An unpublished letter to the LRB on high frequency trading

Lanchester, John. “Scalpers Inc.” Review of Flash Boys: Cracking the Money Code, by Michael Lewis. London Review of Books 36 no. 11 (2014): 7-9, http://www.lrb.co.uk/v36/n11/john-lanchester/scalpers-inc

Dear Sir,

It is striking for John Lanchester to claim that those who believe high-frequency trading is a net benefit to finance (and by extension, society) “offer no data to support” their views. Aside from the fact that he presents such views in the line of climate-change deniers, rather than a perfectly respectable mainstream view in financial economics, it doesn’t really seem like he has gone out looking for any data himself!

In fact there is a wide literature on the costs and benefits of HFT, much of it very recent. While Lanchester (apparently following Lewis) dismisses the claim that HFT provides liquidity as essentially apologia, a 2014 paper in The Financial Review finds that “HFT continuously provides liquidity in most situations” and “resolves temporal imbalances in order flow by providing liquidity where the public supply is insufficient, and provide a valuable service during periods of market uncertainty”. [1]

And looking more broadly, a widely-cited 2013 review paper, which looks at studies that isolate and analyse the impacts of adding more HFT to markets, found that “virtually every time a market structure change results in more HFT, liquidity and market quality have improved because liquidity suppliers are better able to adjust their quotes in response to new information.” [2]

There is nary a mention of price discovery in Lanchester’s piece—yet economists consider this basically the whole point of markets. And many high quality studies, including a 2013 European Central Bank paper [3], find that “HFTs facilitate price efficiency by trading in the direction of permanent price changes and in the opposite direction of transitory pricing errors, both on average and on the highest volatility days”.

Of course, we should all know that HFT narrows spreads. For example, a 2013 paper found that the introduction of an algorithmic-trade-limiting regulation in Canada in April 2012 drove the bid-ask spread up by 9%. [4] This, the authors say, mainly harms retail investors.

The evidence is out there, and easy to find—but not always easy to fit into the narrative of a financial thriller.

Ben Southwood
London

[1] http://student.bus.olemiss.edu/files/VanNessR/Financial%20Review/Issues/May%202014%20special%20issue/Jarnecic/HFT-LSE-liquidity-provision-2014-01-09-final.docx
[2] http://pages.stern.nyu.edu/~jhasbrou/Teaching/2014%20Winter%20Markets/Readings/HFT0324.pdf
[3] http://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1602.pdf
[4] http://qed.econ.queensu.ca/pub/faculty/milne/322/IIROC_FeeChange_submission_KM_AP3.pdf

Is Uber worth $18bn?

James Ball, at The Guardian, thinks that Uber’s implicit $18bn valuation is “a nadir in tech insanity”. His case is that tech firms are overvalued because although investors know this, they always assume there are other “suckers” they can palm their securities off on. That is, they think the other guys are “behavioural” (falling prey to the sorts of biases detailed in behavioural economics and behavioural finance) but they themselves are rational. Ball is responsible for some very good and important work, but I think this particular piece would benefit from the application of some financial economics.

It’s always possible that prices are irrational. And because we can never test investors risk preference separately from the efficient markets hypothesis (the idea that markets accurately reflect preferences and expected outcomes) it’s very hard to work out if prices are off, or just incorporating some other factor (usually risk). This is called the joint hypothesis problem. But when there are two alternatives, there is a reason economists put rational expectations in their models—it’s a simpler, better explanation. Finding truly suggestive evidence of irrational price bubbles is the sort of thing that wins you a Nobel Prize not something that a casual onlooker could easily and confidently observe.

Ball might say that even if irrational pricing is rare because of the strong incentives against it in a normal market, there have certainly been episodes of it in the past. Quoting J.M. Keynes, he might say “markets can remain irrational much longer than you or I can remain liquid”. He might point to the 1999-2000 peak of what’s commonly described as the “dot com bubble”. But I urge Ball to consider a point raised in this email exchange between Ivo Welch and Eugene Fama:

How many Microsofts among Internet firms would it have taken to justify the high prices of 1999-2000?  I think there were reasonable beliefs at the time that the internet would revolutionize business and there would be many Microsoft-like success stories based on first-mover advantages in different industries.

Loughran and Ritter (2002, Why has IPO pricing changed over time) report that during 1999-2000 there are 803 IPOs with an average market cap of $1.46bn (Table 1).  576 of the IPOs are tech and internet-related (Table 2). I infer that their total market cap is about $840 billion, or about twice Microsoft’s valuation at that time.  Given expectations at that time about high tech and the business revolution to be generated by the internet, is it unreasonable that the equivalent of two Microsofts would eventually emerge from the tech and internet-related IPOs?

Has not the second wave of cyber firm success (FacebookGoogle, arguably Apple) been even more impressive than the first wave? It may well be only 25% or 10% likely that Uber turns out to be one of these behemoth firms, through network effects, first mover advantages, name-recognition or whatever—but even if the chance is small the potential rewards are huge.

But Ball may point out that even if this is true, in the (putatively) 90% likely scenario, of Uber being a failure, then all this capital is being wasted. It could be put in the projects he prefers: “green energy, modern manufacturing, or even staid-but-solid sectors like retail”. Even if rational expectations—the idea outcomes do not differ systematically (i.e. predictably) from predictions—and the efficient markets hypothesis are not violated, and risk-adjusted expected (private) returns are equal across industries, it might be that social returns from these staid-but-solid sectors are higher—after all, lots of capital is being apparently wasted when so much goes to Uber.

This does not obtain—from the prospects of society, Uber could deliver huge welfare gains. If it does turn out that Uber has enough in the way of network effects to generate returns justifying its price tag (or more) then it would have to create lots of value, by saving taxi-consumers serious money. If they are using less resources to create the same amount of goods, then they are making society better off. Since society is big and diversified, it can afford to be relatively risk neutral (at least compared to an individual), and take even 9-1 punts on the chance that one memorable, semi-established network might be a particularly good way of running a taxi market.

Markets do set rates: A reply to Julien Noizet

Financial analyst and blogger Julien Noizet has replied to my article on mortgage rates on his blog. It is a good piece, worth reading, but I still think I am right. It is perhaps true that Noizet is right too, because my claim was really very modest: in total, mortgage interest rates do not mechanically vary with the Bank of England’s base rate; we can show this because the spread between them and the base rate varies extremely widely; and since we have very strong independent reasons to expect that market forces largely drive rate moves, that should be our back-up explanation. The implication of this I was interested in was that this meant a hike in Bank Rate wouldn’t necessarily drive effective rates up to a point that would substantially increase the cost of servicing a mortgage and hence compress the demand for (London) housing.

Even if the first graph in Noizet’s blog post did appear to support his narrative that effective market rates follow Bank Rate moves, I’m not sure why these disaggregated numbers matter given that the spread between overall effective rates on both new and existing mortgages varied so widely. If it turns out that specific mortgage types varied closely with Bank Rate but the overall picture did not, then markets still control effective rates, they just do it via a changing composition of mortgages, not by changing the rates on particular products. The effect is the same—and it is the effect we see in the Bank’s main series for effective rates secured on dwellings. But the graph, to me, looks a lot like mine, despite the effect of new reporting standards: mortgage rates are about a percentage point from the base rate until 2008, then they don’t fall nearly as far as the base rate in 2008 and they stay that way until today. If other Bank schemes, like Funding for Lending or quantitative easing were overwhelming the market then we’d expect the spread to be lower than usual, not much higher.

His second big point, that the spread between the Bank Rate and the rates banks charged on markets couldn’t narrow any further 2009 onwards perplexes me. On the one hand, it is effectively an illustration of my general principle that markets set rates—rates are being determined by banks’ considerations about their bottom line, not Bank Rate moves. On the other hand, it seems internally inconsistent. If banks make money (i.e. the money they need to cover the fixed costs Julien mentions) on the spread between Bank Rate and mortgage rates (i.e. if Bank Rate is important in determining rates, rather than market moves) then the absolute levels of the numbers is irrelevant. It’s the spread that counts. But the whole point of my post is demonstrating that the spread changes very widely, and none of Julien’s evidence seems to me to contradict that claim. Indeed, Noizet’s very very good posts on MMT, which stress how deposit rates are much more important as a funding cost than discount rates for private banks, seem at odds with what he’s written in this post. And supporting this story is the fact that the spread between rates on deposits (both time and sight) and mortgages changes much less widely. If we roughly and readily average time and sight on the one side and average existing and new mortgages on the other, the spread goes no higher than 2.3 percentage points and no lower than 1.48.

In general with the post I don’t feel I understand the mechanisms Noizet is relying on, perhaps I’m misunderstanding him, but the implications of his claims regularly seem to contradict our basic models of markets. For example, he says that a rate rise would lead banks to try and rebuild their margins and profitability. But I can’t see any reason why banks wouldn’t always be doing that. The mortgage market is fairly competitive, at least measured by the numbers of packages on offer and the relatively small differences between their prices. I don’t think Julien has presented any mechanism to suggest why banks would suddenly want to maximise profit after a rate rise but wouldn’t beforehand—or why they’d suddenly be able to ignore their competitors but couldn’t beforehand. It’s possible there is one, but I can’t see that he’s explained it. Overall I suspect I’ve missed something crucial, so I welcome any more comments Julien has on the issue.

There’s no such thing as a free minimum wage hike

Paul Kirby, who was head of the No. 10 Policy Unit until last year, has a long post calling for a “dramatic, historic increase” to the minimum wage, bringing the levels from the current £6.10/hour to £10/hour in London and £8/hour in the rest of the country. It’s a bold post, but ultimately most of his arguments fail. In this post I try to address the key points he makes in favour of a hike.

Low wage earners are, overwhelmingly, providing services for domestic consumers within the UK economy. They work in shops, cafes and hotels. They cut our hair, they clean our houses, they look after our kids and they care for our elderly.  They are not  in manufacturing, competing on the price of their labour with other countries. What they do has to be done in this country. Nor is it tradable with other countries. If the Minimum Wage increases, it impacts equally on all of an employer’s competitors, so there is no disadvantage.

Even though nobody can switch to a cheaper hairdresser in India, they can get their hair cut less often, or have their homes cleaned less frequently, or send their children to creches with fewer minders per child or their parents to care homes with fewer carers. Kirby is assuming that demand for domestic services is inelastic – that is, it does not change much according to price. Obviously, this may differ between different services, but in without evidence to the contrary (Kirby gives none) it does not seem reasonable to assume that people’s demand for services will stay the same even if the prices of those services rise.

Bear in mind that a minimum wage increase would only affect the bottom of the market, where you would expect customers to be the most price-sensitive. The economic evidence suggests that increases in the minimum wage lead to slower job growth, particularly for young workers and in industries with a high proportion of low-paid staff.

Raising the lowest wages does not mean that employers simply have to, or will, just cut jobs or working hours to keep the wage bill constant. The evidence is clear that employers find a variety of solutions.  Firstly, they restrain pay growth for their better paid staff. Secondly, they increase prices to consumers. Thirdly, they improve productivity and get more out of each hour that they are paying for. And then they squeeze their profits. Through productivity gains, they either earn more revenue or cut the amount of labour they need.

Employers do not try to ‘keep the wage bill constant’. They try to make a profit on the labour they hire. If hiring an extra manager led to extra profits, it wouldn’t matter that doing so also increased the overall wage bill. A minimum wage imposes a price floor on labour, so any worker whose total productivity is less than the minimum wage floor represents a net loss to their employer – which a profit-maximising firm will respond to by firing the worker. It makes no difference whether or not that firm has ‘restrained pay growth’ for its other workers: if an employee is loss-making at the lowest wage a firm can pay them, a profit-maximising firm will fire them. (Or simply not hire additional workers who would be loss making on net.) Even if firms can only tell the average productivity of their workers, because of information problems, they will demand less labour in total.

On the possibility of raising prices to make the worker profitable, see the previous point: if demand for the service is price inelastic, this might work, but it’s quite a claim to say that this is the case for most minimum wage-supplied labour.

Wages are not the only cost of labour to firms, either. Firms may reduce costs in response to minimum wage increases by cutting back on perks like lunch breaks and sick leave, as Starbucks did after it agreed to pay additional corporation tax in 2012.

Increasing low pay has a limited impact on the overall costs of most businesses. In some sectors, very few earn less than the living wage, e.g only 6% in manufacturing. Even in hotels and catering, which is one of the biggest sector for the Minimum Wage, only 17% of jobs are below the living wage and raising the Minimum Wage to the Living Wage would only add 6% to the wage bill. This is the highest impact for any sector. More importantly, labour is only a proportion of all costs, e.g. 25-35% for restaurants.

Is a 2.1% increase in costs for labour-intensive firms not something to be concerned about? The fact that ‘most businesses’ would not be affected seems beside the point. (The reverse of this is true too: if Kirby’s other points were correct, would his suggested minimum wage hike be a bad idea because it would affect “only” 17% of workers?)

There is no real evidence of any minimum wages in the world adversely effecting employment levels.

This is totally wrong. In 2006 Neumark and Wascher reviewed over one hundred existing studies of the employment impact of the minimum wage. Of these, two-thirds showed a relatively consistent indication that minimum wage increases cause increases in unemployment. Of the thirty-three strongest studies, 85 per cent showed unemployment effects. And “when researchers focus on the least-skilled groups most likely to be adversely affected by minimum wages, the evidence for disemployment effects seems especially strong”.

Few people stay on low-wage jobs for their whole lives: minimum wage work is usually a stepping-stone to something better where employees can acquire human capital. There is evidence that suggests that minimum wages deter young workers from acquiring these skills that allow them to get better jobs in the long run. Note also that minimum wages have been used explicitly to kick away the ladder for minorities: by whites in pre-Apartheid South Africa; by anti-Hispanic campaigner Ron Unz in California; and by, er, Polly Toynbee in a recent Guardian column.

Tyler Cowen reminds us to make sure our views of sticky wages and minimum wages are consistent: if “worker-imposed minimum wages” (sticky wages) lead to unemployment, as most Keynesians (among others, including me) believe, why would “state-imposed minimum wages” not also do so? (“Have you no respect for the law (of demand)?”, asks Will Wilkinson.)

Given that we know that minimum wage increases usually cause some unemployment, why take this chance when we could just give money to poor people directly? As we’ve been saying for years, the difference between the current pre-tax minimum wage and the post-tax “living wage” is roughly as much as a minimum wage worker pays in income tax and national insurance: in other words, if that worker didn’t pay tax, they would be earning a living wage. It looks as if the personal allowance will soon rise to the minimum wage level, but the national insurance contribution threshold needs to rise too.

But let’s go even further: if we replaced the tax credit and welfare systems with a Negative Income Tax (or Basic Income – call it whatever you want), we would top-up the wages of low-paid workers directly. Jeremy Warner calls for this in the Telegraph today, and I outlined something similar a few weeks ago. Yes, I’d like all the standard supply-side deregulations as well, but a Negative Income Tax would act as an insurance policy against the potential down-sides of such deregulations, strengthening workers’ bargaining power and addressing the fears of those who worry that deregulations will hurt some workers.

I understand that many Conservatives are coming to see a minimum wage hike as a political ‘free lunch’ – a popular and surprising way of showing an interest in the welfare of the poor that does not affect the government’s balance sheet. I hope this is not true. Contrary to Kirby’s claims, there are good empirical and theoretical reasons to think that raising the floor on the price of labour will cause more unemployment. And unemployment destroys lives. There are lots of things we can and should do to help the poor right now. Raising the minimum wage isn’t one of them.

Old Economy Steven would have been better off now

An interesting essay from Chris Maisano over at Jacobin Magazine drifts over many topics—full employment, growth since the 1970s and neoliberalism, worker activism and the 40-hour week. Its essential case is that full employment is important, because it makes workers better off in lots of ways, including giving them more leisure time. There are some interesting points in the piece, and I agree that full-ish employment is an important goal, but overall I think it rests on a huge number of misconceptions—indeed data is used in very weird ways, with what I see as obvious questions left entirely uninterrogated.

Maisano points to the “Old Economy Steven” meme, which looks back to an idealised post-war era:

Steven pays his yearly tuition at a state college—with his savings from his summer job! He graduates with a liberal arts degree—and actually finds suitable entry-level employment! … But Steven doesn’t just enjoy the material comforts of Old Economy abundance. He possesses a degree of everyday power scarcely imaginable by working people today. Steven can tell his boss to shove it, walk out and get hired at the factory across the street.

The contrast with popular views about today’s economy, at least since the recession, is obvious. But full employment policies have been demoted—indeed since the late 1970s and especially since central bank independence most developed countries have centred their macroeconomic policies around stable inflation, not high employment. In fact, central banks now see a Non-Accelerating-Inflation Rate of Unemployment (NAIRU) as the optimal situation. But is this an “ideological response” as Maisano suggests?

There will always be some unemployment, from the numerous supply side restrictions on labour, and from job switching, especially with sectoral shifts. Inducing unexpected inflation can temporarily take unemployment below this “natural” level, for example through money illusion—where workers think nominal pay is actually real pay—but it is unsustainable. Once unions and individual workers compute this level of demand growth into their calculations the natural rate will return and the monetary authorities will need to push inflation yet higher to subvert this equilibrium.

Many economists, including Milton Friedman, argue that something like this caused the rampant, out of control inflation of the 1970s, something that was only reigned in by harsh recessions in both the UK and USA (attempting to control wages and prices was an abject failure everywhere). Acknowledging this means acknowledging that aiming for unemployment as close as possible to zero is a bad idea; it is better to aim for the lowest level of unemployment achievable without acceleration inflation. It’s certainly possible to argue that monetary policymakers have failed to do this—but it hardly seems like a specifically ideological development, more like progress in economics.

A second sticking point is how growth has declined since neo-liberalism replaced the post-war consensus as the dominant political framework in at least the US and UK. This is true. But it’s also true that every developed country saw a growth slowdown in the 80s and 90s relative to the post-war era. Economic historians are divided on the causes but since the most neo-liberal countries grew much faster than the more left-leaning states, one’d be hard placed to see that as a key cause. But even though growth has slowed down it has not stopped—and despite a few bumps we are much much richer today than in the 1970s. Just think, if had the opportunity to be whizzed to the 1970s to have the same standard of living as someone in your income percentile did then, would you?

My third disagreement is on hours worked. Maisano heavily implies that the consistently looser labour markets since the 1960s and 1970s have resulted in workers forced to work longer hours. He’s clearly looked at the numbers, since he compares the US’s average 1,778 in 2010 (1,742 on the FRED numbers I’ve seen) worked unfavourably to “continental European and the Scandinavian social democracies”. But is that a germane comparison? To me it seems like the best way to compare the wellbeing of workers now, following decades of neo-liberalism and below-full-employment, and workers then, is to directly compare them. On average, during the 1970s, an employed person worked 1,859 hours (in 1970 it was 1,912 hours), in the ten years up to and including 2011 the average was 1,772.9. Maybe Maisano believes that with a greater focus on full employment incomes would have grown even more and hours would have fallen even faster—but if he thought that maybe he should say it.