How severe has the zero lower bound constraint been?

Eric Swanson

08 November 2014



In December 2008, the Federal Reserve’s Federal Open Market Committee lowered the federal funds rate to essentially zero, and has kept it there ever since. Because physical currency earns an interest rate of zero, it is generally impossible for the Open Market Committee to lower the federal funds rate substantially below zero, since banks would opt to hold physical currency rather than earn a significantly negative rate of return on cash balances held at the Fed. This barrier is commonly referred to as the ‘zero lower bound’.

According to traditional macroeconomic thinking, once monetary policy hits the zero lower bound, there is nothing more the Committee can do to stimulate the economy – monetary policy is essentially ‘stuck at zero’. A corollary of this observation is that fiscal policy becomes more powerful than in normal times because any stimulus from fiscal policy on output or inflation will not be partially offset by monetary policymakers raising interest rates to keep inflation in check. In other words, monetary policy will not act to ‘crowd out’ fiscal policy because interest rates will remain stuck at zero as long as the economy is weak (see, e.g., Mankiw 2013, Chap. 12).

The role of monetary expectations

More recent research, however, has emphasised how monetary policy expectations can alter this reasoning. Reifschneider and Williams (2000) and Eggertsson and Woodford (2003) show that, if the Federal Committee can credibly commit to future values of the federal funds rate, then it has the power to largely work around the zero lower bound constraint. As these authors point out, the economy depends not just on the current level of the federal funds rate (a one-day interest rate), but rather on the entire path of the expected future federal funds rate over the next several years. Put differently, businesses and households typically look at interest rates with maturities out to several years when making investment and financing decisions. Even if the current federal funds rate is stuck at zero, the Committee could continue to push longer-term interest rates lower by promising to keep the federal funds rate low for an extended period of time. In this way, the Committee could continue to stimulate the economy even when the current federal funds rate is constrained by the zero lower bound.

This line of reasoning suggests that monetary policy has probably not been as constrained by the zero lower bound as the traditional way of thinking would imply.  Figure 1 plots the federal funds rate along with the one-, two-, five-, and ten-year Treasury yields. Although the funds rate (solid black line) is essentially zero from December 2008 onward, even the one-year Treasury yield averages close to 0.5% throughout 2009 and 2010, and fluctuates noticeably as the outlook for the economy and monetary policy rose and fell over this period. The two-year Treasury yield is even higher and more volatile. Thus, Figure 1 suggests that monetary policy might not have been very constrained by the zero lower bound until at least mid-2011.

Figure 1.  Federal funds rate target and one-, two-, five-, and ten-year zero-coupon Treasury yields from January 2007 to September 2014

Note: Although the federal funds rate was essentially zero from December 2008 onward, longer-maturity yields were substantially above zero and responded to news over much of this period.
Source: Data are from the Federal Reserve Board.

Just looking at the level of yields in Figure 1 does not provide a very precise answer to the question posed in the title of this paper, however.

  • First, we’d like to be able to test whether the zero lower bound is a statistically significant constraint on a given interest rate (say, the one-year Treasury yield). 
  • Second, we’d like a quantitative measure of how severely any given interest rate is constrained. 
  • Third, the effective lower bound on interest rates might be a bit above zero for institutional reasons – see, e.g., Bernanke and Reinhart (2004) – which implies that an interest rate of 0.5% or even higher might nevertheless be best regarded as severely constrained.

New research

In a recent paper, my co-author and I propose a novel way of addressing these concerns and measuring the statistical significance and severity of a constraint on yields of any maturity (Swanson and Williams 2014). Virtually every business day, a major piece of macroeconomic data is released, such as nonfarm payrolls, GDP, CPI, etc. When these releases differ from financial market participants’ expectations, bond yields respond significantly in a systematic way. For example, if the number of employees on US nonfarm payrolls falls short of expectations, bond yields tend to fall substantially as the outlook for the economy is weaker and the Open Market Committee is less likely to raise interest rates in the near future. Bigger surprises in these data releases cause correspondingly larger movements in bond yields.

Because bond yields respond systematically to macroeconomic data releases and because we observe so many such announcements, we can use them as a benchmark to measure whether any given yield is behaving normally or not. For example, if a given yield is responding to news today in the same way as it always did in normal times (which is taken to be the period 1990-2000), this would suggest that yield is unconstrained. Alternatively, if a given yield used to respond to these announcements but no longer does, then that would be evidence that the yield is completely constrained. A yield could also be partially constrained, which would be the case if it still responds to news, but in an attenuated fashion.

This analysis is discussed in detail in our paper, but the results are summarised and updated in Figures 2-4 here. The thick blue line in Figure 2 plots the estimated ‘sensitivity coefficient’ for the three-month Treasury yield over the period 2001-14. Thinner, light blue lines above and below the main line depict ±2-standard-error-bands around the time-varying point estimates. A sensitivity coefficient of 1 indicates that the yield is responding normally to news, while a sensitivity coefficient of 0 indicates that the yield is not responding at all. Periods when the three-month yield’s sensitivity coefficient is statistically significantly less than 1 are shaded yellow, and periods when that sensitivity coefficient is also not significantly different from 0 are shaded red. Thus, red shaded regions are periods when the three-month yield appears to be completely constrained in its behaviour, while yellow shaded regions are periods when the yield seems to be partially – but not completely – constrained.

Not surprisingly, the three-month Treasury yield’s sensitivity to news is not significantly different from 0 from the spring of 2009 onward, suggesting that the zero lower bound was a severe constraint on that yield over that period.1

Figure 2.  Estimated time-varying sensitivity of three-month Treasury yield to macroeconomic announcements from January 2001 to September 2014. 

Note: A coefficient of 1 indicates normal sensitivity to news, while 0 indicates complete insensitivity. Thin blue lines depict ±2-standard-error bands around the time-varying point estimates. Yellow shaded regions denote sensitivity coefficient significantly less than 1; red shaded regions denote sensitivity coefficient significantly less than 1 and not significantly different from 0.

The results become more interesting as we look a little further out the yield curve.  Figures 3 and 4 report results for the one- and two-year Treasury yield, using the same format as in Figure 2. In Figure 3, we see that the one-year yield continued to respond normally to news throughout 2009 – suggesting no constraint – and was only partially attenuated by the zero bound in 2010 and the first half of 2011. Only beginning in late 2011 do we see the one-year yield’s sensitivity to news fall close to zero, indicating a severe constraint. Similarly, Figure 4 shows that the two-year yield responded normally to news all the way until mid-2011, suggesting essentially no constraint over this period.  Even from mid-2011 onward, the two-year yield has only been partially constrained by the zero bound.

Figure 3. Estimated time-varying sensitivity of one-year Treasury yield to macroeconomic announcements from January 2001 to September 2014. 

Note: The format of the figure is the same as in Figure 2. The one-year yield is only partially constrained beginning in 2010 and severely constrained beginning in late 2011.

Figure 4. Estimated time-varying sensitivity of two-year Treasury yield to macroeconomic announcements from January 2001 to September 2014. 

Note: The format of the figure is the same as in Figures 2 and 3. The two-year yield is only partially constrained beginning in mid-2011 and is essentially never completely constrained by the zero lower bound over this period.

The results in Figures 2-4 are backed up by data from other sources. Interest rate futures and options tell exactly the same story (Swanson and Williams 2014).  Forecasts from the Blue Chip survey of professional forecasters also display the same pattern. In particular, from 2009 to mid-2011, Blue Chip survey participants consistently expected that the Open Market Committee would begin to raise the federal funds rate in just 4 quarters’ time. Only beginning in August 2011, when the Committee issued an explicit statement that it expected to keep the funds rate near zero “at least through mid-2013”, did the Blue Chip forecasters’ expectations of policy tightening get pushed out further, to a horizon of 7 quarters or more.

Implications for monetary and fiscal policy

These results have important implications for monetary and fiscal policy. For monetary policy, they suggest that the Open Market Committee could have done more to ease policy between 2009 and late 2011. In retrospect, it clearly could have lowered one- and two-year yields further by promising to keep the federal funds rate low for a longer period of time. Only beginning in August 2011, when the Committee issued its ‘mid-2013’ forward guidance, do we see those longer-term interest rates fall to the point where they were more significantly constrained by the zero lower bound. Going forward, one of the lessons from the analysis above is the power of the Committee’s explicit forward guidance to influence longer-term interest rates and thereby the economy.

For fiscal policy, the results suggest that crowding out would not have been very different from normal between 2009 and late 2011. Although the federal funds rate was stuck at zero, one- and especially two-year yields were largely unconstrained during this period. As a result, the ARRA fiscal stimulus package passed in 2009 likely raised interest rate expectations in the usual way and crowded out the effects of the stimulus by the usual amount.2 This is one factor that could help explain why the economic recovery has been so disappointing, despite the very large fiscal stimulus package passed early on (see Romer 2012, Frankel 2014, and Fischer 2014 for discussions of the weak recovery and post-mortem analyses of the stimulus).


Bernanke, B S, and V R Reinhart (2004), “Conducting Monetary Policy at Very Low Short-Term Interest Rates”, American Economic Review Papers and Proceedings 94 (2), 85–90.

Eggertsson, G B, and M Woodford (2003), “The Zero Interest-Rate Bound and Optimal Monetary Policy”, Brookings Papers on Economic Activity, Spring, 139–211.

Fischer, S (2014), “The Great Recession: Moving Ahead”, speech at a Conference sponsored by the Swedish Ministry of Finance, Stockholm, Aug. 11.

Frankel, J (2014), “Guest Contribution: The Obama Stimulus and the 5-Year Anniversary of Market Turnaround”, Econbrowser, 26 February.

Mankiw, N G (2013), Macroeconomics, Worth Publishers: New York, NY.

Reifschneider, D, and J C Williams (2000), “Three Lessons for Monetary Policy in a Low Inflation Era”, Journal of Money, Credit and Banking 32 (4), 936–966.

Romer, C (2012), “The Fiscal Stimulus: Flawed But Valuable”, The New York Times, Oct. 20, Economic View.  

Swanson, E T, and J C Williams (2014), “Measuring the Effect of the Zero Lower Bound on Medium- and Longer-Term Interest Rates”, The American Economic Review 104 (10), 3154-3185.


1 We also estimate that the three-month Treasury yield was severely constrained in 2003-04, which corresponds to the Committee’s first use of forward guidance. In June 2003, the Federal Committee lowered the federal funds rate to 1% but felt that the funds rate could not be reduced further for institutional reasons (Bernanke and Reinhart 2004). Instead, the Committee began issuing statements such as “policy accommodation can be maintained for a considerable period,” which were designed to alter financial market participants’ expectations about the future path of the federal funds rate. See Swanson and Williams (2014) for additional discussion.

2 This does not imply that the stimulus was uneffective, only that its effectiveness was no greater than in normal times.



Topics:  Global crisis Monetary policy

Tags:  zero lower bound, monetary policy, Federal Open Market Committee

Professor of Economics, University of California at Irvine