نوبت کاری های سیاست های پولی درون زا و ساختار اصطلاح: شواهدی از بازده اوراق قرضه دولت ژاپن

I construct a no-arbitrage term structure model with endogenous regime shifts and apply it to Japanese government bond (JGB) yields. This model subjects the short-term interest rate to monetary regime shifts, specifically a zero interest rate policy (ZIRP) and normal regimes, which depend on macroeconomic variables. The estimates show that under the ZIRP regime, the effect of deflation (inflation) on lowering (raising) bond yields amplifies on the long end of yield curves, compared with a case with positive interest rates under the normal regime. On the other hand, output gaps’ ability to raise bond yields weakens for all maturities

How are bond yields affected by endogenous shifts in and out of a zero interest rate policy (ZIRP)? Policy shifts are typically endogenous because they depend on the state of economy. For example, a ZIRP may be introduced when the Taylor-rule policy rate, which is a function of real activity and inflation variables, hits the zero lower bound of the policy rate. Policymakers may lift a ZIRP when the exit conditions on macroeconomic variables are satisfied. The answer to this question is relevant for countries currently under a ZIRP and struggling to meet macroeconomic conditions to exit the ZIRP. The answer is of particular importance for Japan, which has already experienced such policy shifts several times. For Japan, small increases in bond yields can strain public finances with the ratio of public debt to GDP already exceeding 200%. I examine this question by constructing a no-arbitrage term structure model with discrete regime shifts. The model has three key features. First, the probability of transitioning from a ZIRP depends on the state variables that appear in the monetary policy rule (e.g., output gap and inflation), allowing the entry and exit conditions of a ZIRP to depend on the state. Second, the model uses discrete regime shifts in the affine term structure framework, addressing possible nonlinearity in the conditional means of short-term yields (Ang and Bekaert, 2002).1 Third, the state vector, which includes the policy rate, depends on the current, rather than the previous, monetary policy regime. This third feature is not trivial: today’s policy rate should depend on today’s monetary policy regime.2 For example, a ZIRP status from the previous period does not guarantee that today’s policy rate is set at the ZLB. Putting differently, if the model did not include the third feature, it would inappropriately allow a policy rate well above zero even during ZIRP periods. The model includes the policy rate in the state vector so that the lagged policy rate can affect the dynamics of macroeconomic variables, as modeled in the standard monetary VAR models (e.g., Stock and Watson, 2001) and several macro-finance term structure models (e.g., Ang et al., 2006 and Hördahl et al., 2006). This paper’s model is related to the existing discrete regime-switching affine term structure models (ATSMs)3 in the following ways. First, most existing models that incorporate the third feature assume a constant transition matrix (e.g., Bansal and Zhou, 2002, Ang et al., 2008 and Hamilton and Wu, 2012). This paper extends these models by introducing state dependent transition probabilities. Second, Dai et al. (2007, henceforth DSY) implement state-dependent transition probabilities and discrete regime shifts (the second and third features) under the data generating or physical (P)(P) measure, while their state vector depends on the previous regime. This paper extends DSY’s work by adding dependence on the current policy rate (the third feature) and by providing formal propositions and proofs and discussion on the link between the PP and the risk neutral (Q)(Q) measures. To apply the model to Japanese government bond (JGB) yields, I consider a model with (i) two observable4 regimes given that the actual Japanese policy interest rate process (Fig. 1) appears to have at least two regimes: a period during which the policy interest rate is near zero and flat (the ZIRP regime) and the remaining periods (the normal regime); (ii) two types of regime evolutions following Oda and Oda and Ueda’s (2007) set up with ZIRP exit rules: one that incorporates the zero lower bound (Type I evolution) and one that additionally includes the forward guidance (Type II evolution). Under the latter evolution, the Bank of Japan maintains the zero rate until some inflation condition is satisfied. (for a discussion on the Bank of Japan’s forward guidance policy, see e.g., Ueda, 2012a and Ueda, 2012b and Ugai’s survey, 2007).The estimated yield curves demonstrate how bond yields are affected by the monetary policy regimes shifts: the effect of deflation (inflation) on lowering (raising) bond yields amplifies on the long end of yield curves under the ZIRP regime. On the other hand, output gaps’ ability to raise bond yields weakens for all maturities under the ZIRP regime. One may draw several policy implications from these findings. First, inappropriate decisions in terminating a ZIRP can heighten risks in the JGB markets. In particular, a delay in lifting a ZIRP while allowing some inflation may lead to higher long-term bond yields (due to the amplified effect of inflation on raising bond yields under the ZIRP regime). In reality, such a delay can occur when the achieved inflation does not reach the targeted level (e.g., below 2%) or the economy continues to stagnate. Second, low JGB yields can be caused by a ZIRP under the continued deflation (due to the amplified effect of deflation on lowering bond yields under the ZIRP regime). This downward pressure on JGB yields, however, will be lifted when a ZIRP ends. Thus there is upward risk on the JGB yields when Japan exits from a ZIRP. Besides examining how bond yields are affected by the endogenous monetary policy shifts, the paper finds that the estimated state-dependent transition probabilities are more persistent when the forward guidance is present. Furthermore, empirical evidence suggests that the evolution with the forward guidance fits the data much better than without it; out-of-sample performance results, however, are mixed. Lastly, the estimated term premia in JGBs indicate that the large bond yield decline in the early 1990s was driven by expectation components, whereas that of the late 1990s was driven by both expectations and term-premium components. Term premia also declined after the introduction of the quantitative easing monetary policy (QEP) in March 2001. This paper proceeds as follows. Section 2 describes a term structure model with endogenous regime shifts. Section 3 describes the specific regime evolutions considered. Sections 4 and 5 discuss the estimation strategy and results. Section 6 concludes the paper.

I construct a no-arbitrage affine term structure model with state-dependent policy shifts. In the model, the state vector depends on the current policy regime. As an application of the model, I examine how the JGB yields fluctuate with macroeconomic variables in the presence of the endogenous monetary policy shifts that incorporate the zero lower bound and the Bank of Japan’s forward guidance. I also analyze yield dynamics by decomposing JGB yields into the expectations and term-premium components. The estimated results indicate that under the ZIRP regime, deflation plays a growing role in lowering JGB yields, especially on the long end of yield curves. On the other hand, output gaps’ ability to raise bond yields weakens for all maturities. As discussed earlier, these results flag upward risks on the JGB yields when Japan exits from a ZIRP or when the country delays in lifting a ZIRP while allowing some inflation. While this paper focuses on interest rate policy, the effect of other nonconventional policy measures, such as current account balances and monetary base, on bond yields should be analyzed in future research. This effect may depend onto what extent boosting current account balances or monetary base can actually induce inflation and stimulate growth. If the extent of quantitative easing effects is large then the likelihood of (interest rate) lift off in the future increases, according to this paper’s analysis of how endogenous policy shifts affect bond yields. As a result, long-term bond yields may rise today with other things unchanged. Lastly, looking forward, I believe it is important to understand not only “normal” bond yield responses to moderate inflation and economic growth, but also the channels that can steeply raise macroeconomic variables, especially inflation, and thus jeopardizing the JGB markets.