Monday, September 2, 2013

Introducing FedSim

How did you spend your Labor Day? Since it's been grey this whole weekend here in N.J., I spent mine on a labor of love.

My Wonkblog colleague Ylan Mui published a post on monetary-policy games on Friday. I liked her post, but all of the simulation games were in fact pretty unrealistic. So I decided to see if I could outdo the Federal Reserve Bank of San Francisco and the European Central Bank: I built a monetary-policy simulation game from the ground-up and realistic enough to give players an accurate impression of the challenges of central banking.

And, after quite a bit of work, here is the result: FedSim. You can download it here for free from Dropbox. (It's a small file at 400 kB.) There are extensive instructions in the game itself, but basically you get to manage monetary policymaking for a decade (2013 - 2023) and the software updates a set of key macroeconomic indicators following your policy choices. The game is designed for use in the classroom -- ideally, to give economics students a fun way to learn about monetary policy.


What's in the game? Noisy measurement. Supply shocks and demand shocks. Sticky prices. Downward nominal rigidity. A short-run Phillips Curve and long-run inflation expectations. NAIRU. Monetary policy that operates with a lag. All this I've realistically calibrated to match the U.S. economy.

As for indicators, you get the unemployment rate, year-over-year headline and core PCE inflation, quarterly annualized percentage growth in real and nominal GDP, the monthly change in payroll employment, and the yield on the 10-year Treasury note.

If you've been reading the blog lately, you'll recognize that the game ties into my growing interest in modeling and quantitative analysis. I built the program in Excel with Visual Basic macros, which was a dumb choice but, hey, it's what I felt most comfortable doing.

I learned a lot of coding, for sure, in writing the program. But more importantly, having to calibrate a realistic model of the U.S. economy taught me a lot about the U.S. economy, and I'll have a post coming on that shortly.

If you play the game, take a screenshot at the end and send it to me on Twitter, by email, or in a comment on this blog post. If you're an educator looking to use this game, you're welcome to do so, and I'd love it if you could send me an email letting me know. If you're a game developer, send me an email, because there's a lot I couldn't do in Visual Basic, and I have a much bigger game in mind.

13 comments:

  1. This is awesome. Way better than the SF Fed.

    I made it to 2016 before leaving the ZLB. Hit the ZLB again in 2020. In between, the unemployment rate was below 4% for 2+ years, at the cost of inflation expectations, which had previously anchored around ~1%, rising to ~3%. Inflation remained around 3% while the unemployment rate started rising in mid-2018, peaking at 7+% in 2020. UNR was stuck there until the end of my term. Same with inflation, around 3%. Stagflation! :(.

    Unfortunately for my priors, my performance became worse the more that I paid attention to NGDP growth.

    Total score: 66. Biggest weakness was the high variance of the FFR, my score was 33 in that category.

    Macro data: http://i.imgur.com/cvyPsDx.png

    ReplyDelete
    Replies
    1. Thanks for playing. My advice (for playing the game) is that you didn't seem to pay enough attention to the lags. That's why you had to rocket interest rates up in 2017. I think you would have been OK with gradual (i.e. 25 basis point every other month) hikes in 2014 and 2015. Inflation expectations would not have been much lower, but the drop in unemployment would have been slower and more sustainable and controlled, so that you didn't need to throw the economy into stagnation in 2018.

      Delete
  2. This is great. You should add forward guidance and/or QE in version 2. Cheers

    ReplyDelete
    Replies
    1. Yes, I was thinking about how to do this. I figured out something with QE. But I'm still puzzling over how to model forward guidance. Ideas?

      Delete
    2. Strikes me the most obvious way to do forward guidance would just be to have input fields for the funds rate in future periods, affecting expectations. (I don't know how much expectations are in your modelling at the moment?)Though I guess you'd have to model in some 'credibility' effect so that the forward guidance isn't effective if you consistently diverge from it.

      Delete
  3. The good news is that inflation doesn't seem to significantly affect economic performance. I increased the fed funds rate in measured steps from 2016-2020, topping out at 5.25%. Still, despite 45% headline and core inflation, unemployment dropped to 0.8%!

    I got a zero on the inflation target and a 1 on inflation and unemployment stability (though unemployment simply kept edging slowly downward), a 91 on unemployment, and an 83 on interest rate stability for an overall of 35.

    ReplyDelete
  4. This was fun! I got 73 on my third try. My lowest sub-score was 67 for interest rate stability, and this was mainly the result of laziness - I preferred big changes to typing small changes over many rounds. The key is indeed the lags, and understanding NAIRU - although I chose to look at it as not caring about inflation until unemployment was at the friction rate (which I think is 5.5).

    Variables that I would add: fiscal policy; savings rate and investment (domestic and foreign); population and working force; trade-weighted exchange rates and terms of trade; income distribution. If you add all this, you can calculate median living standards and add them as part of the score (after all, if everyone is employed, inflation is low, but living standards are dropping, this is hardly a success: I think this was what happened with my simulation).

    If you really want to go crazy, you can add asset prices and bubbles.

    ReplyDelete
    Replies
    1. Yes, I'm trying to encourage people to play like a real-world central bank by giving them a harsh penalty for large-magnitude changes in the policy rate. With a lot of practice, I've found that I've gotten very skilled at timing the policy lag to work out in my favor, but yes, it is easy to get crushed that way. And the fact that I made the inflation penalties for going under NAIRU nonlinear also doesn't help. Don't tell anyone, but the implied NAIRU in the model is 5.3 percent. And thanks for your suggestions. I don't quite know how I would model all of this -- particularly the income distribution, which frankly I think the evidence that monetary policy has an impact on it is young, although of course that evidence is interesting to me.

      Delete
  5. I've only played once and deliberately went crazy... at one stage I had unemployment at -4%, which was interesting! Also after I'd been changing the funds rate wildly for a while it seemed to just throw a hissy fit, and unemployment and both inflation measures locked at zero and wouldn't change whatever I did. So I guess my advice to the new Fed Chair would be to not vary rates between -300% and +200%

    ReplyDelete
    Replies
    1. Ack, yes, I tried to put in a fix that blocks below-zero and above-100-percent unemployment, but it fails if you really hit the gas or slam on the breaks. Then the program itself fails, which is why you got zero.

      Delete
  6. I ran the simulation 100 times using the Mankiw version of the Taylor rule (via new macros), the median score was 79 and the 20th-80th percentile range was 75-82. The vast majority of the time (over 80%), the lowest score came from the stability of interest rates factor--it might be possible to stabilize this a bit by adding some 'memory' of the direction of interest rate change to the 'rule' (e.g., not letting the interest rate fall if you have most recently raised it unless the 'rule' says you should be more than 50 basis points away from current policy, sacrificing a bit of 'accuracy' in the name of stability).

    ReplyDelete
    Replies
    1. Awesome idea. If you might send me the code, I'd love to see it.

      Delete
  7. Nice work! I had a rough time of it my first go, but I'm going to keep at it. Thank you!

    http://www.consciencewarrior.com/2013/09/fedsim-by-evan-soltas.html

    http://4.bp.blogspot.com/-hafZ060p8vk/UkcpKJ9rERI/AAAAAAAAAaE/APFDHOo_WS0/s1600/fedsim1.png

    ReplyDelete