The 10-year treasury

There has been some talk about the 10-year T-note in the business news, so I've updated this graph (here or here) [1] with some of the latest data:

The 10-year is basically where we'd expect it.

...

Footnotes

[1] I did change the weights to weigh the effective fed funds rate instead of for the 10-year T-note in the fit which gives a better description of recent data for both rates -- hence the observable difference in the curves. Basically, this is fit to the EFF and the same coefficients are used for the 10-year rate. Without this change we have this (not a huge difference):

Maybe we should fall back on the least informative prior ...

This made me laugh out loud (H/T Mark Thoma):

Max Burton-Chellew ... said: 'Game theory can be used to predict how a self-interested and rational person will behave in social situations. However, economic games, in which people have to make decisions on how to allocate money to themselves and others, have consistently shown that these predictions fare poorly. In particular, it seems that people are overly generous and altruistic, and appear to be primarily motivated by concerns of fairness rather than maximising income. ...' 

This has led to the conclusion that there are different social-types of people that can be reliably classified in the lab, and that most people care about fairness so much that they are even willing to forego their own success to help others. These people are known as "conditional cooperators", and the results primarily come from one type of experiment called the public-goods game ... 

Dr Burton-Chellew said: 'We found that many of the so-called conditional cooperators are confused and do not seem to understand the public-goods game, appearing to think that being generous towards others will make them money. We primarily demonstrated this by having them play with computers, which cannot benefit from this cooperation, and showing that people behaved the same way regardless. ... '

My emphasis. As I've said before, the least informative prior for the result of the public-goods game is a 50/50 split. You will choose between 0 and N tokens to contribute to the public; a maximum entropy (least informative) prior leads to the prediction of N/2 ... regardless of whether you are playing with computers or humans. Utility maximization, game theory and evolutionary arguments lead to incorrect conclusions. Humans are confused and don't understand what is happening -- how can you expect rational utility maximization to be happening? How can you presume intelligence (as Paul Krugman does)?

Another win for maximum entropy.

Psych!

The last bi-weekly base data seems to have been a bit of a fluctuation (as was hinted at earlier), but there is still a downward trend per the prediction ...

We should begin to know more about the rate we'll approach equilibrium empirically -- not predicted by the information equilibrium model -- by the end of February.

Maybe you should question your assumptions?

Alex Tabarrok points us to an illustration of what can go wrong with people who think they are using Bayesian reasoning -- and with the strange drive to produce counterintuitive results.

In an excellent new paper, "Too Good to Be True", Lachlan J. Gunn et al. show that more evidence can reduce confidence.

Emphasis in the original. They use a police line-up as their example.

Say you have a police line-up and ask witnesses to identify the suspect.  Assuming the prior probability of guilt is 50%, independent witnesses and a "probability of systemic failure" pc for all the witnesses (i.e. the suspect was a clown, but only one of the people in the line-up is dressed as a clown) ... they find that additional positive identifications should lead to less confidence in the result. Here is a graph:

What is happening here is that you should suspect "systemic failure" given the conditions above -- you can tell because it primarily depends on pc. That was our easy out (only option) given the conditions above [1].

But that neglects the other possibilities -- including one big one. The witnesses are not independent in any way. They saw the same thing -- they are correlated by the event motivating the line-up itself.

Imagine you get a group of "witnesses" who go in a room where someone flips a coin. They come out and you ask what the result of the flip was. If 100 people say it was heads, you have to conclude (with suitably high confidence): the people are conspiring (systemic failure), the coin was double heads (systemic failure), or unfair (prior not 50%). But it could also be that the witnesses all saw the same coin flip (or 90 of them did and then 10 saw a different flip that also came out heads).

Basically the paper shows more evidence can reduce confidence in the evidence or experiment if you hold onto your assumptions. If something is too good to be true, maybe you should question your assumptions -- and not chalk it up to conspiracy or incompetence.

Footnotes

[1] There is also the issue of identification accuracy, but your sample of witnesses might not have that distribution of accuracy for whatever reason. This goes into the prior model anyway, so doesn't impact the main thesis here that you should also question your priors ... and not just systemic failure.

Paul Krugman's definition of economics assumes the form of the theory

Assume a can opener ... [wikimedia commons]

I was reminded of Paul Krugman's piece on evolution in an especially good comment here (I wrote about this subject here):

Caution, however. [David Sloan Wilson] himself is considered heterodox within evolutionary biology, being a vocal and triumphalist advocate of group selection and multi-level selection. Mainstream evolutionary biology is still largely individual-selectionist or gene-level selectionist (their equivalent of methodological individualism). So we have a case of a heterodox biologist egging economics on to heterodoxy.

I'm not going to re-hash the evolution argument (which, as with any methodology, is always a question of "what is it good for?"). I do think discussion of Krugman's definition of economics is worthy of a post, though. Here's Krugman:

Let me give you my own personal definition of the basic method of economic theory. To me, it seems that what we know as economics is the study of those phenomena that can be understood as emerging from the interactions among intelligent, self-interested individuals. Notice that there are really four parts to this definition. Let's read from right to left. 

1. Economics is about what individuals do: not classes, not "correlations of forces", but individual actors. This is not to deny the relevance of higher levels of analysis, but they must be grounded in individual behavior. Methodological individualism is of the essence.
2. The individuals are self-interested. There is nothing in economics that inherently prevents us from allowing people to derive satisfaction from others' consumption, but the predictive power of economic theory comes from the presumption that normally people care about themselves.
3. The individuals are intelligent: obvious opportunities for gain are not neglected. Hundred-dollar bills do not lie unattended in the street for very long.
4. We are concerned with the interaction of such individuals: Most interesting economic theory, from supply and demand on, is about "invisible hand" processes in which the collective outcome is not what individuals intended.

Emphasis in the original. The problem is that Krugman essentially assumes the form of the solution to the problem. Talk about assuming a can opener.

1. We do not know for a fact that there is little dimensional reduction in macroeconomic theory. We also do not know for a fact that all macro statistical regularities directly relate to micro agent parameters -- this presumes there are no macroscopic "entropic forces" that arise from properties of the distributions of agents. Essentially, this kind of approach, if applied in thermodynamics, would not correctly describe the stickiness of glue or osmosis ... because of an assumption. I think there is a lot of dimensional reduction and micro agent parameters tend to have little importance (these are kind of the same things) -- but I don't know for sure.
2. This oddly assumes both a) aggregation of self-interested behavior has macro consequences, and b) aggregation of deviations from self-interested behavior has no macro consequences. It's especially odd since we don't know the answer to either question either way. No one has successfully aggregated agents that successfully describe a macroeconomy empirically where the details of the agents matter.
3. Phase space added to an ideal gas is quickly occupied. Is the gas intelligent? A model where humans are randomly wandering (dither) through an economic state space space predicts a certain occupation of economic states (grabbing those 100-dollar bills). Maybe real humans occupy those states faster than dither would predict. That would constitute an empirical test of the role of intelligence. As yet, we have no answer ... so why assume one?
4. Supply and demand can arise from properties of the opportunity set (economic state space) alone with random ("irrational") agents. The invisible hand seems amenable to treatment as an entropic force. Why assume it comes from the detailed parameters in the interaction of agents?

So you can see -- Krugman's definition of economics presumes a particular form for the answer. My opinion is that Lee Smolin's definition is better, and doesn't presume as much:

[Statistical] economics is the study of the collective behavior of large numbers of economic agents.

I'd like to give an even better definition:

Economics is the study of simplifications in the collective behavior of a large number of agents.

The invisible hand is just such a simplification. I think interest rates, inflation and output might be similar simplifications -- leading to real predictive power in economic theory.

And I'd add: The study of the complexities in the collective behavior of a large number of agents is sociology.

...

Update 1/21/2016

This result is hard to square with Krugman's assumption of intelligence:

We found that many of the so-called conditional cooperators are confused and do not seem to understand the public-goods game

Updates to some ongoing forecasts

I've updated some ongoing predictions, mostly this core CPI inflation prediction (new data since prediction in between vertical gray lines):

I also have a new graph for Japan's core CPI (new data since prediction in dark green):

Nothing earth-shattering here, just maintenance.

Time for a trip to Vancouver for a little fiscal stimulus ...

There was some reaction to the fall in the price of the Canadian dollar (e.g. here and here). According to the information transfer model, forex markets are basically highly volatile relative NGDP futures markets (even though they might not know it). This means we might expect a big fall [1] in Canada's NGDP when Q4 2015 and Q1 2016 data come out ...

Update 1 December 2016

Turns out to have been a good forecast.


Footnotes:

[1] Or a big rise in US NGDP, but that seems questionable.

Assuming answers to complex integrals

The basic problem with a lot of discussion of economic methodology is that many people (e.g. [1], [2]) assume they know the answer to this:

$$
\text{(1)}\;\;\; \int dE \;\;\;  \left( \hat{O} | H. \;\; economicus \right) \stackrel{?}{=} \int dE \;\;\; \left( \hat{O} | H. \;\; sapiens \right)
$$

where $E$ is an economy and $\hat{O}$ is some observable operator. The integral is just a notational way of representing an aggregation problem using a given agent model. But we don't know the answer. No one does. Anyone who claims an answer is probably just assuming an answer.

On this blog, I make a case that most of the time

$$ \text{(2)}\;\;\; \int dE \;\;\; \left( \hat{O} | H. \;\; atomicus \right) \simeq \int dE \;\;\; \left( \hat{O} | H. \;\; sapiens \right) $$

where $H. \;\; atomicus$ is an even simpler creature that has no brain whatsoever and simply randomly walks into an office, receives money, and then bumps into products and buys them.

And whenever

$$ \text{(3)}\;\;\; \int dE \;\;\; \left( \hat{O} | H. \;\; atomicus \right) \neq \int dE \;\;\; \left( \hat{O} | H. \;\; sapiens \right) $$
it is generally bad news.

I did show that under certain conditions:

$$ \text{(4)}\;\;\; \int dE \;\;\; \left( \hat{O} | H. \;\; atomicus \right) \simeq \int dE \;\;\; \left( \hat{O} | H. \;\; economicus \right) $$

But at the end of the day there is no answer to equation (1).

Possible slower adjustment for monetary base

Later this week, we'll get the new bi-weekly monetary base data to compare with the prediction, but the source base (that the seasonally adjusted bi-weekly number closely tracks) has gone back up [source]:

What's with the pessimism?

From Wikimedia Commons.

I remember a conversation in grad school with one of the postdocs in our department about some new result in biology over the course of which he called me a "hardcore reductionist" because of something I said that was rather reductionist. However, I'd really just characterize my view as optimistic about figuring out stuff.

A Twitter conversation with Eric Lonergan (see here) about eclecticism in macroeconomics reminded me about my optimism; I realize many people do not share it.

Eric brought up Hume's "uniformity of nature" -- the basic assumption that goes into any quest for universal laws -- suggesting it might not exist in economics. I agreed it was possible -- and always should be in the back of your mind lest you start seeing pattern where none exists. But do we know that for certain? No.

While Eric's view is closer to a case that we should view various economic models as approximations to a (potentially unknowable) theory, Dani Rodrik's push for an eclectic interpretation of macroeconomic modeling suggests a kind of resignation. He says we should give up looking for a big unifying theory of economics because it does not exist. Rodrik's reasoning must be that he and his colleagues haven't found it, therefore it must not exist.

And (via Mark Thoma) I read Daniel Little:

... Does the phenomenon of [social phenomena] admit of a scientific treatment along the lines of Galileo, Newton, or Lavoisier? 

The answer is resoundingly no. Such a goal displays a fundamental misunderstanding of the social world. Social things and processes at every level are the contingent and interactive result of the activities of individual actors. Individuals are influenced by the social environment in which they live; so there is no reductionist strategy available here, reducing social properties to purely individual properties.

This commits two logical fallacies. The current lack of  a scientific treatment does not mean one does not exist. And the lack of a reductionist strategy does not mean there is a lack of any strategy.

As I commented on Little's blog, even his specific example of not being able to begin to understand a city is contradicted by the existence of work by  M. A. O. Ayeni. From the abstract:

For these uses of the system approach, the concept of entropy, introduced from both thermodynamics and information theory, plays a significant role. ... It emerges that the concept of entropy can be used in studies of urban spatial structure as an integrating concept provided that our terms are defined explicitly and unambiguously.

Apparently Little already knows the final result of this research program. However given that he thinks reduction to agents is the only strategy, he probably hasn't considered maximum entropy approaches.

Now I can understand the pessimism in each case. In economics, we seem to be at the end of a failed research program (microfounded rational utility maximization) that began in the 70s. Pessimism is a natural outcome. And in social science (writ large), there is a bias towards consequential human action rather than mathematical laws. I happen to think there is a lot of room for the latter, but that's because I'm pessimistic about human free will.

But even without the information transfer framework for economics, I'd still be optimistic about finding mathematical explanations of most statistical regularities.

The truth is: we don't know. Lack of a big mathematical theoretical framework (or lack of a successful one) is not evidence one does not exist.