Figures 1 through 3 plot the results of this analysis for three-month, two-year, and 10-year Treasury securities, respectively. The solid blue lines show estimates of how sensitive the particular Treasury yield was to news in recent years, compared with the yield’s average sensitivity in the 1990s. Periods when the blue line is close to the horizontal line at 1 in these figures represent times when that Treasury yield responded fairly normally. Alternatively, periods when the blue line is close to the horizontal line at 0 represent times when that Treasury yield was largely or completely unresponsive to news. The range of uncertainty around the yield estimates is shown by the dotted lines above and below the blue lines.
In each figure, periods when the estimated Treasury yield sensitivity is significantly less than 1, that is, significantly less than normal, are shaded yellow. If the estimated Treasury sensitivity is not significantly greater than 0, that region is shaded red. Thus, yellow shaded regions represent periods when the Treasury yield was partially, but not completely, unresponsive to news, while red shaded regions are periods when the yield was essentially insensitive to news.
Figure 1 shows that the three-month Treasury yield varied from highly unresponsive to more than normally responsive to macroeconomic news between 2001 and 2012. From the spring of 2009 through the end of 2012, the yield was either partially or completely insensitive to news. It is natural to interpret this insensitivity as stemming from the zero lower bound, since the federal funds rate and three-month Treasury yields were both essentially zero from December 2008 through the end of our sample. Thus, our analysis finds that the zero lower bound substantially constrained Treasury yields at the shortest end of the yield curve from the spring of 2009 onward, as one would expect.
What is perhaps more surprising in Figure 1 is that the three-month Treasury yield was also partially or completely insensitive to news in 2003 and 2004, a period when the federal funds rate target and three-month Treasury yield never fell below 1%. However, the Fed had lowered the fed funds rate to 1% in June 2003. At the time, a fed funds rate below 1% was considered disruptive to markets (see Bernanke and Reinhart 2004). So, rather than lowering the rate any further, the FOMC switched to a policy of managing the public’s expectations for monetary policy through language used in FOMC statements and other communications. Thus, even though the fed funds rate was not at the zero lower bound, the three-month Treasury yield behaved as if it were constrained by a floor of 1%. The fact that the empirical method in Swanson and Williams (2013) picks up this monetary policy constraint is noteworthy.
Figure 2 shows that the two-year Treasury yield’s sensitivity to news was less attenuated than that of the three-month yield. For example, in 2003–04, the two-year yield closely followed its normal tendency in response to news, implying that it was relatively unaffected by the FOMC’s implicit lower bound of about 1%. Thus, it is reasonable to conclude that monetary policy’s effectiveness on the two-year Treasury yield was essentially normal in 2003–04.
The biggest surprise in Figure 2 is how late and how little the zero bound appears to have constrained the two-year Treasury yield after 2008. Only beginning in 2011 did yields of this maturity become significantly less sensitive than normal to news. Even then, they continued being partially responsive until late 2012. Thus, to the extent that the Fed can influence monetary policy expectations over a two-year horizon, the implication is that monetary policy was about as effective as usual until at least late 2011.
Results for the 10-year Treasury yield, shown in Figure 3, are also remarkable. The sensitivity of the 10-year yield to news was never significantly less than normal over this period. Even in late 2011 and 2012, when medium-term Treasury yields were more constrained, long-term yields remained largely unaffected by the zero lower bound, except possibly in the last few weeks of 2012, when the 10-year yield’s sensitivity did fall appreciably below normal.