Wednesday, July 2, 2014

More on Japanese inflation

I recently said that Noah Smith (and Scott Sumner in a comment on his blog here) had been hasty in their declarations of victory over deflation in Japan. Noah credits both monetary and fiscal policy, while Sumner (naturally) credits just monetary policy.

There was a large jump in the price level that started in April that I was quite shocked by. However it appears as though it is due to the increase in the VAT from 5% to 8% in April. In fact, Sumner predicted a jump in the price level because of the VAT increase almost a year ago (he warned: "Don't be fooled").

Where does this put us with regards to "Abenomics"? I'll still say things are inconclusive. And I'll say it with a graph:

The data points are the blue dots. I show the "starting point" of Abenomics along with the deflationary trend (gray dashed) and 0%, 1% and 2% inflation paths (light blue, with 2% being the "Abenomics" target). The information transfer (IT) model was fit to pre-2013 data (shown in orange) and used subsequent evolution of NGDP and the currency component of the monetary base to show where the model predicts the price level to be (red). I also show a counterfactual path without the stimulus spending during 2013 (the Keynesian component of Abenomics, assuming a multiplier of 1, i.e. ΔNGDP = ΔG) as dotted red line.

It appears as though the data is consistent with 1% inflation, the IT model is consistent with 0% inflation and the IT model without stimulus is consistent with the pre-2013 deflationary path. The post-2013 data are within the model error, though (gray band, 1-σ errors). We'll just have to see more data. However I think the rest of 2014 can potentially be decisive.

PS I made an attempt to subtract out the VAT increase (red points) by linear extapolation from the last two pre-VAT points to the first post-VAT point and subtracting the difference from the last two points. I can't say the result changes much -- the data is still consistent with 1% inflation.


  1. What do economists know about supply and demand.

    They put Q on the horizontal axis of the supply demand chart. But supply and demand are flows. dq/dt or quantity divided by time period.

    Have you ever tried to plot supply and demand from data?

    I have. It is a hard concept having been taught economics. I didnt get lines. I got points in time. Maybe I could imagine the points as the intersection of the S,D curve etch-a-sketch. But, one wonders how failed the supply and demand curve concept is.

    Hah. Even if it was one commodity it assumes every one pays the same price and sells for the same price.

    1. It's not really a single-good market for supply and demand, but the data shows "LM curves" per the ISLM model looking at the data here:

      Additionally, the model doesn't require the "law of one price"; the information signal sent from demand (ID) to supply isn't necessarily perfectly transmitted so that the information received at the supply (IS) is less than or equal to the information sent from the demand, IS < ID. This is like communication in a noisy channel without error correction, and it means the price can take on several values:

  2. Jason, O/T: you may not approve of this, but now that I've made myself into a spamming internet econ nuisance/troll posting links to your "challenge" everywhere... I feel it's best to tone it down a bit... so I didn't provide a direct link back to your blog here, but I'm curious to see if anyone asks about this mysterious new theory that generates one of your recent plots. I'll gladly redirect them if they ask... or feel free to correct me if I got it wrong (I may have oversimplified a bit).

    1. Nick followed up with a dimensional analysis question. I'll be curious to see your answer.

    2. Ha! Well you have single-handedly driven up traffic by about 25%.

      I responded to Nick. The key is that there are constants of integration so that e.g.

      N/N0 = (M/M0)^(1/k)

      So that

      P = (1/k) N/M

      P = (1/k) (N0/M0) (M/M0)^(1/k - 1)

    3. It was pretty funny as I was typing the response to Nick my step daughter was reading over my shoulder "Nuh guh duh puh".

    4. That is funny... depending on how old she is. :D

    5. (4) P=dAD/dM=1κAD/M

      What about the dimentional analys between the first P term and the second term that is a a derivative? There is a money term on the bottom the derivative term and price would be in terms of a monetary unit of account. I doubt price would be in a commodity because you have aggregated every good and seller and or buyers.

    6. "bottom the derivative term..." sould have been

      There is a money term on the bottom of the derivative term. Actual prices would normaly be in terms of a monetary unit of account.

    7. I questioned the dimentions when I saw this equation:

      P=dD/dS=1/κ D/M

      I thought if dD and dS have the same dimentions that would mean P had no dimention at all.

    8. Technically, all of the terms are dimensionless (even "Dollars" or "Yen" are technically dimensionless). Numbers of things (money, widgets) have no "units" to speak of.

      But yes, demand (D or AD) and supply (S or M) have units of money unit (e.g. dollars) in the price level equation (the price level from which we get inflation is dimensionless ... in the US it is typically set so that P = 100 in some year, say, 1980).

      In specific goods markets, demand appears to have units of "value = money * quantity demanded" and the supply has units of quantity so that the price has units of money.

  3. If the IT model predicts 0% inflation and Japan has 5% inflation in 6 months can't you just switch to saying that the central bank really is funding the deficit? Your model has 2 different results depending on if you decide the central bank is funding the deficit. But how do you really decide that? Just by what a central banker says? Other central bankers said they were not funding the government and got hyperinflation (because they really were funding the government). This call decides what your model predicts, and it seems like you just wing it.

    1. If it is your own personal call that decides the predictions of the model then the model is not really making the prediction.

    2. Hi Vincent,

      It's not so much that it is a judgment call as it is that macro data is uninformative. It is hard to say when the hyperinflation started in Argentina:

      However the non-hyperinflation solution is pretty soundly rejected.

      I'd prefer, given the uninformative data, not to switch between solutions at the slightest hint of a disagreement with data. That's why I'm particularly interested in Japan: it may show the model is wrong.

      As far as I can tell, Japanese debt is still being sold on the open market and NGDP isn't increasing relative to the size of the base leading to smaller information transfer index, so there is no compelling reason to switch to the hyper inflating solution.

    3. My mission is to understand hyperinflation as well as possible. In some sense your theory is just saying that if you monetize like crazy you get hyperinflation. But you are not helping to predict when/why that would happen.

      To me it is intuitively clear Japan has gone too far and there is no real chance of avoiding hyperinflation at this point. But I would like a theory that let me quantify this and make reliable predictions for other cases too.

      It is really the debt level and deficit level that push a country into uncontrolled monetization and hyperinflation. But debt level and deficit level are not even inputs to your models, right?

    4. If people went around trusting their intuitions, we'd still be thinking that energy wasn't quantized, that time is absolute, and that the world is flat. :D

    5. ... and BTW, good question in your final sentence.

    6. Vincent,

      You are correct that national debt isn't an input to the price level model. However, the deviations from the price level fits appear to have nothing to do with national debt. In the graph at this link, there are three major deviations from the predicted inflation rate:

      Two (rather strikingly) are in 1973 and 1979, coinciding exactly with oil crises. These are two well-defined supply shocks. The third is an unnaturally low inflation rate that happens in the 1960s for which I have no explanation (also, national debt was still high at the time in the post WWII period). None of these appear to have anything to do with national debt.

      I'd prefer to use Occam's Razor and not add in effects due to national debt until the deviations from the model warrant it. For example, I will wait for the Japan model to fail in a statistically significant way before making modifications to it, whether positing a switch to the hyperinflating solution or an impact from an outside variable. (In truth, if the model fails for Japan, I think I'd rather just wrap up the blog and do a postmortem on the model than continue making "epicycles".)

      For example, I used the fact that Australia had a bunch of US dollar denominated debt when looking at the massive deviation from the model of the Australian interest rate:

    7. If you want cycles I think you need at least ... in your equation.

      You know.

    8. Jason, you write:

      "In truth, if the model fails for Japan, I think I'd rather just wrap up the blog and do a postmortem on the model than continue making "epicycles""

      It's that attitude that attracts me to your blog in the first place! (that's a rarity in the econ blogosphere AFIK). Also, thanks for the link (that's one I'd never looked at closely) and the Occam's Razor comment.

    9. Jason, philosophical question: How strong were your priors when you first started building the ITM? How many surprises did you find when you starting working out the details and comparing with the empirical data? What were some of those surprises?

    10. Tom, regarding your philosophical question, my only theoretical prior was: any new model should reduce to known economics models in some limit (like the quantity theory, AD/AS model or ISLM model for example). I seriously doubted that economics as conducted today was based on wholly wrong-headed views.

      I do have some political priors being somewhat center left. However, that alignment doesn't have any specific economic model associated with it except some sort of vague Keynesianism. My interest in monetary economics (and what I know about it) comes almost entirely from reading Scott Sumner's blog. If you couple this with the theoretical prior, I basically think both Scott Sumner and Paul Krugman are right, if you get the ceteris paribus right.

      The series of posts starting from this one:

      through this one:

      show how I slowly figured out the model worked surprisingly well for the price level and RGDP growth.

      I was also surprised by how well this worked:

      And the single interest rate function working for both long and short term rates was pretty surprising too:

    11. Thanks, that's very interesting.

  4. ... Vincent, take a look at Jason's latest post. Check out how fitting his three parameters to the data in the US from 1960 to 1990 allowed him to fairly accurately predict the price level time series from 1991 to the present. In reality that was probably a contingent prediction: contingent on no switches to the "exogenous" solution of the underlying differential equation.

  5. Debt level, and deficit rate.
    Surplus level and savings rate.
    Stock and flow.

    Note there may be more flows to debt than deficit.

    Question is Gov. debt or suplus equivalent to private entities capital account or net worth?

    The other thing is who are debtor and creditors. Private entities do it different than the government.

    1. By rate I meant rate of flow. (In continuous time)

      If you want descreat time or accounting treatment deficit per period. Same dimentions.

  6. Note there may be more flows to debt than deficit.

    Should have been

    Note there may be more flows to and from debt than deficit.

  7. Charlie Clark plots similar data, and I ask him about it:

    1. The data I was using comes from the BoJ and corresponds to the "core-core": CPI less food and energy. For April and May, the values are the 100.6 and 100.7 respectively, which is a big jump over 98.6 in March.

      The question is whether this jump is due to the VAT increase (as I mention above, citing Sumner on it) or was generated by monetary policy.

      We'll (probably) know if we just have a one-time jump in the price level.

    2. Thanks. BTW, Charlie answered you, and did a new post (he seems interested in what else your model has to say).

    3. I wrote a belated response (I'm on travel again, so I'm only getting to check stuff at the end of the day).

      I'll have to keep watching the Japanese inflation numbers -- if the market predicts something else, this may be a good test! The VAT hike may make it less informative than I'd like, though.


Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.