing operations that are conducted as auctions and standing facilities such as overnight deposit/lending facilities, OMO of the CBRT include PD repo and BIST overnight repo/reverse repo operations which are both conducted on the initiative of counterparties. The PD banks can obtain overnight funding from the CBRT within their predetermined limits at a slightly lower cost than CBRT's overnight lending facility. The CBRT's transactions at BIST repo/reverse repo market acts as standing facilities (2.3.2a and 2.3.2b in Table Al) and assures that the repo interest rates in this market lie within the interest rate corridor set by the CBRT.
The autonomous factors are not related to central bank's transactions but they have liquidity-providing or liquidity-absorbing effects, therefore they have to be taken into account in central bank's liquidity management. (27)
Reserves held by banks consist of two parts: the first one is due to the reserve requirement system in which banks hold reserves over a 14-day maintenance period; the second one is the balances that are above the reserve requirements which banks may hold for precautionary motives. The sum of these two figures--reserves held by banks--can be treated as a residual position. Similarly, Bindseil and Seitz (2001) claim that reserve holdings of the banking system balance the central bank's balance sheet because all operations of a central bank ultimately affect the banks' reserve accounts. By using the simplified CBRT balance sheet, we can derive this residual component using the following equation:
(Al) Reserves held by banks = Open Market Operations + Use of Standing Facilities + Autonomous Factors = (1 -week repo + 1-month repo + PD Repo) + (CBRT O/N lending + BISTO//V repo - CBRT O/N deposit - BIST O/N reverse repo) + (Net Foreign Assets + Other autonomous factors - Currency in circulation - Government Deposits)
Reserves held by the banking system proxy the demand side of liquidity while the right-hand side of the equation mainly consists of liquidity supply factors. By using the equation above, one can state, that by taking into account the net effects of autonomous factors, the CBRT provides liquidity through OMO in such a way that the financial institutions can fulfill their reserve requirements throughout the maintenance period. A loose liquidity management of the CBRT increases the use of liquidity draining standing facilities such as CBRT O/N deposit and BIST O/N reverse repo. On the other hand, the use of liquidity-providing standing facilities such as CBRT PD repo, CBRT O/N lending, and BIST O/N repo facilities increases if the CBRT's monetary policy stance is tight.
TABLE Al Simplified Version of CBRT's Balance Sheet Assets Liabilities 1. Net foreign assets 5. Currency in circulation 2. Open market operations 6. Liabilities to government (Gov. Accounts) 2.1. 1-week refinancing 7. Reserves held by the banking operations system 2.2. 1-month refinancing 7.1. Required reserve balances operations 2.3. Overnight refinancing 7.2. Excess reserve balances operations 2.3.1 Primary dealer repo 2.3.2a. BIST O/N repo 2.3.2b. BIST O/N reverse repo 3a. CBRT O/N lending facility 3b. CBRT O/N deposit facility 4. Other autonomous factors
BIST: Borsa Istanbul
CBRT: Central Bank of the Republic of Turkey
ECB: European Central Bank
EONIA: Euro Overnight Index Average
IRM: Interbank Repo-Reverse Repo Market
LON: Late Overnight Window
OMO: Open Market Operations
PD: Primary Dealer
TL: Turkish Lira
Basci, E., and H. Kara. "Finansal Istikrar ve Para Politikasi." Iktisat Isletme ve Finans, 26(302), 2011, 9-25.
Bech, M., and C. Monnet. "The Impact of Unconventional Monetary Policies on the Overnight Interbank Market," Reserve Bank of Australia Conference, Sydney, Australia, August 19-20, 2013.
Beirne, J. "The EONIA Spread Before and During the Crisis of 2007-2009: The Role of Liquidity and Credit Risk." Journal of Internationa! Money and Finance, 31, 2012, 534-51.
Berentsen, A., and C. Monnet. "Monetary Policy in a Channel System." Journal of Monetary Economics, 55, 2008, 1067-80.
Bindseil, U., and F. Seitz. "The Supply and Demand for Eurosystem Deposits: The First 18 Months." European Central Bank Working Paper, No. 44, 2001.
Binici, M. H. Erol, H. Kara, P. Ozlu, and D. Unalmic. "Interest Rate Corridor: A New Macroprudential Tool?" Central Bank of the Republic of Turkey Research Notes in Economics, No. 13/20, 2013.
Brunetti, C., M. di Filippo, and J. Harris. "Effects of Central Bank Intervention on the Interbank Market during the Sub-Prime Crisis." Review of Financial Studies, 24(6), 2011,2053-83.
CBRT. Press Release on Amendments Related to Open Market Repo Auctions, No: 2011-28, 2011.
--. "Monetary and Exchange Rate Policy for 2013." 2012. Accessed December 2012. http://www.tcmb.gov. tr/yeni/announce/2013/monetary_2013 .pdf.
--. "Monetary and Exchange Rate Policy for 2014." 2013. Accessed December 2013. http://www.tcmb.gov. tr/yeni/announce/2013/monetary_2014.pdf
Joyce, M., A. Lasaosa, I. Stevens, and M. Tong. "The Financial Market Impact of Quantitative Easing in the United Kingdom." International Journal of Central Banking, 7(3), 2011, 113-61.
Kara, H. "KUresel Kriz Sonrasi Para Politikasi (Monetary Policy after the Global Crisis)." Iktisat Isletme ve Finans, 26 (302), 2012, 9-25.
--. "Safeguarding Macroeconomic and Price Stability Amid Capital Flow Volatility." Note Prepared for the BIS Meeting of Central Bank Chief Economists, Basel, March 2013.
Krishnamurthy, A., and A. Vissing-Jorgensen. "The Effects of Quantitative Easing on Interest Rates." NBER Working Paper No. 17555,2011.
Kucuk, H., P. Ozlu, A. Talaslt, D. Unalmic, and C. Yuksel. "Interest Rate Corridor, Liquidity Management and the Overnight Spread." Central Bank of the Republic of Turkey Working Paper, 12/04, 2014.
Linzert, T., and S. Schmidt. "What Explains the Spread between the Euro Overnight Rate and the ECB's Policy Rate?" International Journal of Finance & Economics, 16, 2011, 275-89.
Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, 1983.
Nautz, D., and C. J. Offermanns. "The Dynamic Relationship between the Euro Overnight Rate, the ECB's Policy Rate and the Term Spread." International Journal of Finance & Economics, 12, 2007, 287-300.
Ostry, J. D., A. R. Ghosh, K. F. Habermeier, M. Chamon, M. S. Qureshi, and D. B. S. Reinhardt. "Capital Inflows: The Role of Controls," International Monetary Fund Staff Position Note, No. 2010/04, 2010.
Pradhan, M., R. Balakrishnan, R. Baqir, G. Heenan, S. Nowak, C. Oner, and S. Panth. "Policy Responses to Capital Flows in Emerging Markets," International Monetary Fund Staff Discussion Note, No. 2011/10, 2011.
Szczerbowicz, U. "Are Unconventional Monetary Policies Effective," Department of Economics and Business, LUISS Guido Carli Working Paper, No. 07, 2011.
--. "The ECB Unconventional Monetary Policies: Have They Lowered Market Borrowing Costs for Banks and Governments?" CEPII Working Paper No: 2012-36, December 2012.
Whitesell, W. "Interest Rate Corridors and Reserves." Journal of Monetary Economics, 53, 2006, 1177-95.
(1.) Pradhan et al. (2011) and Ostry et al. (2010) review the policy responses of emerging markets to capital flows in the aftermath of the global financial crisis.
(2.) Kara (2013) explains in detail various policy instruments introduced in the new policy framework and discusses how exchange rate and credit channels are affected by the new policy instruments of the CBRT.
(3.) See Basci and Kara (2011), Kara (2012), and Kara (2013) for detailed descriptions of the policy framework implemented by the CBRT in the post-crisis period.
(4.) The Federal Reserve, the European Central Bank, the Reserve Bank of Australia, the Bank of Japan, the Bank of Canada, and the Bank of England are a few examples. In fact, although many countries adopt the corridor system, there can be differences in its implementation. See Berentsen and Monnet (2008) and Whitesell (2006) for the details of the interest rate corridor system.
(5.) See Bech and Monnet (2013) for an analysis on selected developed countries.
(6.) To be more precise, "CBRT average funding rate" or the "effective policy rate" is the weighted average cost of outstanding funding by the CBRT via Interbank Money Market (overnight lending facility) and Open Market Operations (BIST repo, primary dealer repo, 1-week repo via quantity auction, 1-week repo via traditional auction, and 1-month repo). Simple interest rates are used in calculation.
(7.) See Binici et al. (2013) for the use of interest rate corridor as a macroprudential tool in Turkey.
(8.) One major difference of our approach from the literature mentioned above is that money market spread that we are analyzing does not incorporate any counterparty risk as both types of borrowing are subject to similar types of collateral. Thus, in our case, the widening in spread is due to a change in the operational framework of monetary policy instead of a rise in counterparty risk or credit risk following the financial crisis.
(9.) PD banks refer to selected banks that lead the market.
(10.) The CBRT ceased funding through 1-month repo auctions as of November 15, 2013.
(11.) As the BIST repo/reverse repo market is the largest organized repo market in Turkey, the overnight rate for the repo transactions in this market acts as the "benchmark overnight market rate" for the money markets in Turkey.
(12.) In the traditional repo auction, all the bids are ranked from highest to lowest rate and the amount announced for the auction is allocated according to this ranking. In other words, the rates on successful bids determine the auction repo rate.
(13.) Additional monetary policy tightening policy was terminated as of January 29, 2014.
(14.) By definition, central banks open intra-day repo auctions in order to offset the adverse effects of unexpected liquidity shocks or central banks' liquidity forecast errors. In Turkey, CBRT had used intra-day repo auctions on additional monetary tightening days from 2011 until mid-2012. On these days, CBRT did not provide repo funding via quantity auctions at the policy rate and instead it opened 1-week intra-day auctions via the traditional method. CBRT has not used intra-day repo auctiOons since June 4, 2012.
(15.) One last facility that should be mentioned is the intra-day limit facility that banks can use during the trading hours to meet their temporary liquidity needs by paying a certain amount of commission.
(16.) The Appendix of this study provides details of the liquidity supply/demand factors.
(17.) The CBRT's daily liquidity forecasts are not open to the public. Confidential data that start from 2011 show that forecasts capture the actual net liquidity deficit quite well. For January 2011-May 2013, the correlation between these two series is 97%. In order to strengthen our argument that the net liquidity deficit is largely exogenous to the BIST spread, we repeated the regression analysis using the forecast of this variable (instead of actual values) for the available sample and obtained similar estimation results.
(18.) Average funding rate is a weighted average of different funding rates some of which are determined only once a month at the Monetary Policy Committee meetings. Thus, changes in the average funding rate are expected to be less pronounced than the changes in the overnight rate and the spread is usually led by the overnight rate.
(19.) If the total amount of bids is higher than the announced auction amount, the auction amount is distributed to the institutions via multiplying each institution's bid amount with the ratio of auction amount to total bid amount.
(20.) Market opening reserves are the total liquidity of the banking system at the beginning of a trading day which consists of free deposits of banks and expected net cash flows for that day. The reserves can be taken as a "buffer liquidity" of the banking system which can be used in case of unexpected liquidity shocks which may arise during the trading day. This variable is available at the CBRT web site on a daily frequency since 2007.
(21.) To ease the interpretation of the mentioned funding uncertainty variable, we use it after multiplying by -1. This way, when there are cash outflows (on tax payment days for instance), we observe a rise in funding uncertainty.
(22.) Since January 2011, banks obtaining funds from non-bank financial establishments (such as investment funds) at the BIST repo/reverse repo market are liable for holding required reserves. On the other hand, the IRM is established for repo transactions among banks and these transactions are not subject to reserve requirements. The difference between these two repo markets is reflected on the average overnight repo interest rates, where the rates in the IRM tend to be higher than the rates in the BIST repo market (bank to non-bank transactions) because the rates in general repo/reverse repo market include the reserve requirement holding cost.
(23.) We are grateful to the referee for suggesting this methodology. An ordinary least squares estimation where we remove the bounded observations yields similar results to the censored regression model. The ordinary least squares estimation results are available in the working paper version of this study (Kucuk et al. 2014).
(24.) See Maddala (1983) for details of the estimation methodology.
(25.) The starting date for the conventional policy episode is determined by the availability of the overnight swap rate which is used to calculate the interest rate expectations.
(26.) In Turkey, not only deposits but most major balance sheet items are subject to reserve requirement and the reserve requirement ratios also change depending on the maturity structure. As of the end of 2011, reserve option mechanism (ROM) has been added to the monetary policy mix. ROM allows banks to voluntarily hold a certain proportion of their TL reserve requirements in foreign exchange (FX) and/or gold. Under the ROM, it is difficult to gauge the currency composition of other banks' reserve holdings.
(27.) The autonomous factors in our study include Net Foreign Assets, Currency in Circulation, Liabilities to Government. and Other Autonomous Factors.
HANDE KUCUK, PINAR OZLU, ISMAIL ANIL TALASLI, DEREN UNALMIC and CANAN YUKSEL *
* We thank the referees for helpful comments and suggestions that improved the study to a great extent. We also thank participants at the seminar series of the Central Bank of the Republic of Turkey (CBRT), Borsa Istanbul Finance and Economics Conference 2013, 6th International Finance and Banking Society Conference 2014, and Econ Anadolu Conference 2015. The views expressed in this article do not necessarily represent those of the CBRT.
Kucuk: Research and Monetary Policy Department, Central Bank of the Republic of Turkey, Ankara 06050, Turkey. Phone +90 312 5075407, Fax +90 312 5075732, E-mail email@example.com
Ozlu: Research and Monetary Policy Department, Central Bank of the Republic of Turkey, Ankara 06050, Turkey. Phone +90 312 5075457, Fax +90 312 5075732, E-mail firstname.lastname@example.org
Talash: Markets Department, Central Bank of the Republic of Turkey, Ankara 06050, Turkey. Phone +90 312 5075273, Fax +90 312 5075289, E-mail email@example.com
Unalmic: Research and Monetary Policy Department, Central Bank of the Republic of Turkey, Ankara 06050, Turkey. Phone +90 312 5075413, Fax +90 312 5075732, E-mail firstname.lastname@example.org
Yuksel: Research and Monetary Policy Department, Central Bank of the Republic of Turkey, Ankara 06050, Turkey. Phone +90 312 5075452, Fax +90 312 5075732, E-mail email@example.com
TABLE 1 Descriptive Statistics of the Overnight Spread (a) Standard Time Period Mean Deviation Complete sample January 2005-May 2013 0.024 0.856 Old framework January 2005-May 2010 -0.087 0.377 New framework May 2010-May 2013 0.226 1.318 (a) BIST overnight rate minus CBRT average funding rate. TABLE 2 Explanatory Variables and Expected Signs Explanatory Variable and Its Notation Explanation CBRT weekly and longer-term Weekly and longer-term funding ([cbt.sup.funds.sub.t]) maturity funding by the CBRT as a ratio of the total TL reserve requirements of the banks Net liquidity deficit CBRT's net overnight funding ([L.sub.f]) at BIST repo market + PD repo usage as a ratio of the total reserve requirements of the banks Maturity structure of CBRT The ratio of weekly CBRT funding ([maturity.sub.t-1]) funding to the total of weekly and longer-term maturity CBRT funding [DELTA] Upper bound of Change in CBRT's overnight corridor ([DELTA][ub.sub.T]) lending rate [DELTA] Lower bound of Change in CBRT's overnight corridor ([DELTA][lb.sub.t]) borrowing rate [DELTA] Bid-to-cover ratio The bid-to-cover ratio in the ([DELTA]bid/cover) open market repo auctions conducted by CBRT Funding uncertainty Ratio of the expected ([uncertainty.sub.t]) intra-day cash outflow/ inflow to the amount of reserve balances at the beginning of the day Liquidity distribution among banks Ratio of the volume of ([H.sub.t-1]) overnight repo transactions among banks in BIST IRM to the total volume of overnight transactions in the two BIST repo-reserve repo markets Interest rate expectations The difference between the ([i.sup.E.sub.t]) overnight swap rate and the CBRT's funding rate Cumulative reserve position of banks Cumulated average reserve ([RP.sub.t-1]) fulfillment since the beginning of the maintenance period as a ratio of the average reserve requirement Dummy: First day Dummy for the first day of (D.sub.t.sup.first day]) the maintenance period Dummy: Last day Dummy for the last day of the ([D.sub.t.sup.last day]) maintenance period Dummy: Balance sheet Dummy for the last working ([D.sub.t.sup.balance sheet]) day of the quarter-end Dummy: Tax payments Dummy for the days of high ([D.sub.t.sup.tax]) tax payments by the banks Dummy: Auction amendments Dummy for the period after ([D.sub.t.sup.auctionamendments]) August 2011 (limits are imposed on the total bids by the banks) Explanatory Variable and Its Notation Expected Sign CBRT weekly and longer-term - funding ([cbt.sup.funds.sub.t]) Net liquidity deficit + ([L.sub.f]) Maturity structure of CBRT + funding ([maturity.sub.t-1]) [DELTA] Upper bound of + corridor ([DELTA][ub.sub.T]) [DELTA] Lower bound of - corridor ([DELTA][lb.sub.t]) [DELTA] Bid-to-cover ratio + ([DELTA]bid/cover) Funding uncertainty + ([uncertainty.sub.t]) Liquidity distribution among banks + ([H.sub.t-1]) Interest rate expectations + ([i.sup.E.sub.t]) Cumulative reserve position of banks - ([RP.sub.t-1]) Dummy: First day + (D.sub.t.sup.first day]) Dummy: Last day - ([D.sub.t.sup.last day]) Dummy: Balance sheet + ([D.sub.t.sup.balance sheet]) Dummy: Tax payments + ([D.sub.t.sup.tax]) Dummy: Auction amendments - ([D.sub.t.sup.auctionamendments]) TABLE 3 Estimation Results for the New Policy Period (May 20, 2010-May 6, 2013) Standard Coefficient Error Dependent variable: BIST spread Dependent variable (-1) 0.64 *** 0.05 Dependent variable (-10) 0.04 0.03 Policy variables CBRT weekly and longer-term funding -0.20 *** 0.06 Net liquidity deficit 1.88 *** 0.34 Maturity structure of CBRT funding (-1) 0.54 *** 0.17 [DELTA] Upper bound of corridor 0.55 *** 0.06 [DELTA] Lower bound of corridor -0.32 ** 0.12 Liquidity demand and conditions [DELTA] Bid/Cover ratio 0.13 *** 0.03 Funding uncertainty 0.35 *** 0.13 Liquidity distribution among banks (-1) 0.48 *** 0.17 Interest rate expectations 0.21 *** 0.04 Cumulative reserve position of banks (-1) 0.31 0.36 Dummy: First day 0.17 ** 0.08 Dummy: Last day -0.28 *** 0.09 Dummy: Balance sheet 1.18 *** 0.31 Dummy: Tax 0.29 *** 0.09 Dummy: Auction amendments -0.07 0.05 Constant -0.55 0.41 Scale 0.53 *** 0.03 Log likelihood -567.4 Left censored observations 57 Right censored observations 39 Uncensored observations 667 Note: Standard errors are adjusted for autocorrelation and heteroskedasticity. *** and ** denote 1% and 5% significance levels. TABLE 4 Standardized Estimation Results for the New Policy Period (May 20, 2010-May 6, 2013) Standard Coefficient Error Dependent variable: BIST spread Dependent variable (-1) 0.64 *** 0.05 Dependent variable (-10) 0.04 0.03 Policy variables CBRT weekly and longer-term funding -0.08 *** 0.02 Net liquidity deficit 0.18 *** 0.03 Maturity structure of CBRT funding (-1) 0.10 *** 0.03 [DELTA] Upper bound of corridor 0.06 *** 0.01 [DELTA] Lower bound of corridor -0.05 ** 0.02 Liquidity demand and conditions [DELTA] Bid/Cover ratio 0.09 *** 0.02 Funding uncertainty 0.07 *** 0.02 Liquidity distribution among banks (-1) 0.08 *** 0.03 Interest rate expectations 0 23 0.04 Cumulative reserve position of banks (-1) 0.01 0.01 Dummy: First day 0.13 ** 0.06 Dummy: Last day -0.21 *** 0.06 Dummy: Balance sheet 0.90 *** 0.23 Dummy: Tax 0.22 *** 0.07 Dummy: Auction amendments -0.05 0.03 Constant -0.05 *** 0.02 Scale 0.40 *** 0.02 Log likelihood -382.6 Left censored observations 57 Right censored observations 39 Uncensored observations 667 Notes: Standard errors are adjusted for autocorrelation and heteroskedasticity. *** and ** denote 1% and 5% significance levels. TABLE 5 Estimation Results for the Conventional Policy Period (November 29, 2007-May 18, 2010) Standard Coefficient Error Dependent variable: BIST spread Dependent variable (-1) 0.71 *** 0.08 Dependent variable (-10) 0.09 0.06 Policy variables CBRT weekly and longer-term funding -0.27 *** 0.08 Net liquidity deficit 2.50 *** 0.51 Liquidity demand and conditions [DELTA] Bid/Cover ratio 0.04 * 0.02 Interest rate expectations 0.01 0.02 Cumulative reserve position of banks (-1) -0.10 0.21 Dummy: First day 0.12 *** 0.04 Dummy: Last day -0.02 0.06 Dummy: Balance sheet 0.22 ** 0.10 Dummy: Tax 0.08 0.07 Constant 0.06 0.23 Scale 0.21 *** 0.02 Log likelihood -46.8 Left censored observations 450 Right censored observations 0 Uncensored observations 184 Note: Standard errors are adjusted for autocorrelation and heteroskedasticity. ***, **, and * denote 1%, 5%, and 10% significance levels. TABLE 6 Standardized Estimation Results for the Conventional Policy Period (November 29, 2007-May 18, 2010) Standard Coefficient Error Dependent variable: BIST spread Dependent variable (-1) 0.71 *** 0.08 Dependent variable (-10) 0.09 0.06 Policy variables CBRT weekly and longer-term funding -0.29 *** 0.08 Net liquidity deficit 1.00 *** 0.20 Liquidity demand and conditions [DELTA] Bid/Cover ratio 0.11 * 0.05 Interest rate expectations 0.03 0.24 Cumulative reserve position of banks (-1) -0.03 0.06 Dummy: First day 0.40 *** 0.14 Dummy: Last day -0.07 0.21 Dummy: Balance sheet 0.70 ** 0.32 Dummy: Tax 0.27 0.24 Constant -1.34 *** 0.20 Scale 0.67 *** 0.06 Log likelihood -262.04 Left censored observations 450 Right censored observations 0 Uncensored observations 184 Note: Standard errors are adjusted for autocorrelation and heteroskedasticity. ***, **, and * denote 1%, 5%, and 10% significance levels.
|Printer friendly Cite/link Email Feedback|
|Author:||Kucuk, Hande; Ozlu, Pinar; Talasli, Ismail Anil; Unalmis, Deren; Yuksel, Canan|
|Publication:||Contemporary Economic Policy|
|Date:||Oct 1, 2016|
|Previous Article:||Job reallocation, financial crises, and policy reforms: evidence from Korea.|
|Next Article:||Contemporary Economic Policy 2015 editor's report.|