I find Larry Summers’ argument in his latest Financial Times column linked above quite persuasive: that the Fed should wait until at least the beginning of 2016 to raise rates. In addition to assessing the economic situation, he points to a possible reason for bias in the Fed’s decision making that rings true to me. Larry writes:

I doubt that, if rates were now 4 per cent, there would be much pressure to raise them.

Although positive versus negative would be more salient than it rationally should be in any case, I think the zero lower bound makes a rate of zero seem more special than it really is. Knowing that the zero lower bound is a policy choice rather than a law of nature makes monetary policy choices at an interest rate more obviously like choices at a 4% interest rate. In either case, the decision should be made based on the behavior of inflation, output, and labor market measures, not on whether the interest rate seems high or low in an absolute sense. I don’t agree with Larry Summers’ take on what interest rates are likely to be once the economy has fully recovered, but his take point to the truth that there is great uncertainty about exactly what the medium-run natural interest rate and the long-run natural interest rate are. 

And since there is still in all likelihood an output gap, the short-run natural interest rate could be quite a bit lower. See my posts 

• The Medium-Run Natural Interest Rate and the Short-Run Natural Interest Rate
• On the Great Recession

Another Argument for Waiting to Raise Rates: Interest Rate Smoothing Should Be a Thing of the Past

In addition to Larry Summers’ arguments for holding off on raising rates, I have a conceptually quite distinct argument: the Fed can afford to wait to raise rates, because it can always raise rates very fast if it needs to later on.

Contrary to what optimal control models suggest, monetary policy committees around the world tend to believe the fallacy that (although events can sometimes overrule this) it is a good thing to raise and lower rates slowly. This belief shows up as policy rate movements being predictably in one direction for a long time, with small steps along the way. Optimal control models suggest that instead a policy rate should look a lot more like a random walk modified by some drift and mean reversion. That means that optimal monetary policy should have lots of reversals and dramatic movements in the interest rate when there has been relatively big news since the last meeting. 

Let me address one myth: that Mike Woodford has shown that interest-rate smoothing makes sense. I would be glad to be corrected, but I think this myth arises because Mike talked about the Fed carrying about affecting the bond markets (and more generally macroeconomic) expectations of future rates. Just as backward-looking state variables have forward-looking costate variables, bond market expectations are like a forward-looking state variable for the Fed; those bond market expectations have a corresponding backward-looking costate variable. 

As an analogy, in working toward my dissertation, I did an unpublished efficiency-wage model (which you can see and freely download here) in which, to motivate workers with an expectation of future pay making a job valuable, there is a backward-looking costate variable that can be interpreted as “seniority.”

Such backward-looking costate variables giving guidance about doing the right thing in relation to bond-market expectations contribute additional drift terms to the optimal policy rate, but it still seems to me that over a six-week span of time between FOMC meetings, the variance of news is sufficient that the effect of news should typically be substantially larger than the sum of all drift terms on the policy rate. Hence the metaphor of a muddy random walk.