I haven't heard from my buddy Josh in awhile as school and responsibilities have left him no time for blogging. It's welcome to hear from him again:
"Sorry that it's been so long. School has overtaken my life, including blogging, but I still manage to read posts from time to time (so keep up the good work). Anyways, I wanted to let you know that I'm currently working on a paper that uses Granger-causality tests to demonstrate the relationship between commercial bank loans, the monetary base, monetary aggregates, NGDP and the Fed funds rate from 1971-2008. A draft version should be available soon but the general conclusions strongly support the endogenous money hypothesis.
Just wanted to point out that someone else is working on these topics."
"Recently I’ve noticed New Economic Perspectives blog author Dan Kervick repeating some version of the following question in response to any claim about the monetary base:
“What empirical evidence exists for that claim?”
"I’ve done Granger causality tests on the monetary base over the period since December 2008 and find that the monetary base Granger causes the real broad dollar index, the S&P 500, the DJIA, commercial bank deposits, commercial bank loans and leases, the PCEPI, and 5-year inflation expectations as measured by TIPS."
"The empirical fact that QE has an effect on the exchange rate of the dollar means that the Exchange Rate Channel is effective. (It’s always been curious to me how people can believe in the liquidity trap when Bulgaria, Denmark and Switzerland all successfully control their exchange rates despite being at the zero lower bound.) Does MMT address how exchange rates are set? (I honestly don’t know.)"
"The empirical fact that QE has an effect on stock prices (something which is obvious to nearly everybody) means that the Tobin Q Channel, the Wealth Effects Channel, the Balance Sheet Channel and the Household Liquidity Effects Channel are all effective. The usual MMT response is that “the markets are irrational”.
"The empirical fact that QE has an effect on inflation expectations and the PCEPI means that the Unexpected Price Level Channel is effective. The effect on the PCEPI is not easy to observe, but most people acknowledge that QE has an effect on inflation expectations. Again the usual MMT response is that “the markets are irrational”.
"The empirical fact that QE has an effect on commercial bank deposits, and loans and leases, means that the Bank Lending Channel is effective. This is particularly significant in debunking MMT (as well as Accomodative Endogeneity) since they claim that this is completely impossible."
"But here’s by far biggest problem with Dan Kervick’s question."
"I’ve read some of the research published by MMT economists such as Randy Wray (e.g. “System Dynamics of Interest Rate Effects on Aggregate Demand”). Mostly it consists of constructing stock-flow models using parameters derived from “neoclassical” economics estimates. I’ve never seen a single empirical MMT research paper, or even any evidence that they have any knowledge of econometrics at all.
"Sorry that it's been so long. School has overtaken my life, including blogging, but I still manage to read posts from time to time (so keep up the good work). Anyways, I wanted to let you know that I'm currently working on a paper that uses Granger-causality tests to demonstrate the relationship between commercial bank loans, the monetary base, monetary aggregates, NGDP and the Fed funds rate from 1971-2008. A draft version should be available soon but the general conclusions strongly support the endogenous money hypothesis.
Just wanted to point out that someone else is working on these topics."
Here is his blog though as he says he hasn't been able to keep it up lately.
Still, that he is writing a paper on the MB and it's relation to NGDP is very welcome news. It's funny that he's going to use Granger-causality tests as this was what Mark Sadowksi invoked the other day to justify the claim that the tripling of the MB in 2008 and the fact that we subsequently saw no effect should be put aside as he's proven that MB Granger-causes all kinds of things-like a rising stock market.
Here was Mark Sadowski the other day during this debate on whether a doubling of the MB would lead to a doubling of NGDP and-'almost' doubling the price level:
"Recently I’ve noticed New Economic Perspectives blog author Dan Kervick repeating some version of the following question in response to any claim about the monetary base:
“What empirical evidence exists for that claim?”
"I’ve done Granger causality tests on the monetary base over the period since December 2008 and find that the monetary base Granger causes the real broad dollar index, the S&P 500, the DJIA, commercial bank deposits, commercial bank loans and leases, the PCEPI, and 5-year inflation expectations as measured by TIPS."
"The empirical fact that QE has an effect on the exchange rate of the dollar means that the Exchange Rate Channel is effective. (It’s always been curious to me how people can believe in the liquidity trap when Bulgaria, Denmark and Switzerland all successfully control their exchange rates despite being at the zero lower bound.) Does MMT address how exchange rates are set? (I honestly don’t know.)"
"The empirical fact that QE has an effect on stock prices (something which is obvious to nearly everybody) means that the Tobin Q Channel, the Wealth Effects Channel, the Balance Sheet Channel and the Household Liquidity Effects Channel are all effective. The usual MMT response is that “the markets are irrational”.
"The empirical fact that QE has an effect on inflation expectations and the PCEPI means that the Unexpected Price Level Channel is effective. The effect on the PCEPI is not easy to observe, but most people acknowledge that QE has an effect on inflation expectations. Again the usual MMT response is that “the markets are irrational”.
"The empirical fact that QE has an effect on commercial bank deposits, and loans and leases, means that the Bank Lending Channel is effective. This is particularly significant in debunking MMT (as well as Accomodative Endogeneity) since they claim that this is completely impossible."
"But here’s by far biggest problem with Dan Kervick’s question."
"I’ve read some of the research published by MMT economists such as Randy Wray (e.g. “System Dynamics of Interest Rate Effects on Aggregate Demand”). Mostly it consists of constructing stock-flow models using parameters derived from “neoclassical” economics estimates. I’ve never seen a single empirical MMT research paper, or even any evidence that they have any knowledge of econometrics at all.
What empirical evidence exists for MMT? None."
Sumner often has tried to disparage me personally when I've tried to engage him on his blog by claiming that since I'm not an economist I should just let him talk I guess and say nothing other than shake my head in agreement. He's often said since I don't understand economics I shouldn't write this blog which shows that he is certainly not very good at psychology-no wonder he hates Keynes so much.
I don't even have a problem admitting that's there's plenty in economics I don't know though I'm determined to learn. I'm not as clueless as Sumner wants to claim but I know very well I don't know everything far from it. I'm just not quite so naive or so stupid as to look to him to learn these things.
I've felt that I've often felt like something of a voice in the wilderness on the need to engage Sumner's ideas-I admit engaging him is not for everyone. Still the ideas must be engaged. As for what I don't know about economics, let me admit that I could use some elucidation on exactly what Granger-Causality is. I'm aware of it, and know that it's the product of the Nobel Prize winning economist, Clive Granger.
I know that is has many great uses in figuring out true causality in dealing with correlations and isolating them from noise. However, he himself said when he received his reward that GC has been used in all kinds of 'ridiculous' ways.
So Josh, if you have time, maybe you could elucidate on GC a little. What are legitimate and illegitimate uses of it? I know for starters, that Granger-Causality is not causality itself. If we say that X Granger-Causes Y this doesn't necessarily mean that X causes Y.
In trying to get at the truth someone with the formal training you have but that I lack will be very helpful in getting there.
P.S. I think that I have a natural feel for economics-I certainly I'm fascinated by it, but no doubt I'm an intuitive thinker-for me I kind of think that thinking intuitively about it and thinking as a trained academic economist do are kind of a relation like that between a piano player who reads sheet music and one who plays by ear. In piano I can actually do a little bit of both. In economics I'm mostly intuition though I want to understand the academic side adequately-to make sure they don't get over on me for one thing!
The reason why economists prefer models to intuition is intuition can't be repeated but models work for everyone even those with little intuition.
Mike,
ReplyDeleteThanks for the post and support! Before I continue I should point out that I'm still a novice in using GC, but will try to elucidate its uses anyway.
As you correctly point out, Granger-causality tests (similar to any statistical test) cannot prove actual causality. GC basically looks at whether the lagged values of variable X improve the prediction of the current value of Y. In other words, GC demonstrates that changes in X not only occur prior to Y but hold predictive value for Y. Since any hypothesis of a causal relationship implies that X occurs prior to Y, a GC relationship supports a causal hypothesis.
As I see it, a significant issue with GC results is the impact of expectations. For example, it has been shown that stock prices Granger-cause earnings per share (X --> Y). In spite of this result, most people would not argue that stock prices cause changes in earnings. Instead, a typical response is that expectations of changes in earnings cause changes in stock prices, resulting in the statistical anomaly. Most people therefore maintain the hypothesis that earnings cause changes in valuation (Y --> X).
As I mentioned in a separate reply to Mark Sadowski, I think its conceivable that the relationship could be spurious for other reasons. We'll have to await his response on that subject.
Lastly, the paper is also most ready (or at least I thought so). I just read Mark Sadowski's comment about not first differencing the variables in order to obtain accurate results. I have to admit that my paper uses first differences since that was the method I learned in school, saw used most frequently in published articles and read in my textbook. While I don't yet known enough to outright disagree with Mark on this subject, I am skeptical that such an egregious error would not only occur in countless publications but not elicit any strong objections (a point made by Mark). Given how far along the paper already is, I doubt I will make the suggested modifications now, although I will certainly consider his suggested method going forward (and as a robustness check of the results).