Why is it so expensive to create an NGDP futures market?

When Eric Crampton first suggested the idea of creating an NGDP futures market in New Zealand, I was intrigued.  It would be outside the oppressive regulatory regime of the SEC.  When Robert Quigley-McBride said it could be set up for $1500, I was pleasantly surprised.  I might be able to raise those funds from readers.  Hell, I could pay that myself, except my wife would probably (quite reasonably) ask why I had to do so, when I’d already devoted six years to this project, mostly unpaid.  So I didn’t ask her.

Then Robert said it would cost an extra $30,000 to set up the accounting infrastructure to handle all the interest-bearing margin accounts, if we adopted my complete proposal.  When I read that I thought to myself “fat chance of raising that kind of money.”  And in fact we did not raise $31,500, we raised more than twice that sum, in less than 2 days.  Approaching $70,000.  And without even asking for small donations. I’m stunned.  (Special thanks to Ken Duda and Gabe Newell for huge donations, but the others were also very generous.)

But as all good philosophers know, every success just creates new problems.  Now I started to worry about liquidity.  There is no NGDP futures market in the US, precisely because there is very little demand for NGDP futures contracts.  And this market would be in New Zealand.  Who would trade?  My initial proposal in academic papers was to have the Fed run the market, and the interest rate on margin accounts would be raised high enough to assure liquidity.  The Fed could easily afford to do that for a market with several $100 million in margin accounts.  But this current proposal would merely pay market rates (now quite low); the funding would be to set up the complex accounting system. The more money we raised, the more guilty I felt about the prospect of wasting the money of idealistic donors.

I have some ideas on how the trading volume concern can be addressed, which I will present below.  But I am also looking for your ideas.  Before rushing in, however, try to remember two things:

1.  Don’t look at the donations thus far as a big pot of money to spend.  I am determined to do this as cheaply as possible, given that we can be confident of succeeding.  I am determined to rethink the plan in such a way as to come in well under the promised donations thus far, so I can present the donors with a plan that has a high likelihood of success, and also tell them we only need X% of their promised donations.  Where X is well under 100.

2.  This is not a normal futures market, and hence normal rules don’t apply.  Normal futures markets are trying to meet customer demand for hedging and speculation.  There is little such demand for NGDP futures.  We are trying for a demonstration project showing that it would be possible to create a public good that is highly useful in policy analysis, by leveraging the “wisdom of crowds.”  So don’t rush in and tell me it must be done “this way” because pork bellies are traded this way.

I’ll get the ball rolling, but in the end this isn’t just my project.  You must also convince both the experts at iPredict and the donors that your proposal is feasible.  So even if you convince me it may not happen.  Here’s one thought I have:

Robert told me that a direct subsidy might be much cheaper to administer than a complex set of interest-bearing margin accounts.  Other accounts at iPredict do not earn interest.  So perhaps we could do something like the following:

Let’s suppose our budget ends up being $41,500.  The $1500 sets up the market, and $40,000 is used to subsidize trading.

Instead of investing margin account money, let’s think of direct subsidies to traders.  We could plan on a 5 year horizon for the project (then raise extra funds later if we are successful.)  That’s 20 quarters of operation to fund.  In that case we can afford a $2,000 subsidy each quarter.

One way of doing the subsidy is to have each winning contract equally share the $2000.  That would guarantee some trading, as if only one person bought contracts they would be guaranteed to win the entire $2000!  Overall I’m strongly interested in a subsidy scheme that does not allow for a complete failure of the market.

Say what you will about Paul Krugman, he’s a genius at intuition.  He once remarked  (I’m over-simplifying) that all of economics boils down to two jokes:

1.  That can’t be a $100 bill, someone would have picked it up already (the EMH)

2.  No one goes there anymore, it’s too crowded.  (S&D)

My plan is a subsidy not a price, so it sort of reverses #2.  “Lots of people buy NGDP futures, as the market is so uncrowded that they have a good chance of winning the subsidy.”  Of course that’s nonsense, but it’s a way of thinking about why my proposal would be a sort of foolproof plan for insuring at least some trading.  Low volume is hard to conceive as an equilibrium outcome. That’s what I’m aiming for.

Some commenters prefer a plan with a contract linked to the specific future NGDP, not just a range of values.  I like that better too, as it provides precise point estimates.  But iPredict found that people much prefer to trade things like win/loss in elections, rather than vote share.  Again, we need volume for success.  Vote share contracts are boring because the vote share rarely diverges more than a couple percent from public opinion polls.  Then I thought about tying the contract price to the growth rate of NGDP, which is more volatile than the NGDP level.  But the starting point can undergo revisions, which makes the growth rate more ambiguous.  So we didn’t end up there, but I suppose the contract price could be the percentage difference between the actual initial announcement of NGDP, and the estimate of 4 quarter previous NGDP on the very same day (if we were dealing with a one year contract.)  It’s something to at least consider.

Other proposals would subsidize the market maker to assure more liquidity, but I worry that while it’s a good idea, that might still not overcome a lack of interest in US NGDP (outside the US). Recall the goal is to derive a public good—the public’s expectations of US NGDP growth, even if they have no interest in trading without a subsidy.  Did I mention that volume is the problem?

Because I did not ask small contributors to help out, I am going to ask my non-American readers to give this market a shot, if only for the fun of it.  It will be set up to be either a zero or positive sum game; what casino offers those odds?  American readers can recommend this to their foreign friends and colleagues.  Eventually I will also ask other bloggers to point out the value of the experiment, if it succeeds.

We are also working on issues of how best to deliver the donor funds to NZ, hopefully in a way that earns a tax deduction.

I suppose it’s slightly embarrassing to be rethinking things this late in the game, but I’ve never much cared about losing face, and do care a lot about getting this done right.

And finally, on a personal note, the last few days have been very gratifying.  A few hours after the initial post I had not gotten any donations, then for another hour only one.  I flashed back to my Cub scout days where I only sold one box of cookies, because I was too shy to approach people. I’ve never been good at asking people to do me favors, and would have been a horrible salesman. “You actually don’t need these encyclopedias, to be honest it’s all on Wikipedia.”  The wonderful thing about the anonymity of the internet is that it allows a loser like me to actually have some small moments of success.  The downside is that it also allows me to be somewhat more of a jerk than I am when actually meeting people face to face.



50 Responses to “Why is it so expensive to create an NGDP futures market?”

  1. Gravatar of Jason Smith Jason Smith
    4. October 2014 at 10:22

    Hi Scott,

    An idea to get some of your US readership involved would be a donation system set up in the US to fund that subsidy in the iPredict market — donations that are accompanied by NGDP projections, where the correct ones get a mention on your blog.

    Also — you could turn the market into a yes/no question by using a log-linear extrapolation from the past two (or more) quarters and the contract be a greater than/less than choice relative to the projection.

    Good luck with your endeavor!

  2. Gravatar of Joel Aaron Freeman Joel Aaron Freeman
    4. October 2014 at 10:45

    Dumb question. Why is the market NGDP in the United States? If Americans can’t participate, then who will be interested in a US measure? New Zealand’s NGDP is possible, but that’s a very small country. How about NGDP for the Eurozone as a whole? NGDP for Europe is a number of tremendous international significance, and one Europeans might be interested in trading.

  3. Gravatar of Nick Nick
    4. October 2014 at 11:43

    First, a suggestion:
    You might want to consider being more promotional with the money … don’t necessarily split it into even lumps for the subsidy. It’s not how a business would do it, I think.
    You should have a larger ‘marketing’ budget to start off with. try to make a splash and generate interest immediately.
    It might also be prudent for the purpose of the experiment to always subsidize one quarter more than the other three, or try a different method of subsidy in one of the four quarters. Then you could start to see how much good the subsidy is doing or if a different structure generate a different result.
    Congratulations! I wish it were legal for me to trade on the platform…

  4. Gravatar of ssumner ssumner
    4. October 2014 at 12:57

    Jason, Joel and Nick, Those are all three very good ideas, just what I am looking for.

    Jason, I just may do something like what you suggest, I especially like the first idea.

    Joel, People pay more attention to US NGDP than eurozone NGDP (where it’s often a country by country focus.) But I do think your point has a lot of merit, and with the extra money collected we could consider at least a basic NGDP contract for the eurozone at little cost, perhaps just once ever 4 quarters, or we could try for 12 month growth rate forecasting. Something that would not double the number of contracts, because that spreads bettors too thin.

    It’s also the case right now that the eurozone NGDP growth rate over the next 2 years may be both more uncertain and more consequential for global welfare. Recall the importance of the 2011-12 double dip recession.

    Nick, Good point, and I’ll pass that along to Robert, who will have a better idea of how effective various forms of promotion are. But front loading the promotion does make a lot of sense.

  5. Gravatar of Clayton Daley III Clayton Daley III
    4. October 2014 at 13:12

    Subsidizing the market may help kick it off, but it isn’t sustainable or scalable. If the market gets sufficient publicity and still needs a monthly “prize” after year 1, it’s likely running on borrowed time.

    I could flesh these out further, but a couple lines of thought are:
    1) You must find a model that creates scalable value for direct participants in the market.
    – At scale, market participants (on average) break even (or worse with fees) so “gambling” is a shaky value proposition.
    – Who has risk highly correlated with NGDP? Who benefits from a nominal hedge? Who can’t hedge risk elsewhere? What about importers?
    – If you find someone with that risk, is anyone willing to buy the other side?
    2) If you were convinced that someone would benefit from participating, would you be willing to write the whitepaper and “market” the value to those organizations?
    – Instead of cash prizes, consider spending most of the money on marketing, white papers, speeches, etc.
    3) Once the platform is built, the marginal cost of new markets is nearly zero.
    – Does anyone have a robust technical platform they’d give, lend, or license?
    – Could you build (and fund) the platform in a liquid market that isn’t exposed to US regulation? If profitable, that market could fund all technical development, letting the US market tag along. It could even provide a sustainable source of “profit” to subsidize the US market.
    3a) If importers care, NZ and Australia do tons of trade with China. How about a Chinese NGDP market?
    – Would the Chinese government be willing to buy the other side? How much credibility would spill over from such a commitment? If it makes sense, who would you want/need to persuade inside China?
    – Would the Aussie or NZ government be willing to buy the other side of domestic transactions as an economic stabilization policy? Similar questions as the last bullet.
    – Would someone else take that bet?

    I think the question “who benefits from direct participation” is what you really need to pose to your audience (and colleagues). My impression (without deep technical analysis) is that NGDP is a tight proxy for the market portfolio (but clarifying this for readers may help them help you). In that case, the question sounds like an econ 101 question about pricing and trading beta.

  6. Gravatar of Rajiv Sethi Rajiv Sethi
    4. October 2014 at 13:13

    Hi Scott,

    I only just became aware of this initiative (thanks to a David Levey email). I’ve studied and written about various kinds of prediction markets quite a bit so here are a few thoughts. Maybe some or all of these have already been brought to your attention by readers; if so I apologize in advance.

    First of all, it should be a real money peer-to-peer market with a public order book in which a fair amount of depth is visible. There is no need to subsidize anyone: if US residents are permitted to open accounts and trade they will do so despite the fact that it is a grey area legally. Most of Intrade’s volume came from the US. It may take a while to gain traction but it will get there if you are patient.

    Second, it the contracts should be winner-take-all and there should be relatively few of them. The prospectus must specify very clearly how they are to be settled, based on a public announcement. Intrade had contracts on Mubarak’s departure that were settled prematurely and then reversed. There will be chaos if there is any ambiguity regarding settlement.

    Third, and perhaps most important, trader margin has to be kept credibly secure. You can’t have even the slightest suspicion of another MF Global style commingling of funds. This is what did Intrade in. In particular the funds you have raised should be set aside for operating costs, including perhaps the retention of an auditor. This is a better use of funds than subsidies. No amount of subsidy will generate volume if the funds deposited are not credibly safe.

    Fourth, the contracts should be simple, and reference an easily computed and verified magnitude. For instance, four contracts which specify NGDP growth of 2,4,6,8 percent relative to a public baseline. Each contract pays the long position if growth exceeds the referenced amount, and pays the short position otherwise. This is just an example, but it will give to a probability distribution not just a point estimate. Only the contracts for which there is genuine disagreement will see heavy volume.

    Finally, execution has to be close to instantaneous. Once an order is placed the margin should be adjusted and the order book refreshed right away, whether or not the order is marketable.

    Not sure if this helps at all, but it’s an interesting experiment so I thought I’d give you my two cents.

  7. Gravatar of anon anon
    4. October 2014 at 14:50

    Robin Hanson has written extensively about subsidized prediction markets, both in his academic papers and on his blog. Anyway, my own hunch is that simply providing the automated market maker with more money would give a pretty good bang for the buck – it would narrow the bid-ask spread, which stands in the way of any subsidy, while not distorting the generated probabilities as other schemes might.

    Honestly, I don’t get why you guys are not adding a linear contract _in addition_ to the binary bins. It would provide a direct estimate for _expected_ NGDP, which many people care about. (And the combination of both types of contracts might allow observers to describe the consensus probability distribution in a quite detailed way). ISTM that vote share in elections is a bad analogy here.

  8. Gravatar of dtoh dtoh
    4. October 2014 at 15:14

    Rajiv’s comments are good. I would make it even simpler.

    The futures contract has a nominal amount of $1000.

    Start with contracts on quarterly GDP going out four quarters.

    The value of the contract is based on quarterly YOY GDP growth relative to a 2.5% target (or any other number around there). E.g. growth is 2.5% then contract payout is $1000; 5% – $2000; 0% or below – $0. Use the BEA advance quarterly announcement as the benchmark to determine final settlement amount.

    You need a good settlement/margin system. Since the contract is valued on GDP growth, it will be fairly volatile which will make it more interesting but also require higher margin. (If you want lower volatility and margin requirements, you could value the contract on growth relative to 90% of the year earlier quarterly GDP number, but I would make it simple to start with.)

    If you want to subsidize the thing. Just open an account and put the donations in the account. Set up the account so that on a daily basis it automatically sells a fixed number of contracts at a market clearing price if volume for the previous day was below some floor level.

  9. Gravatar of dtoh dtoh
    4. October 2014 at 15:27

    And forget the binary bins. Bad idea.

  10. Gravatar of Lawrence Indyk Lawrence Indyk
    4. October 2014 at 16:22

    I think some basic wisdom when doing anything for the first time as an experimental trial is:

    1. Keep it cheap
    2. Keep it simple
    3. Keep it intuitive
    4. Make it easy to generate meaningful insights, lessons learned, and potential future refinements from post facto analysis.
    5. Trust people with relevant experience

    I think all this argues for a pure Direct Subsidy on the simplest kind, and smallest number, of futures contracts.

    The amount of the subsidy could “float” in order to always guarantee the production of the minimally acceptable amount of volume (in other words, you are letting the market decide that too, instead of making an arbitrary choice).

    In the alternative, a few rounds of trial and error should help discover the level of subsidy most likely to create the desired effect, and you could stick with that for a longer period, after which you could reassess.

  11. Gravatar of Lorenzo from Oz Lorenzo from Oz
    4. October 2014 at 16:35

    Tenured professors do not get to describe themselves as losers 🙂

    Terminally shy, hopelessly diffident perhaps 😉

  12. Gravatar of Eric Crampton Eric Crampton
    4. October 2014 at 16:45

    I’d suggest that everyone who’s interested go have a look at the existing iPredict website to see how it works. There’s a standing visible order book of bids and asks on every contract; the API lets you get farther into the order book if you’re really keen.


  13. Gravatar of benjamin cole benjamin cole
    4. October 2014 at 16:58

    Kudos to all. I will contribute if you can take paypal. No shame in evolving the idea…iterative adaptation got us this far…

  14. Gravatar of honeyoak honeyoak
    4. October 2014 at 17:28

    One thing that I can recommend is that you make contracts be settled in terms of a multiple of positive basis points of GDP growth. For example if GDP is expected to grow by 2% in Jan 1, have the the contract valued at $2 (or $20,$200 depending on the multiplier). This way if GDP growth is 2.2%, the contract would be valued at $2.2 for a 10% profit. The same would be true if GDP growth is negative (with margin calls of course).

  15. Gravatar of Bob Bob
    4. October 2014 at 18:19

    I am glad donations went well. I was surprised you were so pessimistic about your funding objectives though, given how many wealthy readers you have. Have you checked GabeN’s net worth? It’s not quite as having Bill Gates in your readership, but it’s close.

    If you were more interested in micro, I am sure you could convince him to give you access to his Steam Marketplace transaction treasure trove. There are dozens of papers just waiting to be written on top of that data.

  16. Gravatar of Saturos Saturos
    4. October 2014 at 19:09

    Gabe Newell is now like one of my 10 favourite guys on the planet.

  17. Gravatar of H_WASSHOI (trader) H_WASSHOI (trader)
    4. October 2014 at 22:14

    Where can I find&buy the futures?

  18. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 04:46

    Eric, I looked at the introductory video and didn’t see an order book on the trading screen so there must be some other way to get to it. Otherwise it seems to operate much like Intrade. Separation of trader funds is credible and settlement terms are clear. There are limits to account size (like IEM) but these are not terrible. I don’t see why they don’t just create the market right away, once contract specs have been decided. Seems easy enough since the infrastructure is in place.

    But I don’t see why Scott doesn’t just contact the CME, which already offers futures contracts that reference temperature, rainfall and hurricanes. This seems to be the easiest path to high volume and I can’t imagine any regulatory obstacle.

  19. Gravatar of Scott Freelander Scott Freelander
    5. October 2014 at 05:48


    Have you considered trying to get Robert Shiller involved? Yes, he might be busier than ever after his Nobel, but as you know, he favors a greatly expanded role for prediction markets in providing certain services/indicators. I think I once read he wanted something similar to an NGDP futures market, but perhaps for different reasons. Obviously, he could give your effort much more publicity and may have good ideas to add.

  20. Gravatar of ssumner ssumner
    5. October 2014 at 06:57

    Clayton, Those are very good points, but I really don’t think it’s realistic to believe that very many people would ever benefit from participating without a subsidy. Like the $100 bill joke, why doesn’t the market already exist? The goal is not to have the market run forever, but just long enough to convince economists that the Federal government should take over the duty. They have deep pockets, and could easily afford a few million per year in subsidy, forever.

    Rajiv, Thanks for those very helpful comments, I will pass them along. The only point I disagree with is the first. I don’t think it’s realistic to assume adequate trading without a subsidy. And I’m not sure why so many commenters think it is. I’m told Goldman Sachs or one of the other huge financial firms tried to set up a NGDP futures market in Chicago and failed because they got very little trading. Why does everyone assume no subsidy will be needed? I don’t get it. And regarding the laws, I’m not worried about Americans, I understand they can get around the rules, I’m worried about iPredict being crushed the way the US crushed Intrade, and online poker, and a bunch of Swiss banks.

    But even if I disagree with your first comment, you have lots of other great points.

    anon, Yes, I have relied on Robin’s expertise in forming my views.

    Adding a linear contract is an option I would personally like to see.

    dtoh, You said:

    “The futures contract has a nominal amount of $1000.”

    That would immediately remove 90% of our business, and our biggest concern is volume. I’ve traded at Iowa Electronic market, but would never even consider a market with a $1000 minimum. There are many people like me.

    Otherwise some very good ideas, some of which are already incorporated.

    Thanks Lawrence, those are very good ideas.

    Eric, Thanks for that info.

    Ben, Good idea, I use paypal all the time, it’s very convenient.

    Bob, Yes, I once had an interesting phone conversation with some people at Valve who work with data.

    Wasshoi, It’s not set up yet, but it will be before too long. At iPredict. The link is in Eric’s comment.

    Scott, Just yesterday I was thinking of contacting him. I will at some point.

    Everyone. I’m hearing a lot of suggestions that we should consider linear contracts, and/or contracts for other regions. Regarding linear, almost everyone thinks growth rates are better. Let’s think about the exact structure. Should it be:

    Growth from initial estimated NGDP for 2014:4 to the initial estimated of NGDP in 2015:4, or should it be the growth rate from 2014:4 as estimated when 2015 data comes out, to the initial estimate of 2015:4? I think I prefer the latter, as it addresses the concern about rebasing GDP, like when they added R&D to investment a few quarter back, and everything shifted higher by a percent or so (including older data). It’s a true growth rate, using constant methodology.

    I like dtoh’s idea, but might opt for $100 contracts, with 4% as the baseline US NGDP growth. The contract would have zero value below zero, but if that prospect seemed possible we could add binary contracts for negative numbers. With 8% NGDP growth the contract would be worth $200.

    We could start out with both iPredict’s preferred approach, and also the linear contracts, and do a sort of horserace to see interest in the two options. But obviously the linear approach gives you a better point estimate. On the other hand I seem to recall that with a bunch of binary options Robert tells me that one can derive point estimates fairly accurately.

    I can imagine these regions might be popular, given the location:

    1. New Zealand NGDP

    2. Australian NGDP

    3. British NGDP

    4. Eurozone NGDP

    Another option is a parallel RGDP market. By looking at both variables in combination you could either validate or refute the entire Real Business Cycle theory by observing the market responses to central bank announcements.

    Old Keynesian says no response at the zero bound. RBC says NGDP but not RGDP. MM and New Keynesian say both, MM is more confident on that point.

  21. Gravatar of dtoh dtoh
    5. October 2014 at 07:39

    I’ve actually negotiated at length in the past with the CME (Melamed) to set up a new futures contract. It’s a very high hurdle.

  22. Gravatar of Clayton Daley III Clayton Daley III
    5. October 2014 at 07:44

    By the $100 bill argument, you might as well shut your blog down. If changing fed policy actually created value, they’d have already done it. QED.

    Fortunately, there’s plenty of evidence that $100 bills exist… like these billion dollar bills (http://graphics.wsj.com/billion-dollar-club/).

    I believe you’re chasing a billion dollar bill. It *may* be the fed policy change, but please don’t ass-u-me that’s the only one. Give Popper… or the scientific method… or Sherlock Holmes a chance. Prove that subsidization is the best/only option by:

    – Describing the theoretical nature of the NGDP futures
    – Describing the entities that would benefit from trading that theoretical construct
    – Establishing that these entities doesn’t exist

    … and get a published paper out of it. If you find that they exist, but aren’t “interested”, you have an education problem. For most startups that’s a huge red flag and they’d be best to run away. In this case, you’d be the perfect man for the job.

  23. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 07:58

    Eric, I take back what I said about settlement terms being clear on iPredict, they are hopelessly vague. For example the markets for the 2014 US elections don’t specify whether Dem control excludes independents who caucus with Dems. The corresponding IEM contract does exclude independents so the likelihood of neither party in control now stands at above 20% there. But on iPredict its 1%, presumably because traders think independents count as Dems. Recipe for disaster if not clarified.

    Scott, if your contributors simply opened accounts for 100 NZD each on iPredict they would collectively generate enough volume to bring others in, without subsidies, and as long as they made good bets on average they could contribute to volume indefinitely. Intrade had negative subsidies (participation fees) and plenty of volume.

  24. Gravatar of Peter K. Peter K.
    5. October 2014 at 07:59

    Off topic, from a thoughtful liberal:

    The war on poverty is being won in Brazil:


    “This return to growth, plus the government’s use of increased revenues to boost social spending, has reduced Brazil’s poverty rate by 55 percent and extreme poverty by 65 percent. For those in extreme poverty, the government’s internationally renowned conditional cash transfer program (Bolsa Familia) provided 60 percent of their income in 2011, up from 10 percent in 2003. A hefty increase in the minimum wage – 84 percent since 2003 after adjusting for inflation – also helped quite a bit.

    Unemployment has fallen to a record low of 4.9 percent; it was 12.3 percent when Lula da Silva took office in 2003. The quality of jobs has also increased: the percentage of workers stuck in the informal sector of the economy shrank from 22 to 13 percent.”

  25. Gravatar of dtoh dtoh
    5. October 2014 at 07:59

    On the contract denomination, don’t forget you don’t put up the face value of the contract. Just the margin. At 4% growth rate it’s less volatile. I don’t know what the margin requirements are at iPredict, but you’re probably talking a couple hundred bucks maximum…. maybe a lot less (Tsy Bonds are around 2 or 3% of the face on large exchanges.)

  26. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 08:26

    Okay, so I opened an account on ipredict and played around a bit. It looks like this is *not* a peer-to-peer market like Intrade, where traders interacted with each other. No limit orders are permitted, so the order book is populated by an algorithmic market maker that sets prices subject to some convex cost function with bounded loss. This means I can’t outbid the current best bid, all I can do is accept the current best offer if I want to buy, thus trading agains the algorithm. And the spread is quite large so there is a strong disincentive to trade. Plus there are trading fees and withdrawal fees. I suspect that anyone used to trading in conventional double auctions will hate the design. I certainly do.

    If I’m wrong about any of the above claims about design I hope someone will correct me. I haven’t spent much time on the site (and don’t plan to).

  27. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 08:44

    Spoke too soon… just managed to place a limit order. Now I’m not sure if there’s an automated market maker at all, and it does seem to allow peer-to-peer trading. Better shut up till I’ve really figured this exchange out.

  28. Gravatar of ssumner ssumner
    5. October 2014 at 09:46

    Saturos, I agree about Newell.

    Clayton, That’s a fair point about the $100 bill, but consider:

    1. Policies are not market outcomes. It’s easier to get the market to be efficient than it is to get the Fed to be efficient.

    2. Remember I’m not claiming there is a market for this contract, we are trying to create a prediction market to get a public good—NGDP expectations. That may require subsidy.

    But, yes, sometimes I am prone to excessive fatalism.

    Rajiv, Good point about the US Senate elections.

    Peter, I agree, but what they don’t tell you is:

    1. The anti-poverty program, which is somewhat effective, was started by the previous center-right government.

    2. Brazil has a very low GDP per person, especially for a country that exports commercial airliners. And it still has a very unequal distribution of income by global standards. That combination is not good. And unlike China (and even India) it is growing very slowly in per capita terms, say 1% vs. 7% in China in the past three years.

    So there are also lots of negatives in the record. But yes, they did do a good job in reducing abject poverty, and should be applauded for that. It’s a question of what kind of expectations should we have for a country that (unlike Congo) is capable of exporting commercial airliners. I’d say at least somewhat high, say Portugal, or at least Mexico, in terms of per capita GDP.

    dtoh, Yes, good point about margins. But we don’t want to exclude small traders for whom even a potential $1000 loss is scary. Middle class people think in terms of $100s, for a gamble. The big players can just do multiple contracts, like in the stock market.

    Rajiv, On your last two points, let me know what you find on that question you are unsure about. I suppose the subsidy money could be used to make the transactions costs zero for traders, at a fairly modest cost for the exchange if we assume modest amounts of trading. And of course if a huge amount of trading develops we win anyway.

  29. Gravatar of Thomas Aubrey Thomas Aubrey
    5. October 2014 at 09:49

    “This is not a normal futures market, and hence normal rules don’t apply. Normal futures markets are trying to meet customer demand for hedging and speculation.”


    If you can’t get banks to make a market based on your comments above (I think you are right) then you can still run a market without actual money (with a small cost). I implemented this successfully in a project with the World Economic Forum in 2006 for oil price spikes and for the spread of bird flu (H5N1)


    The results, as you’d expect, were better than individual oil analysts expectations, and for bird flu far better than any top down model from the WHO. A lot of the data we produced was published here:


    Using some basic software to run it, (I suggest you use http://www.lumenogic.com, Emile, who set this up, is a great guy), you could easily run it on your blog given the type and quantity of visitors you have. Key thing is to get the question formed right which is where non-money based predictive markets can often fall down. Once you start producing data, you may well find that you can then start selling it to the Fed and generate a decent return:) Happy to help out if this is of interest.

  30. Gravatar of Clayton Daley III Clayton Daley III
    5. October 2014 at 10:19

    @Thomas (et al), does a prediction market generate any “private” data? Is there anything the Fed couldn’t get by looking at the price data?

    If so, you could look for other entities (insurance companies, underwriters, investors) who might be willing to subsidize the market in exchange for data. In fact, the model already exists in web search (http://www.studentprofit.co.uk/make-money/get-paid-to-search-the-web-3-websites-to-make-money-from-1501.html). Basically the Google (free search for ad revenue) taken to the extreme.

  31. Gravatar of Thomas Aubrey Thomas Aubrey
    5. October 2014 at 10:31

    Clayton, for the Fed to use the data they would need to acquire a license according to the terms set down by the market itself. It’s possible there might be some “private data” too but the first point is the main one.

  32. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 11:44

    Scott, if you can decide on the structure of the contracts, just have your readers open accounts on iPredict and simply propose the market. It takes less then two minutes to open and fund an account by credit card and the money seems well protected. The funds you have raised could then be used to pay the trading fees for just this market (by arrangement with the exchange). I think this is a better subsidy than changing payouts. The fees are small so what you have raised will last a long time.

    I’ve been playing around a bit with the exchange and it does seem to be peer-to-peer with visible limit order books. The segregation of trader funds seems credible. Settlement ambiguity is a problem but they have already clarified the Senate contracts (I had submitted a query this morning) so they are responsive.

    The only thing left to do is to design the contracts. To fit their model, each contract will have to pay 1 NZD if the event occurs, zero otherwise. This winner-take-all structure is the best for liquidity in any case, since it allows bets as small as a few pennies. The upper bound on bets is set by a limit to account size, just like IEM.

  33. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 13:31

    Okay, before you have your readers open and fund iPredict accounts, please note the following (emailed in response to a query by me):

    “iPredict’s current policy with regards to international withdrawals is that we will make an international withdrawal for you if you are intending to close out your account, but we will not make international withdrawals otherwise.”

    So it’s very easy to deposit via credit card but impossible to withdraw winnings while keeping the account open. This will be a dealbreaker for most traders outside NZ. Subsidies can’t fix this.

  34. Gravatar of Eric Crampton Eric Crampton
    5. October 2014 at 15:03

    Hi Rajiv,

    A few notes:
    – The subsidised market maker seeds the order book with bids and asks around a seeded start point along a curve defined by the amount of money put in for that market. The greater the subsidy provided the automated trader, the more bids and asks it puts in and the more contract orders it’ll have at each bid or ask.
    – You can’t automatically tell which orders in the book are from the computer and which are from traders, but you can make guesses. If you place an order into the book, rather than accepting a current offer, you will always wind up trading with a human. If you buy or sell by taking a current bid or ask, you may or may not be trading with the computer.
    – On trading fees: it’s the withdrawal fee that’s the large one. Trading is otherwise pretty cheap. The system earns very little because the regs here, as I understand things, require that deposited funds be held in an account that must be able to pay out all depositors on demand, and so earns negligible interest. It wouldn’t be nuts to have some of the donor funds be used to make that easier for traders in these contracts, but it might require some back-end changes to make that possible. Robert’s the one who knows how that all works.

    Note that Robert’s the definitive authority on this stuff. I was academic advisor to iPredict a few years back when I was at Canterbury and when Matt Burgess was their CEO. I’m now at a NZ think tank and happy to help out, but I definitely have no ability to speak on iPredict’s behalf.

  35. Gravatar of Kevin Erdmann Kevin Erdmann
    5. October 2014 at 16:20

    Scott, I think you are too pessimistic about the market for hedging and speculating in NGDP. There are many futures markets currently operating that are no less abstract than NGDP – volatility futures, weather, housing indexes, etc.

  36. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 17:20

    Eric, thanks for the helpful explanation. I’ve exchanged a few emails with Robert today and am starting to get a sense of how this thing works. The quote about international withdrawals is directly from him.

    It seems that iPredict is a hybrid of a peer-to-peer exchange (like Intrade and most financial markets) and an automated market maker (like PredictWise, except with real money). I spotted the algorithm first but then realized that some large bids and offers were inconsistent with a convex cost function and must have come from real traders. That got me wondering whether I was mistaken about the algorithm but your comment cleared this up. It seems that the automated market maker keep prices consistent across linked contracts, and prevents arbitrage opportunities from appearing.

    I don’t see any problem with opening a set of NGDP futures contracts on this exchange, following the same rules and winner-take-all structure, with each contract having face value of 1 NZD. But the exchange needs to loosen rules for international withdrawals otherwise liquidity will be very limited.

  37. Gravatar of ssumner ssumner
    5. October 2014 at 18:18

    Thomas, Actually I would like to do what you suggest, but I am horrible with computers. I hire a guy to do tech support. So if you really would like to help, send me an email with the link and enough information to set everything up, except for the question, which I will write. I may do 2 contests, Q4 NGDP, and growth between 2014 Q4 and 2015 Q4.

    I’ll forward him the email and then add the questions. I presume this is something that he can set up easily with the link, say in less than an hour?

    my email address is Ssumner@bentley.edu

    Rajiv, Again, thanks for the help. Your comments are very important. We need to find out whether the donor funds could be used to subsidize the cost of allowing international withdrawals w/o closing the account.

    On the other point, I’d probably like to do both kinds of subsidy if possible, at least initially. Lower trading costs (or even zero), and subsidized payout. It’s very important to get off to a good start, so we don’t have bad PR for this experiment.

    I’ll tell you a funny story. I’ve made two bets at IEM in my entire life. In 2008 I bet about $60 on Obama. I won $100 and left the money in the account. In 2012 I returned to bet on Obama again, and was told I had about $60 left.


    It seems inactive accounts were charged a monthly user fee, and I had dropped back to about $60. So I bet on Obama again, and won again. This time I immediately withdrew my $100.

  38. Gravatar of Rajiv Sethi Rajiv Sethi
    5. October 2014 at 19:14

    Scott, the international withdrawal issue could be solved with a simple change in policy. No subsidy is needed since the international transfer fee is paid by the account holder in any case. I can’t see any rationale for the current policy unless they want to restrict international participation.

    Intrade used to mail out paper checks on demand for a small fee with a limit on withdrawal frequency. IEM allows free withdrawals by paper check up to twice a month. Easier withdrawal is necessary for broad participation.

    I’ve very surprised to hear about the IEM story since I’ve had long periods of inactivity in my IEM account over the past decade and never paid a penny.

  39. Gravatar of Mike Sax Mike Sax
    5. October 2014 at 22:56

    “The downside is that it also allows me to be somewhat more of a jerk than I am when actually meeting people face to face.”

    Still the guy with the biggest gap between his Internet and face to face persona is Stephen Williamson. I saw him on a video where he spoke somewhere at a conference and he was as agreeable as a church mouse.

    I kind of wished I was there so I could see what happened if I mentioned Krugman to him-maybe that would have bought Internet SW out.

  40. Gravatar of Njnnja Njnnja
    6. October 2014 at 05:25

    Following up on a comment from Thomas (and probably others). Make sure to do something to allow paper trading. If it isn’t already a part of iPredict, then it needs to be done in addition to iPredict. Your target customer (active, individual investors) are typically going to paper trade for a bit before devoting money to it. Good luck!

  41. Gravatar of Morgan Warstler Morgan Warstler
    7. October 2014 at 03:16

    Scott, I’m good for something. Where do you collect it?

    I’m glad thing have moved to the subsidy model, bc thats how the Fed ought to run it anyway as the transmission mechanism thru SMBs. Lose however much is necessary on pump months to the bettors, and hit them, but less than full amount when money needs to leave system.

    Question for traders: Assume the NGDP prediction market works, is there some weekly trade that Scott could execute based on the current prediction or change in the prediction?

    Meaning could Scott take the $2K for the quarter and throughout the 13 weeks, make 13 trades, that could increase or decrease the $2K?

  42. Gravatar of Scott Sumner Scott Sumner
    7. October 2014 at 08:54

    Rajiv, You said:

    I’ve very surprised to hear about the IEM story since I’ve had long periods of inactivity in my IEM account over the past decade and never paid a penny.”

    That makes me even angrier. Perhaps my account was too low, or inactive longer than yours. I will look into the international transaction fees.

    njnnja, I’ll pass that on.

    Morgan, I’ll let you know–we are still working out the details, which get more complicated every day. Mostly due to the elephant in the room, the US GOVERNMENT.

  43. Gravatar of Tim Cameron Tim Cameron
    7. October 2014 at 14:20

    Ask IG — an authorised futures dealer in Australia and NZ — to create the market for you? http://www.ig.com/au/welcome

  44. Gravatar of ssumner ssumner
    8. October 2014 at 04:46

    Tim, Thanks, but a for profit group won’t work. Are they for profit?

  45. Gravatar of Clayton Daley III Clayton Daley III
    8. October 2014 at 10:55

    What’s the problem with a for-profit operator? If you put the donations in a non-profit, I assume that non-profit could hire a for-profit to operate the exchange for you. That way you could access an existing platform (eliminating development cost) while ensuring your donations aren’t co-mingled with someone else’s profits.

  46. Gravatar of ssumner ssumner
    9. October 2014 at 05:06

    Clayton, Yes, that might work. But if you give directly to a for profit you have the principal/agent problem.

  47. Gravatar of Clayton Daley III Clayton Daley III
    9. October 2014 at 06:03

    I’d be shocked if you didn’t have at least one lawyer-reader who’d setup a non-profit for you. Then it’s just a question of finding the operator and designing the contract. For example, you could pay the operator:

    – A (low) fixed fee per month to maintain the market (so they don’t lose money if it doesn’t go anywhere) and
    – A fixed fee per $1 of US NGDP contract purchased (with some monthly cap)

    On top of any preexisting incentives to the operator (promotion of their platform to your users, spread on margin), they now have a direct, financial incentive to try and generate volume for you.

    You decide whether to specify how they spend the subsidy or let them figure out how best to encourage that volume… likely to include promoting the marketplace to existing users. Naturally, they’d be even more effective at persuading existing users if you could describe (in a white paper) who benefits.

    Of course, you also need to be the cynic and make sure the payments exclude trades that obviously “game” the system (i.e. they can’t just trade their own futures at the end of every month), but a good lawyer can help you with this.

  48. Gravatar of Ray Lopez Ray Lopez
    9. October 2014 at 21:12

    Three cheers for Professor Sumner! This man, like Schiller, deserves a prize! Single handed he: (1) created the notion of targeting NGDP, a subtle difference from targeting real GDP and/or inflation, unemployment and other metrics (2) he got the attention of the prominent economist blogger T. Cowen (3) he influenced the US Federal Reserve to consider, it seems, targeting NGDP (4) he steered clear of being labeled right-wing or left-wing; I was impressed since early on I figured Sumner was a flaming liberal of the Krugman type, but a more careful reading of his posts shows he is not (5) he brought honor and prestige to his current employer, a university in California, and, last but not least, this latest project, (5) he created a prediction market in GDP, the first one in the history of humankind.

    Huzzah!x3! Small but large point: iPredict has very low volume. I hope the GDP prediction works and I will do my best but it will be tough unless we all chip in, all of us reading this. With the low volume at iPredict, even a dozen active traders can actually make a difference.

  49. Gravatar of ssumner ssumner
    10. October 2014 at 07:24

    Clayton, Thanks for the idea.

    Ray, Thanks for the compliments, although Bentley is actually in Massachusetts.

  50. Gravatar of Ben Ben
    13. October 2014 at 11:30

    I’ve turned up late to this party and Rajiv has bottomed out a lot of the points I’d have on the exchange so I’m going to shift gears and suggest something else in addition to the market.

    Scott, alongside with your focus on NGDP your other constant and resonating message is the (seeming) failure of (some) economists to change views based on data (see hawk/dove fed policy post). So it got me thinking, why not have a ranking system for traders. Each trader when setting up their account can opt-in or opt-out for a league table in which they are assessed on % return (rather than total return to offer some protection for small cap traders) per quarter/per year/per forever. It would give both incentive (query if you use subsidy money to reward performance in this table a la bonuses as well as providing bragging rights) and also allow those with enough conviction/ego to put their reputations on the line? I’m not saying Krugman/Mankiew would sign up but surely their acolytes might; heck I’d even run a position based on their posts/stated position. Then we don’t just have an NGDP market but a market for ideas. Greed is good/an exceptionally good test of models

Leave a Reply