Commenter Ted asked me the following question:
What will money look like in 100 years? In 1,000 years? In 10,000 years?
I’m no more able to answer that question that a citizen of the Holy Roman Empire (or a Mesolithic human) would be able to describe our current monetary system. So I’ll try to answer an easier (but still quite difficult) question:
What will money look like in 50 years? In 100 years? In 200 years?
My current views on monetary policy are quite unconventional. My belief that these views are true can be seen as an implied prediction that the policies will eventually be seen as conventional, and thus will be adopted.
In 50 years:
In 50 years, I expect something like NGDP targeting, level targeting. I expect the Fed will be willing to sell short unlimited NGDP futures contracts at a price implying 5% NGDP growth, and buy unlimited NGDP futures contracts at a price implying 3% NGDP growth. Interest rate targeting will no longer be needed, rather the New York Fed will be instructed to adjust its balance sheet (and/or interest on reserves) in a way that avoids having the Fed take a risky position on NGDP futures.
I see a 50-50 chance that paper currency will have been abolished.
In 100 years:
It is extremely likely that paper currency will have been abolished by 2119. The Fed will allow private citizens to have checking accounts at the Fed. This is because it’s unthinkable to have a monetary system where ordinary people cannot hold base money.
The Fed will use AI to determine the optimal target for monetary policy. This target will be similar to NGDP, but not identical. NGDP will be estimated on an hourly basis, as “big data” will provide the Fed with near real time access to aggregate nominal spending on goods and services, as well as nominal domestic income.
In 200 years:
Currency zones will no longer coincide with nation states. There are two possible models that could replace national money, the Eurozone and China. In the Eurozone, 19 countries share a single currency. In China, there are three currencies (four if you include Taiwan) in a single country. I don’t have strong views on which model will win out, but it seems highly unlikely that currencies will continue to coincide with national boundaries, that far out into the future. Instead, currency zones will typically cover areas that are either larger or smaller than nation states—perhaps both. What possible logic would there be for the current set-up, in a world likely to be radically different from today?
Surprisingly, I believe that wage/price stickiness will continue to be a problem in the year 2219.
I am not at all confident about any of these predictions, except for the eventual decline of paper currency.
READER COMMENTS
Brett
Jan 6 2019 at 1:22pm
I think by 2100, interchangeability and digital payment systems means that for the average person “national currencies” don’t really mean much. As far as you can tell, you can just fly from the US to Japan to India, paying with whatever your mobile device might be by then – and everything else is handled on the back end out of sight. They only become meaningful if you’re moving large amounts of money between currencies.
Matthias Goergens
Jan 7 2019 at 4:33am
Yeah, currencies will probably have a longer impact as a unit of account than as a means of exchange.
Don Geddis
Jan 7 2019 at 6:46pm
Exactly. Most people (and Brett seems to be one of them) focus far too much on the Medium of Exchange function of money. But macroeconomic significance comes from the Unit of Account function. When you take out a house mortgage from the bank, what is it exactly that you promise to pay back 30 years from now? When you accept a job, what exactly will be in your paycheck, a year from now? Those questions will have answers, and those answers will matter. Thus “currencies” will continue to matter.
Scott Sumner
Jan 6 2019 at 2:03pm
Brett, That’s one reason why I don’t expect currency zones to follow national boundaries forever.
robc
Jan 7 2019 at 9:05am
couldn’t currency “zones” also be a result of getting rid of central banks and going back to a model of banks issuing their own currency?
Alan Goldhammer
Jan 6 2019 at 2:30pm
In looking back at the last two months, the only place that I’ve used paper money is the barber shop.
Matthew Waters
Jan 6 2019 at 3:10pm
Even over 200 years, I do not think the monetary system will change neqrly this much. Path dependency is ridiculously strong.
In 1818, money existed as coinage, banknotes or deposits. Today, money exists as coinage, banknotes or deposits.
The Federal Reserve still technically exists like the first two national banks and 1800’s chartered state banks. It has:
1. A specific government charter for a private corporation. The Fed Banks have extremely restrictive corporate charters, but nevertheless a corporate charter.
2. Initial capital paid in.
3. Liabilities consistenting of banknotes and deposits.
The liabilities are no longer for specie, but the technical structure remains.
Mark Z
Jan 6 2019 at 3:38pm
I think people a thousand years ago actually wouldn’t find modern money that unthinkable (paper money, that is; credit cards would seem bizarre). Deposit banks existed during the middle ages, promissory notes existed. In some places (China, for example) paper currency had even been used. I think they’d view paper currency as very optimistic about the credibility of the state (though they’d acknowledge states can devalue precious metal currencies almost as easily), but not otherworldly.
Scott, may I ask why you think price stickiness will persist? Clearly, digitization has led to increasingly dynamic pricing (not just thanks to Uber and AirBnB, but also Amazon) and in some cases dynamic wages (again thinking of Uber). Do you just think this trend won’t go far enough to render stickiness irrelevant? And relatedly, are you familiar with Christopher Sims’s article arguing that price stickiness is caused by ‘rational inattention?’ I’m wondering how the persistence (or lack thereof) of price stickiness depends on its etiology.
Matthew Waters
Jan 6 2019 at 3:48pm
Private Fed accounts also have large technical issues.
Currently, for Treasuries, Fed reserves, Treasuy checks and postal money orders, the government relies on warranties from well-capitalized banks. Banks then make the appropriate cost/benefit analysis. They have to both please their customer, minimize their cost and indemnify both the customer and Fed.
Morgan Ricks’ Fed account proposal glossed over these issues. The Fed account proposal could change to:
1. Full-reserve (or 100% Treasury-backed) banks with adequate capitalization for indemnity, or
2. FedWire opened up to private entities. The same strict security standards exist as current Fed account holders. Any fraudulent transfer has irrevocable loss.
artifex
Jan 6 2019 at 4:08pm
The Federal Reserve System has barely existed for more than a century. It might not exist in 100 years. I give it 10% probability, but more if we condition on human civilization still existing then.
Scott Sumner
Jan 6 2019 at 9:19pm
Mark, I believe that people prefer sticky prices, and especially sticky wages.
Mark Bahner
Jan 7 2019 at 12:55am
I think within the next 10-20 years, as the combined power of global computers begins to exceed the combined power of global human brains, the global annual per-capita real GDP will begin to increase by more than 7 percent per year. That means a doubling time of less than a decade.
About that time, the global per capita GDP will be about $20,000 (purchasing power parity). So 10 years later $40,000…ten years after that $80,000. I expect the global per capita GDP in the year 2100 to exceed $10,000,000 (year 2019 dollars).
So what will money be like in 2119 when the average person is making $10+ million per year? I don’t think there will be money.
Matthias Goergens
Jan 7 2019 at 5:48am
Apart from the industrial revolution (that’s still sort of ongoing in some countries), productivity growth seems have almost universally slowed around the globe. Even Moore’s law seems to be slowing. And we are getting less benefit out of each new available FLOP. (Standard issue of diminishing returns, when flops were expensive, we only used them on really important stuff.)
So all my sympathies for the ‘singularity’ aside, the outside view looks like slowing growth is the null hypothesis.
Ed Zimmer
Jan 7 2019 at 2:49pm
If you’ll read the engineering journals, you’ll see that the technological revolution is not slowing, but accelerating at a mind-blowing rate. You’re not seeing it in the economy, and likely won’t, for at least two very good reasons. First, the transition from capital to technology – whereas capital was about doing more with more, technology is about doing more with less, ie is inherently deflationary. Second, the lag-time from the development of technology to it’s economic application is unbelievably long. There exists a huge communications gap between business leaders and the engineering community – so much more CAN be done than IS being done.
Mark Bahner
Jan 7 2019 at 11:30pm
Yes, we’re in a lull between two eras of per-capita economic growth.
In the late 20th century, we had era of rapid per-capita economic growth caused by increasing human population and increasing economic freedom. The percentage growth rate of human population peaked in 1968 at 2.09% per year, and the absolute population growth rate peaked in 1988 at approximately 92,900,000 people per year. Economic freedom in the world (as measured by such means as the Heritage/Frasier economic freedom index) increased greatly as China and India opened their economies, but yearly gains are slowing. So this era has essentially come to the end.
The era of growth from machine intelligence hasn’t even started (contrary to what you may have read/heard). Using a 20 quadrillion calculations per second (call it 20 petaflops for familiarity and simplicity) as the capacity of the human brain, or one “human brain equivalent (HBE)” the whole world had only 5 HBEs in 1998 (i.e., all the computers in the world were capable of only 100 quadrillion calculations per second, or 100 petaflops). In 2018, it was about 5 million HBEs (very small compared to the human population of 7.5 billion actual brains). But in 2038, it will be about 5 TRILLION HBEs, which will dwarf the human population.
Moore’s Law, as it was originally precisely formulated, is not very relevant to economics. What is far more relevant to economics would be a “corollary” of Moore’s Law, which is that the computing power (calculations per second) per $1000 is doubling approximately every 1-2 years. As of the end of 2017 (the latest date for which data are available on Wikipedia) the rate of growth hasn’t slowed down.
There are diminishing returns, but there are also synergies that aren’t available with less computer power. For example, Velodyne lidar systems are having more and more lasers, producing more and more data. It would have been pointless to do that, if computers weren’t improving in data handling.
As I wrote, in 2038, assuming the present computing power trend continues, the number of “human brain equivalents” due to worldwide computers will be 5 trillion. And if the trend continues just another decade, it will be over 1 quadrillion HBEs. The only way that could not dramatically increase the economic growth rate would be if more human brains do not increase the economic growth rate.
Cliff
Jan 7 2019 at 11:52pm
While I am sympathetic to your conclusions (although I think unfriendly AI is quite likely), this:
“The only way that could not dramatically increase the economic growth rate would be if more human brains do not increase the economic growth rate.”
Is obviously wrong, since human brain processing power equivalence is not he same thing as a human brain.
Mark Bahner
Jan 8 2019 at 11:52am
I don’t think you appreciate the scale (magnitude and rapidity) of the coming change. I’m taking 20 quadrillion instructions per second (call it 20 petaflops) as a “human brain equivalent (HBE)”. And if the trend of the last ~30 years continues for the next 30, I’m talking about over a quadrillion HBEs (1,000,000,000,000,000). So even if I’m off by a factor of one million–that it takes a million times more computing power to be functionally equivalent to a human brain–we’re still talking about 1 billion human brain equivalents by 2048.
And the chances that I’m off by a factor of one million are incredibly small. Look at the NVIDIA Pegasus autonomous vehicle computer. NVIDIA thinks that when paired with appropriate sensors (e.g. cameras, lidar, GPS, etc.) it will be capable of full Level 5 autonomy (the highest level…can drive anywhere under any condition). The NVIDIA Pegasus is capable of 320 trillion operations per second…that’s approximately a factor of 60 less than the value I’m assuming.
Don’t get me wrong…I think economists should be doing much more research about just what are the relationships between computer power and human brains with respect to economic performance. But again, the only way that adding 1 quadrillion times 20 petaflops of computer power to the world will not dramatically increase economic growth rate*** is if adding huge numbers of human brains does not increase economic growth rate.
P.S. ***One significantly tricky aspect about adding computers is that they work for free. So conventional GDP calculations–based on conventional inflation calculations–can appear to show no GDP growth where there is actually significant GDP growth. That is, the total value of goods and services may increase significantly, without being able to account for the value in GDP calculations. To give an example, whereas 20 years ago, a tourist might buy a cell phone, a digital (or film) camera, and a translation book, a smartphone can handle all those things. So GDP calculations might show a decrease in economic activity (no cameras bought, no translation books bought), but the true value has actually increased…no need to lug a separate digital camera or buy a translation book.
RPLong
Jan 7 2019 at 12:52pm
I’m sympathetic to the view that paper currency will be abolished, but upon further reflection, I consider it unlikely that in the future money will be 100% digital. There has to be some tactile store of value that cannot be compromised in the event of a major technological failure of some kind (natural disaster or whatever). I do not believe that humans would be so foolish as to entrust their entire economic system to machines that cannot be operated during a power failure.
I agree that money as we know it will change drastically, but I guess my imagination is too limited to conceive of a world in which some non-barter transactions must be conducted using physical money of some kind.
Matthew Waters
Jan 7 2019 at 2:50pm
Most of the financial system is electronic going back to the 60s. Half of all dollars are paper, but most of the paper dollars are held overseas. Most securities have electronic registration of owners at the transfer agent and depositories.
The financial system has hidden the electronic-ness really well. There are a lot of disaster recovery plans, including a catastrophic natural disaster.
Paper currency has worse security and reliability than electronic currency, all told. What paper currency has is strong privacy guarantees.
Todd Kreider
Jan 8 2019 at 6:11pm
Mark: Every couple of years I look at Wikipedia’s MIPS (Millions of Instructions Per Second) to show that there has been no slowdown. Which three chips would you pick from around 2003, 2010 and 2017 to show this?
(Paper money won’t be around after 2030-2035 and in steep decline by 2027.)
Mark Bahner
Jan 8 2019 at 9:37pm
Hi Todd,
I think the easiest thing to do is to go to the Wikipedia article titled, “FLOPS” and look at the years, 1997, 2007, and 2017. We have:
1997 –> $46,000,000 per teraflops.
August 2007 –> $57,000 per teraflops.
June 2017 –> $60 per teraflops.
October 2017 –> $30 per teraflops.
So the average for the two 2017 values is $45 per teraflops, which is almost exactly a factor of a million (approximately two to the twentieth power) lower than the value of 1997, twenty years previously. So the cost per teraflops has been cut in half about every year for the 20 years from 1997 to 2017. The value for August 2007 falls approximately along that line.
P.S. Some interesting notes with respect to all this:
1) The NVIDIA Pegasus (car-driving computer), at 320 teraflops, would have cost…$14.7 BILLION in 1997 ($46,000,000 per teraflops times 320 teraflops).
2) The NVIDIA Pegasus would have been more powerful (in terms of calculations per second) than all the computers on earth, combined in 1988! (Which coincidentally, was the year I bought my second computer…I think it ran at 18 megahertz, had 2 meg of RAM, and a hefty 80 megabyte hard drive. I think it cost $1600.)
Todd Kreider
Jan 8 2019 at 10:20pm
Hi Mark,
Thanks!
Comments are closed.