Evidence of Effectiveness of Unconventional Policies

Although the theoretical literature on unconventional monetary policy has blossomed extensively over the past 20 years, empirical analysis of the impacts has been much more limited. This is not surprising, given that until recently, only Japan in the period of 1999–2006 provided data on the experience of monetary policy at zero interest rates, at least since the Great Depression, and on quantitative easing between 2001 and 2006. Some analyses of the US experience with zero interest rates and unconventional policy are just beginning to emerge, e.g., McAndrews (2009). Unfortunately, this means that the bulk of the “empirical” analysis of unconventional monetary policies has been conducted with various kinds of macroeconomic models such as vector auto-regression (VAR) models. This raises questions about the validity of the results, as they depend critically on the extent to which the models capture the underlying behavior of the economy. Given the complexity and nonlinearity of the economy, one has to approach the results of this literature with a good deal of skepticism. This section reviews the evidence for the effectiveness of different kinds of unconventional monetary policies.

Analyses of unconventional monetary policy impacts face a number of other methodological problems. A number of different policies may be adopted at the same time, which makes it difficult to tease out their separate effects. Second, one has to identify the “counterfactual,” i.e., what would have happened in the absence of such policy steps? Third, it is necessary to identify the extent to which a specific announcement was a surprise to the market. Fourth, spillover effects may be important, i.e., market conditions in one country may be influenced by easing measures in another country.

Commitment effect: There is a lot of evidence that announcements by central banks do affect market expectations about future policy, which should not be surprising, since market participants regard the direction of monetary policy to be an important influence on the path of interest rates and markets. Table 2 [ PDF 8.8KB | 1 page ] shows the results of a number of empirical studies of this effect. Most studies focused on the impact on the yield curve, which is the first and most obvious link in the transmission mechanism. Others looked at the impacts on credit spreads (which may be more relevant for the current crisis), real output, and inflation.

Early studies such as Fujiki, Okina, and Shiratsuka (2001) and Kuttner and Posen (2001) relied on casual inspection and came up with opposite conclusions regarding impacts on the yield curve in Japan. Later studies adopted more formal approaches. Japanese authors all found significant impacts of the commitment effect on the yield curve using a variety of methodologies, including time series/cross-section modeling of the yield curve and combinations of macro models with either a Taylor Rule decision function for monetary policy or an econometric model of the yield curve. Perhaps the most thorough of these are Okina and Shiratsuka (2004) and Baba et al. (2005b). However, Bernanke, Reinhart, and Sack (2004) argued that some of these studies did not adequately control for how rates would have moved in the absence of the policies adopted. Using an “event study” and “macro/finance” model approach, they analyzed the effects of Fed and BoJ statements on expected short-term interest rates, decomposing the impact into very short-term, 1-year forward, and 5-years forward. They found that, in the case of Japan during the period of the ZIRP, effects were statistically significant, but of a much smaller magnitude than in the case of the US—only about one-third as large. Also, they tended to be concentrated at the shorter end of the yield curve—the impact on the long end of the curve was relatively modest. Notably, they found no impact of BoJ statements on year-ahead expectations of short-term interest rates, a significant difference from the results for US Federal Reserve statements. This suggests that the BoJ statements themselves may not have been framed carefully enough to have maximum impact on market expectations.

Oda and Ueda (2007) used a similar approach and found evidence that the commitment effect did tend to lower interest rates. However, the effect was most pronounced for rates of three years or less, and was much more pronounced after the economic recovery began. This suggests that the commitment effect is least effective when it is most needed, i.e., when the economy is still in recession.

Fujiki and Shiratsuka (2002), Fujiwara et al. (2005), and Baba et al. (2005a) all found significant effects of the commitment effect in bringing down credit spreads as well. For example, Baba et al. (2005a) found that the switch to the ZIRP and then the adoption of definitions of ending deflation were correlated with a reduction in bank credit spreads.

Reifschneider and Williams (2000), using a macro model with a modified Taylor Rule that took into account past deviations in output resulting from the zero bound, found that the commitment effect had a significant impact on output and inflation in the US. Fujiki and Shiratsuka (2002), Fujiwara et al. (2005), and Braun and Waki (2006) all found positive impacts on output and inflation in Japan using a similar approach. However, these results have to be treated with skepticism, since, as mentioned above, they depend on the accuracy of the macro models used, which, in most cases, are highly simplified. Therefore, the overall conclusion appears to be that commitment effects do stabilize market expectations about the path of short-term interest rates and thereby tend to lower long-term rates. However, at least in the case of Japan, these effects were not large enough to affect expectations about the real economy and inflation sufficiently to produce effects on those variables.

The Fed reintroduced its commitment effect language in its statement on 16 December 2008 when it cut the target range for the Fed funds rate to 0–0.25%, noting that “…the Committee anticipates that weak economic conditions are likely to warrant exceptionally low levels of the federal funds rate for some time” (US Federal Reserve Board 2008). Figure 4 [ PDF 22.3KB | 1 page ] shows that longer-term and short-term interest rates all fell sharply in December 2008. In particular, the one-year T-bill rate fell more than the three-month T-bill rate, and also considerably more than the two-year T-note rate. This suggests that the market expected that rates would stay “very low” for up to one year, pointing to a significant commitment effect. Of course, it may simply have reflected the market's worsening assessment of the economic situation, but the differential performance between the one-year note and the two-year note is suggestive. Interestingly, no other central bank has adopted a formal commitment regarding its policy during the current global downturn.

Quantitative easing: The findings of the literature about the impacts of quantitative easing on interest rates and economic activity are generally positive, but more tentative than in the case of the commitment effect. The results of some major studies are summarized in Table 3 [ PDF 22.3KB | 1 page ].

Regarding the yield curve, Baba et al. (2005b) found that the level of CABs was significant in explaining the “expectations theory” component of interest rates, i.e., the market's expectations of future short-term interest rates, but not the risk premium. They suggested that the level of CABs could have functioned as a signaling mechanism to strengthen the commitment effect. However, they cautioned that this correlation could be spurious, and that statements by the BoJ Governor made at the same time as the CAB level announcements could have been the main factor instead. On the whole, they concluded that the BoJ's monetary policy worked mainly through the commitment channel in the period 1999–2004. Bernanke, Reinhart, and Sack (2004) found evidence suggesting that Japanese bond yields were lower than otherwise would have been expected, but their model did not differentiate between the impacts of the commitment effect and the CAB level. Interestingly, Okina and Shiratsuka (2004: 94) concluded that the instantaneous forward rate curve “…was hardly influenced by the increase in the target level of the current-account balance at the BoJ from over 6 trillion yen to 10–15 trillion yen on 19 December 2001. This indicates that the strengthening of quantitative monetary easing was not perceived as sufficient stimulus to curb deflation, coupled with low economic growth.” This suggests that there was little independent contribution from quantitative easing beyond that of the commitment effect, which did seem to flatten the yield curve.

There is some evidence that the ample provision of liquidity did ease banks' funding constraints and shrink credit spreads. Baba et al. (2005a) found a positive effect of increasing CAB levels on reducing the dispersion of bank credit spreads in the interbank market. They noted that as the BoJ had to fund successively higher CAB levels, it had to move further out along the yield curve to conduct its operations, which tended to flatten the yield curve. They concluded that both the commitment effect and quantitative easing probably tended to reduce credit spreads in the interbank market, although they were unable to quantify their relative contributions.

Figure 5 [ PDF 38.2KB | 1 page ] shows the relationship between the CAB level (inverted) and the 10-year JGB yield. The impact of the CAB levels on bond yields is not clear, as bond yields began falling in 2002 before CABs started to rise, and began rising again mid-2003, even though the CABs continued to rise until late 2004. Moreover, bond yields began to rise sharply in late 2005, well before the fall of the CABs in mid-2006. CABs have begun to rise again in 2009, but there has been little evident impact on bond yields.

The Bank of England is the only major central bank to have adopted quantitative easing during the current global financial crisis, as it set a target of £75 billion for reserve deposits on 5 March 2009 and subsequently raised this to £125 billion on 7 May 2009 and again to £175 billion on 6 August 2009. Figure 6 [ PDF 19.2KB | 1 page ] shows the spread between the three-month sterling London interbank offered rate (Libor) and the base rate. The figure shows that the spread narrowed rapidly after 20 March 2009, although it is not clear if this was affected by other factors as well. This may provide some evidence of the effectiveness of quantitative easing in reducing credit spreads. The Bank of England's (BoE) QE policy was actually a mix of both quantitative easing and qualitative easing; although it targeted the level of reserve deposits, it accomplished this primarily through purchases of UK government bonds (“gilts”) rather than short-term paper. Meier (2009) found these purchases to have been effective in lowering both gilt yields and interbank rate spreads.

Regarding impacts on output and inflation, Bernanke, Reinhart, and Sack (2004) ran simulations of QE policies on simple macro models of the US and Japan. They found that increases in CAB levels did have positive impacts on output and prices in both countries, although, again, the impacts for Japan were much less than those for the US. Using a similar approach for Japan, Honda, Kuroki, and Tachibana (2007) found positive impacts on output but not on prices. They identified equity prices as the main channel by which the QE policy affected output, which implies that the portfolio-balance effect was the main transmission mechanism. However, the lack of impact on prices is puzzling, and does cast doubts on the validity of the model and the robustness of the conclusions.

Kuttner and Posen (2001) used a VAR model to test the impact of the monetary base on broad money and prices, and found no significant impact since 1990, which was not particularly surprising, given that the normal credit transmission mechanism was not functioning then. They did not find evidence that the QE policy tended to weaken the yen either. However, their results were somewhat limited by the fact that M0 had shown little volatility during the estimation period, since it preceded the BoJ's adoption of QE policy.

On the whole, the evidence for a significant impact of the QE policy in addition to the commitment effect on interest rates, output, and inflation looks limited. However, the evidence that the QE policy helps to ease tightness in credit markets appears to be stronger. As will be discussed in the next section, qualitative easing targeted at specific asset markets is probably a more efficient way to lower credit spreads. Strikingly, the Bank of Japan did not return to quantitative easing during the latest downturn, despite the fact that growth has been far weaker than it was in 2001–2002 during the previous recession. This at least suggests that the Bank of Japan does not have much confidence in its efficacy, aside from its role in easing stresses in the financial sector.

Qualitative easing: Empirical research on the effects of qualitative easing is more limited (see Table 4 [ PDF 9.6KB | 1 page ]). Much of the analysis focuses on purchases of long-term government bonds, perhaps the logical first step toward unconventional purchasing operations. Shiller (1990) analyzed the attempts of the US Fed to influence the shape of the yield curve during the 1960s (“operation twist”), but was not able to find evidence that the effect of the policy was significant. However, this may be simply because the operation was not large enough. Bernanke, Reinhart, and Sack (2004) found that the announcement by the US Treasury in February 2000 that it would probably stop issuing 30-year bonds had a statistically significant impact in lowering yields on 20-year Treasury bond yields compared with Treasury bonds with shorter maturities. They also found evidence that purchases of US Treasury bills by the Japanese Ministry of Finance in 2003–2004 may have been consistent with a decline of bond yields of 50–100 basis points, but the evidence was not conclusive, as the contribution of other factors could not be ruled out, and the deviations could have been due to chance. They concluded on a positive note, however, stating that:

If the Federal Reserve were willing to purchase an unlimited amount of a particular asset—say, a Treasury security—at a fixed price, there is little doubt that it could establish that asset's price. Presumably, this would be true even if the Federal Reserve's commitment to purchase the long-lived asset were promised for a future date. (Bernanke, Reinhart, and Sack 2004: 60)

Nonetheless, the Fed's recent experience with buying US Treasuries has not been obviously successful. In the six months since the beginning of 2009, the Fed's outright holdings of US Treasury securities rose by US$172 billion (36%), but this did not stem a sharp rise in bond yields during that period, as the 10-year bond yield rose from 2.2% to 3.6% (see Figure 7 [ PDF 18.6KB | 1 page ]).

Of course, rates might have been even higher otherwise, but the Fed's basic goal of holding down mortgage loan rates was not achieved. Interestingly, the level of Fed holdings of Treasuries declined substantially between late 2007 and mid-2008, but this does not seem to have put any upward pressure on bond yields. Baba, Ho, and Hordahl (2009) also noted that declines in bond yields in response to bond purchase announcements by both the Fed and the BoE were very short-lived. These results probably should not be surprising in light of the huge size and liquidity of the US Treasury market (total value of US$6.8 trillion in March 2009) relative to the Fed's purchasing operations, which were announced to total US$300 billion, or 4.4% of the total.

As noted above, the Bank of Japan also significantly increased its outright purchases of JGBs during the period of the QE policy between August 2001 and October 2002. However, as was the case with quantitative easing, Oda and Ueda (2007) concluded that the BoJ's increased purchases of JGBs did not lead to a significant portfolio-rebalancing effect. Figure 8 [ PDF 37.6KB | 1 page ] shows that the increase in the level of outright purchases during period did lead the decline in 10-year JGB yields. However, this decline could also be attributed to the decline in US bond yields during the same period. Notably, Japanese bond yields rose in line with the increase in US bond yields from mid-2003, presumably reflecting the global economic recovery, even though outright bond purchases remained high at ¥1.2 trillion per month. Even more suggestively, Japanese bond yields rose since December 2008, again in sympathy with US bond yields, even though the outright purchases were increased dramatically further to ¥1.8 trillion per month by February 2009. Therefore, the most one can say is that the purchases may have diminished the extent of the increase of bond yields, but it is difficult to determine the size of this impact. The large size and liquidity of the Japanese bond market suggests that operations of this kind would have to be large indeed to have a substantial and lasting effect.

The Bank of England embarked on a large-scale program of purchasing UK government bonds (“gilts”) beginning in March 2009 in order to fund its target for central bank reserve deposits described above. Figure 9 [ PDF 37.6KB | 1 page ] shows the relationship between government bond yields and the level of reserve deposits. Its purchasing operations have been comparatively aggressive, as the BoE accumulated about 17% of total tradable government bonds in about four months (Financial Times 2009). Nevertheless, bond yields still rose by about 30 basis points between March and July 2009, after the start of the bond purchase program, so the effect looks somewhat limited. Taking into account relative movements of US and European bond yields over the same period, Meier (2009) estimated that four months after the announcement of the QE policy, it had lowered gilt yields by a range of at least 35–60 basis points, a significant, but not huge decrease.

Qualitative easing may be effective in reducing some credit-related financial stresses. For Canada, Yuan and Zimmerman (1999) used a dynamic general equilibrium model to analyze the effects of monetary easing and changes in required loan-to-deposit ratios on credit availability. They found that direct easing of loan standards was much more effective than conventional monetary easing in counteracting a credit crunch. McAndrews (2009) found that the Fed's TAF and central bank swap programs were effective in reducing spreads between the US (federal funds purchases and sales) and European (Eurodollar deposit) interbank markets. Figure 6 above shows that the announcement by the Bank of England on 13 October 2008 that it would provide unlimited dollar liquidity to the banking sector appears to have been the key factor in easing funding pressures in the interbank market at that time. We discuss the case of Bank of Korea's currency swaps in the section on developing economies below.

The Fed's recent qualitative easing moves appear to have had a significant impact on easing credit spreads of various kinds. For example, the “Ted” spread (the spread between the three-month Libor rate and the three-month Treasury bill rate) peaked at over four percentage points in November 2008, but began to fall rapidly thereafter (see Figure 10 [ PDF 14.7KB | 1 page ]). During this period the US monetary base roughly doubled in size to about US$1.7 trillion as a result of the combined impacts of purchases under the TAF, CPFF, and other programs, plus the Fed swap arrangements with other central banks. It is difficult to identify the relative effects of these different programs, but the combined impact appears to have been substantial. Because banks were the major beneficiaries of these moves, it is reasonable to see this reflected in the Ted spread, which mainly indicates the market's assessment of risk in the banking sector.

Spreads in the US money market and CP market also eased dramatically after the implementation of various Fed liquidity programs described above, including the AMLF in September 2008 and the MMIFF and CPFF in October 2008. (Capital injections into nine major US banks, announced on 14 October 2008, probably contributed to this easing effect as well.) Figure 11 [ PDF 31.9KB | 1 page ] shows that spreads of both Libor and financial commercial paper over Tbill rates declined sharply beginning in November 2008, and have since largely normalized.

The start of the ZIRP and the commitment effect in December 2008 probably had a further downward impact on spreads, although this is more difficult to confirm in terms of timing.

In contrast, spreads of corporate bond yields over those of US Treasury bonds did not peak until December 16 (see Figure 12 [ PDF 31.9KB | 1 page ]). This is precisely the day that the Fed announced the shift to the zero-interest-rate policy and the commitment to maintain it for “some time,” which is very strong evidence that this was the key factor, rather than the Fed's balance sheet activity. This suggests that rate expectations were more important for the non-financial corporate sector, rather than the direct effects of toxic items on the balance sheet. (To be sure, the spread for AAA bonds did spike briefly higher in March 2009, but the spread for BAA bonds clearly peaked in mid-December.)

The Bank of Japan also undertook a number of credit-easing measures, including “Special Funds-Supplying Operations to Facilitate Corporate Financing”(19 December 2008), outright purchases of commercial paper (22 January 2009), and outright purchases of corporate bonds (19 February 2009). Figure 13 [ PDF 29.6KB | 1 page ] shows that the spreads of Euroyen deposits and CP over financing bills declined markedly in the first three months of 2009. However, the spread for Euroyen deposits remained considerably more elevated than that for CP, indicating residual concerns about the financial position of Japanese banks.

These results suggest that qualitative easing or credit easing measures are not very effective in affecting the level of government bond yields, but can be quite effective in reducing spreads of rates of other financial products over those of risk-free rates, especially short-term rates. This is particularly so when concerns about liquidity and solvency lead to a credit crunch that essentially prevents certain markets from functioning normally. In other words, the central bank can successfully unplug logjams arising from a scarcity of funds in a particular segment. However, there is little evidence that such measures can affect inflation expectations or the demand for credit at the macro level. Finally, these studies typically look at individual countries in isolation, and hence may miss spillover effects. For example, Fed policies implemented in US markets may have helped to relieve stress in overseas markets.

Download this Paper [ PDF 294.1KB| 34 pages ].

Post a Comment

We welcome your feedback on this publication. Post a comment. ADBI is not obliged to acknowledge or publish comments and may abridge or edit them before web posting.

Comment(s)

There are [0] comment(s) for this entry. Post a comment.