My market-driven nominal GDP forecasting system says that markets expect something like 3.8% nominal growth in the next year.
That’s the average forecast from the 10 best models I could build, they take into account changes in asset prices over the last year (lags going back 3 and 4 quarters proved the most accurate, out of sample). If assets became overvalued, it happened over a year ago.
To my mind, that’s pretty good evidence that financial markets as a whole haven’t gotten ahead of the nominal economy.
Markets are probably not greatly undervalued either. The Fed’s likely to keep nominal growth going along at 3.0%-4.0% indefinitely.
Personally, I’m buying stocks with any free cash that I get.
The standard neoclassical economics line is that there are fundamentally no constraints on human population levels. As long as technology advances, new resources can be tapped and those we already have can be better used. Because more people means more ‘idea generators’, technology automatically increases with the population level. The idea is that the more desperately poor peasants in Bangladesh, the greater the odds that one of them will invent cold fusion. Let’s say I’m not convinced.
I can however imagine that if you could go back in time to the 1970s and magically boost the birth rate in industrial countries by 25%, while maintaining the realized immigration flows, that current GDP per head would be about what it is now. That is, today’s developed economy is not all that Malthusian.
BTW, in case you weren’t in the loop, an economy is ‘Malthusian’ when birth rates equal death rates. When the horse collar is invented, food output rises (GDP) followed shortly by higher birth rates, which gobble up the overplus and bring GDP per head back to the subsistence level. To get higher living standards, you need a higher death rate or a lower birth rate. More nuns or more wars.
There is a long run (centuries) tendency for our not-so-Malthusian economy to revert to its natural Malthusian state. In a land where the population is below the subsistence-level carrying capacity, alleles that favor high fertility will increase in frequency (say genes that make you less likely to use contraception). That we have low fertility in industrial countries is just because our genes are ill-configured for modern abundance and the welfare state. In the environment of the late middle ages, which industrial populations are more or less genetically programmed to live in, working long hours to buy stuff was adaptive, so today’s middle class focus on buying new cars and drinking fancy beer instead of whelping the vast hordes of unwashed, underfed urchins that could technically be funded by a modern income. It’s biological madness! We’re like the house cat that won’t drink still water, a behavior that was adaptive in an earlier environment lingers on when the environment radically shifts. Selection is of course at work remedying this unnatural situation, we’ll have to wait a few generations, but fertility should steadily climb back in the industrial world so long as we have so much extra food.
A return to the true Malthusian dynamics that ultimately govern all living systems is for the long run though, ‘when we’re all dead’. The world we live in now is not immediately Malthusian. Still, it seems to me that some markets are Malthusian, even if the economy as a whole isn’t. It also seems that the number of times Julian Simon’s name is dropped has little—or even no—effect on the supply of said goods.
I’d define a good as functionally Malthusian if a big increase in demand for said good would not be followed by a meaningful and sustainable increase in supply.
Here are some examples of Malthusian Goods:
Wine recovered from the Titanic
Housing in San Francisco
Highway space in the blue states (i.e. where you’d actually want to live)
Admissions to Yellow Stone National Park
Bayreuther Festspiele tickets
It’s no accident that it’s common for Russians (just richer than Mexicans by GDP/head) to have family summer cottages, while the Dutch (high GDP/head) live on boats: European Russia has low population density, and the Netherlands is plainly overpopulated.
There are some goods the market just can’t make. The supply of shoe box condos in Singapore can easily respond to any conceivable increase in demand, but you need a secluded lake to build a secluded lake house. Likewise, fish stocks can be managed, but once the optimal management policy is in place, supply is set by the weather. And I’m sorry to say that farmed fish is an abomination and hardly a substitute for the real thing.
Policy and cultural failings can also turn situations that need not be zero sum into undignified Malthusian messes. Highways are a good example. The number of drivers in America and has grown rapidly since the interstate system was built in the 1960s and 1970s, yet highway capacity has not kept pace. Environmental concerns, greedy unions and dysfunctional government are more to blame than a shortage of concrete and asphalt, but the constraint remains. Every new driver on the roads means slightly worse traffic. This good is functionally Malthusian, we don’t have the social capital to do anything about it, and we’re not likely to in the future either. The same can be said of water supplies in California and Texas. Might we recognize our policy impotence in this area when making decisions in domains where the policy space is bigger?
I’m not saying it’s the end of the world if we all have to live in Hong Kong style cities. My point is just that there are many aspects of the good life that can only be broadly enjoyed by a population if density is low. This isn’t talked about enough and economics blog readers would be well-served to grapple with the implications.
I’ve been working on some immigration posts between flights and marathon Stata sessions. Maybe I’ll even get them up. In the meantime, consider this Bloomberg headline, presumably written in all seriousness: U.S. Will Have Something Other Countries Want: A Big Labor Surplus.
Who wants a big labor surplus? Zuckerberg? Iceland and Finland have dear labor and those societies seem to be getting along alright. For all the fretting about Japan, unemployment is low and living standards are high.
I want labor shortfalls to rescue the labor share of GDP. I especially want tight supply in lower-skilled labor markets, so that people who work with their hands can maintain a better relative consumption bundle (lower inequality) as the economy robotifies.
GDP’s important but it’s only part of building a healthy society. Making a country a nice place to live and looking out for the welfare of citizens is the goal of economic policy, not maximizing the total level of GDP.
In the end of March, I put up a new charting system at http://www.efficientforecast.com as well as completely overhauling the back end.
The system is no longer plagued by unsychronized prices (blindly scraping at 11:55 and treating everything as having an 11:55 time stamp). My scraper now pulls the time stamp for each price, loading it into a minutely frequency table. All prices for a given three minute span need to be filled in or interpolated before the system will pull principal components and update the forecast. Each forecast is now accurate to within three minutes of the reported stamp.
I’ve also moved operations to an Amazon EC2 instance (a server hosted by Amazon). This means I’m no longer at the mercy of Pennsylvania’s second world electrical infrastructure, or unreliable internet . I can’t recommend Amazon web services enough. It wasn’t easy to setup, but it was a lot easier than any of the alternatives would have been.
Here’s a screen shot of the current graph:
Note the vertical bar in the graph’s middle. This shows the switch from one quarter to the next. The system isn’t trying to forecast NGDP exactly 365 days from today, but rather the figure the BEA will publish 4 quarters hence. Thus there is a discontinuity in the forecasts when we go to a new quarter. Think of the vertical bar as separating two contracts, one that traded until the end of the quarter, and one that took its place after the quarter’s close. The two series are in someway related, but are ultimately measuring NGDP expectations for different quarters.
If you view the forecast at a one or five day window, you’ll get it at 3 minute interval frequency. This is how you’d want to view it if you were curious about a specific event (say a Fed speech). Wider windows will cause the graph to shift to a half hour frequency. As the time series grows, I’ll need to figure out how to deal with the number of observations. I have some ideas here.
Sorry to my Japanese readers for not adding the Japan system like I said I would. However, it took a while (9 months) to figure out how to fix all of the issues with the U.S. system that popped up last year, and to teach myself how to do the web dev work as I’m essentially solo on the project now. I’m please with it now and not really itching to make any major changes to the econometric side just yet. I still think I’ll make time to do Japan and the UK this year.
Maybe I’ll put a post up about how the new forecast averaging system works. The gist is that I searched a massive model space and then did rolling, year ahead out of sample forecasts to evaluate the models. I ranked the models by average accuracy and forecast error variance and took the top 10 models. The systems thus runs 10 forecasts every three minutes and averages them. The system is actually not that sensitive to whether I take the top 20 or top 30 models, but 10 seems a worthy number.
I’m also putting together an FAQ to put on the site. Emails justinpirving [at thing] gmail.com or comments on that topic are most welcome.
1.Nothing against the second world, we in quasi dysfunctional countries got to stick together
Lars Christensen had a nice post up a few days ago on the causal relationship between NGDP and RGDP.
Lars’ post reminds me that shockingly few people who work in the greater macro economics field (especially in the financial world) have any better than a ‘Financial Times’ level understanding of the world. Most think of “GDP” as being 1. RGDP and 2. either driving inflation or, oddly, somehow decoupled from inflation all together. When NGDP is even considered, its thought of as some side effect of inflation and output. Few come out and put it like this but it’s typically lurking under the surface.
If you read Scott Sumner’s golden age posts from 2009-2010, you will find it hard to think of NGDP as—to use Lars’ wording—the quasi residual that so many treat it. The causal direction is obvious you think about it. How is buying and selling is done? The buyer goes to the seller, gives them cash, this is nominal spending. No cash, no sale. The seller hands over a real good or service, this is the real production. Note how it comes after the nominal spending—causality. Spent cash begets real production.
I don’t go to my woodlot-raised-swine dealer and say “what inflation factor would you assign to 30 2005-dollars worth of hamhocks?”, I ask him “how much for hamhocks?” and handover the green paper. The green paper spurs the farmer to breed more of his sows and/or raise prices. The nominal spending kicks it all off.
I read The Storm of War in 2012, a well written, uppermiddle brow overview of the Second World War. The book has a graph in it showing how allied bombing was able to halt growth in German war production in 1944. This is shown by way of a war production index, and when I saw it I remember thinking that it’d be tricky to come up with a useful weighting rule for the components of such an index. What if tank production grows 100% but 88mm shell production falls 50%? How to you convey in an index that there are twice as many tanks but fewer shells to fire from them? Moreover, how do you quality adjust tanks when designs change or when they’re built with substandard alloys (as was the case with most German armor after Kursk)? I mulled the problem for a while and concluded that there’s probably no “right” way to do it but maybe the ambiguities wash out over a wide enough range of production items. It then came to me that this index was just RGDP, for a subset of an economy.
RGDP is just a production index.
It’s useful to have production indices. In the case of RGDP a production index tells you how ‘big’ the economy is, which is something we want to know. Production indices (think industrial production or manufacturing production) are understood to be measuring factor variables. I suspect that fewer people would think of RGDP as setting inflation/NGDP if real output were reported as an index instead of being scaled into “constant dollars”.
From what I understand, it might be right to think of RGDP as determining NGDP in planned economies like the Soviet Union’s. The state picks production goals for all goods in the economy and then assigns prices to them. In this case nominal GDP wouldn’t have much meaning.
What do you think of this analogy: NGDP is iodine and the economy has primary hypothyroidism. ?
Seth Roberts helped me sleep better. He shed a light on research problems within medicine, the moral failures of modern academia and saved more than a few from oral surgery through his research on omega 3 supplementation.
He was a true scientist and I will miss his posts greatly.
The current version of EfficientForecast.com is flawed in a number of ways. Foremost among these flaws is that the system takes market prices as they are given in a single moment, with no regard to the time stamp on said prices. That is, if the price of an asset at 11:26 is listed on a web source as $50, it goes into my book as $50 @ 11:26, even if the price was delay and carried a time stamp of 11:05 on the source site. This is a problem because commodity and bond market prices are released with a greater lag than stock prices, so the forecast for say 09:45:00 EDT, might have the S&P 500 quote from 09:40:00, but copper prices from 09:30. The errors from this desynchronization tend to average out, and wont have much of an effect on the system, hour to hour, but it does mean that when monetary policy is announced, the effects ‘ripple’ through the forecast. First the S&P will move, then bonds and then commodities, this gives the forecast a ‘whip’ pattern.
Another problem is that TIPS yields are updated infrequently (on the free source I use). This means that when news breaks, the five-year yield will move, but TIPS wont, causing another distortion because the 5-year TIPS spread is an important input in my system.
The last major problem with the system is that it uses the level of 5-year treasury yields. This is problematic because, under QE, treasury yields are no longer a reliable indicator, higher yields meant easy money for QEI and QEII, but then yields fell on the news of QEIII, while stocks, commodities and TIPS spreads pointed to easy money. Tapering or fears of Tapering have typically meant higher yields. Euro Zone and geopolitical tensions are also a source of distortion on the yield level.
Since mid-January, I’ve been running (in parallel to the existing program) a new version of Efficient Forecast which fixes all these issues.
The new system features data desyncronization of no more than three minutes. It fills in gaps in the 5-year TIPS index quote using a highly accurate system of mapping individual TIPS bonds (quotes every minute) to the ‘official’ 5-year Index, and uses yield curve spreads instead of the 5-year yield. It’s not finished yet, as I’ve found out a few days ago that the source I use for dollar index quotes hasn’t had any intraday variation for…a while. This is easy to fix. Oh, and the new system averages over three models, and I’m exploring the out of sample properties of others. When I’m satisfied that it’s stable, and when I’ve exhausted the good-out-of-sample model space, I’ll started publishing the average forecast on the site.
I’d also like to roll out a new, faster graph for the site, though it’s looking like I’ll have to teach myself Java script to do that…so it could be a while. Japan and UK versions are goals for 2014, though I haven’t looked at data availability for those much.
Oh right, so the title of this post alludes to Yellen’s press conference today.
Here’s how the system reacted to Yellen’s speech. Forgive the excel graph
The plot shows the intraday year-ahead forecasts for March 18 and 19. The straight line for the start of March 19 (today) is because I made an error loading the system this morning and had to reset it at lunch. When the system has missing data points, it interpolates, thus the straight line. The take away from the graph is that the system reacted in an intuitive way.
Even though the S&P 500 (and 100) are only two of the eleven assets used in the system, and despite the dollar index input being ‘frozen’ at this morning’s open for the whole day, the new program gave a sequence of forecasts broadly similar to the shape of the S&P 500 after Yellen started speaking. It looks like you’d imagine a true NGDP futures market would look.
I consider this a good sign that the next evolution of Efficient Forecast is on track.
P.S. if the graph is hard to see, here are some forecasts from today with time stamps, Yellen spoke at 14:00 EDT:
Figures are the expected % change in NGDP over 2014Q1 to 2015Q1
|2014-03-19 13:51:00 EDT||3.859|
|2014-03-19 13:54:00 EDT||3.859|
|2014-03-19 13:57:00 EDT||3.860|
|2014-03-19 14:00:00 EDT||3.872|
|2014-03-19 14:03:00 EDT||3.855|
|2014-03-19 14:06:00 EDT||3.854|
|2014-03-19 14:09:00 EDT||3.800|
|2014-03-19 14:12:00 EDT||3.765|
|2014-03-19 14:15:00 EDT||3.766|
|2014-03-19 14:18:00 EDT||3.770|
|2014-03-19 14:21:00 EDT||3.774|
|2014-03-19 14:24:00 EDT||3.768|
|2014-03-19 14:27:00 EDT||3.770|
|2014-03-19 14:30:00 EDT||3.798|
|2014-03-19 14:33:00 EDT||3.825|
|2014-03-19 14:36:00 EDT||3.810|
|2014-03-19 14:39:00 EDT||3.804|
|2014-03-19 14:42:00 EDT||3.800|
|2014-03-19 14:45:00 EDT||3.793|
|2014-03-19 14:48:00 EDT||3.801|
|2014-03-19 14:51:00 EDT||3.799|
|2014-03-19 14:54:00 EDT||3.806|
|2014-03-19 14:57:00 EDT||3.803|
|2014-03-19 15:00:00 EDT||3.807|