...Archive for October 2015

Advice for Twitter

In my accustomed role of moralist scold, I probably shouldn’t want to help Twitter. I should want it to be replaced with an open-source, distributed, decentralized ÐAPP or something. And at some level I do want that. But for now, Twitter is the only social whatnot I actively use, and when I don’t hate the service I sometimes really like it.

I keep reading articles about how Twitter is dying or failing or crashing or whatever. I have ideas! I don’t know whether anything I propose might justify Twitter’s current or recent-past stock price. I could care less. But here are some things I would do to make Twitter a better business, in multiple senses of the word better.

  • Ditch the whole ads and analytics model Twitter currently aspires to. It’s an evil business model, no matter how well it works for companies that once promised not to do evil. Putting aside the moralism, Facebook and Google pretty much have a lock on that gold mine. Apple, not a don’t-be-evil company by any measure, has seen the writing on the wall and chosen to differentiate itself by not spying on people. Twitter should do the same. Just stop it, and make a big show of stopping it. Twitter, to its credit, has a decent record overall of standing up for people’s rights to use its platform for controversial political speech. It should keep doing that, stop doing this, stop spying, and make respect of user privacy an integral component of its brand.

  • Charge the people who get the most from Twitter. Don’t give “producers” micropayments, take their money. People who interact on Twitter, especially those who gain a large-ish following, get a lot more out of Twitter than people for whom it is just one more passive information source. This is tricky: if people have to pay to tweet from the start, they would never get enmeshed in the service deeply enough for it to become worth paying for. What Twitter needs is a payments model that lets casual users Tweet for free indefinitely but cajoles power users into paying to support the business.

  • That business model exists. It falls right out of the experience of power users. Over the years, I have wasted tens of hours trying to shave a few characters off some vain idiocy I imagined to be clever and witty in order to conform to the 140 character limit.

    It would be a terrible idea for Twitter to let users pay to escape the character limit arbitrarily. Twitter’s raison d’etre is the micro in microblogging. No one should be embedding essays into people’s timelines. However, it would not kill Twitter if I could pay a little to slip in an extra few characters rather than waste five or ten minutes reworking, abbreviating, or mangling the perfect quip that happens to clock in at 143.

    My suggestion is exponential pricing. One extra character, a tweet length 141, costs 1¢. Two extra characters costs 2¢. Three extra characters costs 4¢. n extra characters costs 2n-1 cents. By 28 extra characters, the cost spirals to more than $1M dollars. I think it’s fair to say that this pricing model would enable people to go a few characters over without worrying about it while keeping intact the “around 140 character” user experience. I think I’d end up paying $20-ish a month under this model, but I’d be grateful for the time saved and the preserved quality of expression relative to my current use of the service. And of course, users would continue to have the option to treat 140 as a hard limit and pay nothing. The tradeoff between perfectionism and money would be entirely at each users’ discretion.

  • Twitter’s default algorithm should remain real-time, reverse chronological feeds whose contents are fully determined by users’ choice of followers. But critics are right to point out that the standard feed may not provide the best curation for everyone. Since Twitter would be out of the corrupt and corrupting business of selling eyeballs for ads, it could leave the choice of alternative curations entirely at each user’s discretion. Twitter could offer a menu of algorithms that it thinks will be useful to different categories of users, even impose a default (in a tab distinct from the standard feed) that users would be able to replace or augment. Twitter could define an API through which outsiders could publish their own algorithms for curating timelines that would then become available to users in a kind of (free) “app store”.

  • It should be possible to Tweet stuff to people. Real stuff, goods and services. Twitter should try to earn users’ trust and collect meatspace identities and addresses while promising to hold them in strict confidence. Without necessarily knowing the true identity of the recipient, Twitter users should be able to send gifts to Twitter handles. Twitter would partner with various merchants to define the range of available goods, and work with them to fulfill.

    Since Twitter itself would know the transacting parties, it would be able to deter abuse. Individual users could define whether they accept gifts and from whom. Only from people I follow? A special Twitter list? Anyone and everyone? Users decide. Users could also decide whether to let gifts surprise them at their doorstep, or could require on-line approval before an order is processed. Obviously, this would be a very easy service for Twitter to monetize. They would just take a percentage of the cost of the gift as a service fee. Gifts might be represented by a special kind of glyph in tweets (that one can click to inspect), and might be included in both public tweets and direct messages. Attempts to publish Tweets with gifts to users not configured to accept them should be blocked, and Tweets with gifts that recipients will screen before accepting might be withheld from publication until the gift has been approved. There should be modest limits to the dollar value of tweetable gifts individually, and to the total value transferred from one person to another over a period. The intent would not be for Twitter to become a payments system or a means of sending large bribes. Over the years, I’ve wanted to send people books, a delivery of chicken soup, a stuffed animal or a card. Obviously digital goods would be easy, gift certificates or music or an ebook.

    As with exponential pricing of characters, no one should be forced from their existing practices by this new option. Users who don’t wish to send or receive gifts could continue to use the service as they currently do, without providing any identifying information to Twitter. They would remain as pseudonomous as they currently are (perhaps even moreso, if Twitter flamboyently forewent the kind of spying on users that is now the standard practice of ad-supported sites).

  • Twitter should go back to making itself open to outside developers and third-party clients. I am delighted to read that it plans to do so. Obviously, Twitter has a credibility problem, having crushed outside developers when it decided it wanted to gather eyeballs to ads and control what they see. The service behaved abysmally. Twitter should do everything it can to commit to never doing anything like that again. Having a business model that makes money independently of how people encounter or publish tweets would render their devotion to a new glasnost less suspect. If there are things they will need to restrict (like, say, a free tweet-longer service embedded in clients), they should think about that carefully and publish those restrictions in advance.

Translating “net financial assets”

Steve Roth at Asymptosis offers a remarkable, detailed discussion of Modern Monetary Theory’s notion of “private sector surplus” with an emphasis on aggregate accounting. Roth’s core point is well taken: “Private sector surplus” (equivalently the increase in “private sector net financial assets”) should not be conflated with the economic saving of households. As Roth points out, household sector saving is the difference between household sector income and household sector (noninvestment) expenditure. “Private sector surplus” is likely to increase household sector income, and so in that sense it forms a component of household sector saving, but it is quantitatively small relative to total household income — especially, as Roth emphasizes, if you use comprehensive measures of income that include capital gains and losses. [1]

Roth is right about all this. But I think he is talking past MMT economists a bit. Roth invites us to think about comprehensive saving by households. But that’s very far from what MMT’s baseline sectoral balances decomposition claims to capture. Instead “net financial assets” capture only the financial position of the aggregated (domestic) private sector, including both households and businesses. MMT enthusiasts sometimes mix these things up, and when that happens it should be called out. But this confusion has been called out a lot over the years, and I think for the most part MMT economists have become pretty precise in expressing themselves. There is a great deal of value in the MMT decomposition, not as a measure of household saving, but of something else entirely. Let’s try to understand it.

I’m going to steal from my own, old post, a derivation of the MMT-standard decomposition (usually attributed originally to Wynne Godley). We start with a tautology:

Every financial asset is also some entity’s liability. The sum of all financial positions is by definition zero. So we can write:


Suppose that, quite arbitrarily, we divide the world into a “foreign” and a “domestic” sector. Then we have:


Suppose that, again arbitrarily, we decompose the domestic economy into a public and private sector:


Substituting into our previous expression, we get


We can also write this in terms of changes or flows. Since the sum above must always be zero, it must be true that any changes in one sector are balanced by changes in another:


Two of the flows in the equation above have conventional names, so we can rewrite:




The highlighted Equation 8 is where the action is. NET_PRIVATE_DOMESTIC_FINANCIAL_POSITION is often described as net financial assets of the domestic private sector. MMTers (domestic) “private sector surplus” is precisely ΔNET_PRIVATE_DOMESTIC_FINANCIAL_POSITION.

The crucial thing to understand is what the net means in net financial assets. It is precisely financial savings net of domestic real investment by the private sector. It is the farthest possible thing from a comprehensive measure of household savings. It is private sector savings excluding the vast preponderance of household savings, which is backed by private sector assets (whether owned by households directly or owned by businesses who then issue financial claims to households). “Ordinary” private sector savings either doesn’t show up as financial assets at all (a home without a mortgage is just a real asset owned by a family, like a television or a baseball card), or else they “net out” when we aggregate, because one private sector entity’s asset is precisely extinguished by another private sector entity’s liability. If a household is “long” a share of stock, a firm is “short” that same position, and owes the household whatever that claim represents. The aggregate financial position of the private sector combines the financial positions of businesses and households, so the financial claims of households against firms are matched by mirror image liabilities of firms to households. They annihilate one another like matter and antimatter.

So why do we care about this odd sliver of savings? Why do MMT economists make it so central to their analysis? Private sector net financial assets are “special” precisely because they are not backed by domestic real assets, but instead by promises that are credibly independent of domestic real asset values, especially promises of states. Saving that takes the form of real stuff, whether that stuff is directly held or hidden behind financial claims, is inherently risky. House prices fall. If you own a factory, or shares in a firm that owns a factory, the factory can burn down. Even if you hold a diversified stock portfolio, you will find it subject to wild swings in value. If you own private sector debt, you expose yourself to credit risk. If you own a diversified portfolio of domestic stocks and bonds, your own circumstances and that of your investment portfolio will be correlated in an unpleasant way. The times when you lose your job and need to draw on savings are likely to be the same times when stocks have crashed and people are defaulting on their debts. People desperately covet assets that are divorced from the risks of the domestic real economy. And that is precisely what “net financial assets” are.

Net financial assets are special, because they serve insurance functions that assets produced by the domestic private sector simply cannot provide. When households are risk-averse, they covet these assets especially. For firms, these assets offer protection against insolvency risk that real assets, whose values both fluctuate idiosyncratically and covary with the real economy, cannot provide. MMT economists often suggest that if the public sector fails to accommodate the private sector’s appetite for net financial assets, recession and financial instability will result. That makes sense. It’s conventional, if a bit vapid, to describe recessions as times when “animal spirits” are low, when people are risk averse. But what matters is not the courage in people’s hearts (or lack thereof). What matters is how people behave. If people’s behavior is counterproductively risk averse, you can encourage greater risk-taking by offering insurance. That’s precisely what injections of “net financial assets” into an economy provide. If firms are teetering on the brink of bankruptcy, you can flood the economy with safe assets they can use to shore up their balance sheets to reduce their risk of default. That’s precisely how the United States saved its banks in 2008 (for better or for worse). The headline bailouts and TARPs and accounting forbearance were all expedients to keep those firms alive until a flood of assets immune to correlated private sector collapse could find their way onto bank balance sheets (with the help of opaque subsidies). Those special assets are “net financial assets”.

“Net financial assets” are a heterogeneous category. They include both claims against the domestic state and claims on foreign public and private sectors. A claim on a foreign firm in foreign currency does not provide the same insurance as claim against the domestic government in domestic currency. Nevertheless, claims on the foreign sector do provide insurance against domestic shocks that do not impair the foreign counterparty. And note that contrary to naive financial theory, which predicts developed economies will net-accumulate claims on emerging economies to invest in their growth, in practice emerging economies tend to net-accumulate claims on developed economies. The insurance function of safer foreign assets outweighs the investment function of accepting foreign capital (or at least it has since the Asian Financial Crisis). For firms and households in an emerging economy, foreign claims and claims on government are both useful insurance. In developed as well as emerging economies, negative positions with respect to foreign creditors increase the domestic private sector’s exposure to risk as surely as indebtedness to the state would, assuming debt contracts are uniformly enforced.

All this terminology — private sector surplus, net financial assets, etc. — is associated with heterodox, lefty MMT, but it maps very nicely to discussion of “safe asset shortages” in the mainstream financial press or Gary Gorton’s schtick on the importance of “informationally insensitive” assets. The main difference has to do with whether we can or ought to rely upon the domestic private sector to produce these kinds of assets. The MMT analysis, by construction, excludes private sector “triple-A” assets, where people like Gorton emphasize a role of private sector in producing assets that might provide this sort of insurance. The MMTers have it right. The domestic private sector simply cannot produce assets that provide insurance against systematic risks of the domestic economy without the help of the state. (Gorton tacitly recognizes this when he suggests the state should supervise and guarantee assets produced by shadow banks like it insures bank deposits. No thank you.)

The insurance function of “net financial assets” is not unambiguously a good thing. Net financial assets are special precisely because they provide insurance against systematic risk. When net financial assets are claims on foreign debtors, they are not so problematic, they just represent a form of diversification that can insure against domestically (but not globally) systematic shocks. Claims against the domestic state, however, offer safety to their holders in a manner that can be quite dangerous to the rest of us. “Insurance” against a truly systematic shock is necessarily a zero-sum game. If we are all collectively poorer, the only way the state can make some claimants whole is by shifting their share of the aggregate loss to people who don’t hold the government’s promises. We’ve experienced this very painfully over the past decade, as both the European and American policymakers refused to accept any risk of inflation (thereby prioritizing the value of past promises). Policymakers chose to make absolutely sure that holders of state assets would be made whole in real-terms, and imposed severe costs on debtors and the marginally employed to do so. (I think policymakers overshot the inherent zero-sum-ness of providing insurance during a systematic shock and have played a sharply negative-sum game.) It would be better, I think, if states downgraded the insurance they provide by weakening the promise they make to asset holders from price stability to an NGDP path target. And I worry much more than I think most MMT economists do about the unjust distribution of risk-bearing that might accompany a large stock of net financial assets very unequally distributed. (Unusually, I’m with Greg Mankiw on this one.) I think the economy includes people who are already overinsured by their stock of net financial assets, and those people tend disproportionately to accumulate new issues. So we should think more about how we can accommodate private sector entities’ need for some degree of insurance by redistributing existing net financial assets rather than creating new ones.

This sentence is a pithy conclusion.

[1] Should you? Or should you use a NIPA-style accounting that “looks through” capital gains? That’s a complicated question. Looking through capital gains entirely is clearly unsuitable, because household capital gains include revaluations of shares due to e.g. retained earnings, which represent increases in the quantity of real assets that households’ shares lay claim upon. This kind of capital gain is economic saving. But what about price appreciation of the existing housing stock? On the one hand, this is certainly perceived by individual households as real wealth and a form of savings. On the other hand, housing price increases also represent a kind of liability on the part of households and households-to-be who are not yet homeowners. If our object is to study distributional questions, differences between households, capital gains of this sort should obviously take center stage. They are real wealth to the households who enjoy them, and often represent costs in some form to households who don’t. But if we are aggregating, it’s not as clear that mere repricings of existing assets should be included in household saving. Unfortunately, it’s almost impossible in practice to disentangle gains due to real growth in the assets backing claims from repricings of existing assets. Consider our prototypical example of “mere repricing”, price appreciation of an existing home. If a home remains entirely unchanged in an unchanged neighborhood in an unchanged city, then that is mere repricing. But perfect stasis is impossible outside of a thought experiment. If an “existing home” is renovated, and its price appreciates, how much of the price appreciation is “mere repricing” and how much of it reflects the change in real assets? If the neighborhood surrounding a largely unchanged home improves dramatically, then the house is a more efficiently deployed real asset, and a change in price may reflect new real value, which does constitute a form of economic saving — a claim on real growth — rather than mere repricing. Similarly, while stock appreciation can reflect an increase in the quantity or real-economic usefulness of the assets backing shares, it can also result from a mere revaluation of largely unchanged firms. When the Fed eases, stock prices rise, but the real assets that back them are not meaningfully improved. I think on the whole Roth’s choice to include capital gains in his analysis of aggregate household saving is right, much better than the alternative choice of ignoring it. But neither choice can be “correct”.

Update History:

  • 24-Jul-2015, 1:15 p.m. PDT: “Note that the The insurance function of “net financial assets”…”