Rise of the machines

When Dr. Tuomas Sandholm’s “Libratus” software defeated four champion poker players in Pittsburgh this February, it suggested that machine intelligence may have finally evolved to the level where it can excel in imperfect information scenarios.

The implications of this win and similar advancements are widespread, and when coupled with the steadily increasing prevalence of ‘big data’ and alternative data sets in investing, suggest that the active management investment universe is undergoing a dramatic evolution.

To better understand the resurgence of interest in the systematic hedge fund (HF) space, from the perspective of both managers and investors, our Strategic Consulting team analyzed data from 64 hedge funds across systematic, discretionary, and hybrid strategies; and 25 investors representing ~$240bn in total HF assets. The objective was to present an overview of the growth trajectory and business/product mix of systematic hedge fund strategies; measures of systemic risk; fund performance across market conditions; and the use of big data throughout the investment process.

Systematic Strategies within the Hedge Fund Industry

The study estimates the “systematic strategies” segment of the hedge fund industry at roughly 17% of the total industry size, comprising approximately $500 billion in assets under management, the bulk of which (~$450bn) is found in stand-alone systemic products and the remainder of which (~$55bn) are allocations within larger discretionary multi-strategy funds.

Size of Systematic Strategies within the HF Industry

Source: 2017 HFI, HFR, review of manager marketing materials and publicly available documents, Strategic Consulting analysis. 

* CTA – Commodity Trading Advisor; ** We estimate that 70-80% of these strategies are managed futures, with the remainder in EMN.

The Strategic Consulting team also found a meaningful increase in investors’ levels of comfort with systematic strategies in recent years. Traditionally, systematic strategies have been more difficult to understand and less conducive to transparency as a result of their secrecy and/or complexity. However, this greater acceptance of systematic strategies has been a driver of inflows to systematic managers.

Comfort with Allocations to Systematic Strategies

Source: 2017 Strategic Consulting analysis; 2015 data from Strategic Consulting Report “Bracing for Impact” and 2016 / expected 2017 data from Strategic Consulting report “Turning the Tide” .

At the same time, this growth has led to concerns around crowding in equity quant strategies in particular and the potential for another 2007-style ‘quant crash’. That said, the team’s measures of risk suggest that these concerns may be overblown.

Big Data and Machine Learning 

Recent advancements in ‘big data’ and ‘machine learning’ have also driven change in how systematic managers incorporate data, technology and analytics in their investment process. The Strategic Consulting team found that 54% of the systematic managers in their sample are now employing alternative and ‘big data’ sources such as web scraping (i.e., a technique to extract large amounts of data from websites, social media data, satellite data and credit card data). Additionally, 62% of the systematic managers are using machine learning techniques within the investment process.

'Big Data' - Current Usage

Source: 2017, 1. Strategic Consulting survey results and analysis (3Q16-1Q17) – refers to % of interviewed managers incorporating each data source in their research process. 2. IDC

Among discretionary managers, 24% are using ‘big data’ and approximately 50% have incorporated quantitative techniques across the information gathering, idea generation, portfolio construction / risk management, and performance analysis components of the investment process.

Quantamental - Use of 'Big Data', Quant Techniques

Source: 2017 Strategic Consulting survey results and analysis (3Q16-1Q17) 

1. Range of 45%-55% covers responses across all four categories (info gathering, idea gen, port construction / risk mgt, performance analysis)

Final Considerations

In general, while the expectation of the managers and investors interviewed was that systematic strategies and techniques will continue to grow in importance within the hedge fund industry, the team found that there are areas where human judgment is likely to retain the advantage for the foreseeable future, and a hybrid approach – a blend of systematic and discretionary approaches – may well be the most popular in the long-term.

Banking Reform in China: Too Little, Too Late?

China’s economic expansion requires a responsive and responsible banking system to keep money flowing. This means financial reform in how China does business or else face severe financial implications having global consequences.

China’s economy has grown in leaps and bounds for a number of years. While it is ambitious, there are dangers along the way that could spell problems for China and the regional and global economy. China must catch up with the economic superpowers of the United States, Japan, and the European Union, but to do so, it must make key reforms in its financial industry or else lay the foundation for economic disaster. More specifically, there must be reforms in China’s commercial and shadow banking sectors. If these reforms do not occur or are neglected for an extended time, China is trading short-term economic growth for long-term financial calamity. The key question is: Is it too little, too late for these reforms?

The Need for Commercial Banking Reform

One area of reform is China needing more independently-owned banks. China’s banking industry is dominated by state-owned commercial banks that have historically funneled financial capital into government run projects including state-owned enterprises (SOEs). While this may help to spur on China’s economy by providing jobs and financial growth, many SOEs lose money and would probably be refused loans by banks elsewhere, globally. The problem is that the Chinese government is building up certain companies and industries while discouraging others, rather than letting free-market forces make that decision.

Read Global Liquidity Reaching a Tipping Point

China has concentrated its banking industry into the “Big Four”. These are the Industrial and Commercial Bank of China, the Bank of China, the Agricultural Bank of China, and China Construction Bank which are responsible for almost 50 percent of all China’s loans. This dangerous oligopoly provides for significant market distortions while not allowing for competition by foreign or domestic banks, especially those privately owned.

This also hurts the efficiency and effectiveness of the Chinese banking industry since less profitable banks thrive by making loans to poorly run SOEs. The Big Four also blocks out smaller and mid-size banks since depositors perceive them as more stable and less risky. This creates a false sense of security for depositors until there is a major loan default by an SOE.

China also has a problem with its central bank, the People’s Bank of China (PBOC), setting a ceiling on retail deposit interest rates and a floor for lending rates. There must be interest rate liberalization for China’s banking system by the market rather than set by the PBOC. Attempts were made within the past year to relax interest rate restrictions by having a variable interest rate for deposits.

In 2014, the PBOC permitted banks to set deposit rates at 20 percent instead of 10 percent. The PBOC has also given banks flexibility to increase deposit interest rates from 1.3 to 1.5 times the benchmark rate. The problem remains that the full interest rate liberalization is only a pipe dream. There must also be market-based prime lending rates as a benchmark rate. This can encourage more savings by depositors and reward businesses that are run profitably.

Reforms for the Shadow Banking Sector

One area that sorely needs reform is China’s shadow banking sector. Shadow banking allows Chinese firms to acquire financial capital without concern of financial regulations or governmental constraints. If a company needs funds for expansion, or to stay afloat financially, shadow banking involves non-bank financial intermediaries providing capital outside of governmental oversight.

On the one hand, Chinese businesses can acquire necessary funds when needed. However, as an unregulated industry, many feel this could lead to financial disaster if a loan default or that a third party must suddenly come up with money to guarantee a debt. Moody’s estimates China’s shadow banking industry at $8.5 trillion.

China’s policymakers recognize the problems and risks of shadow banking and are attempting to address them. President Xi Jingping has acknowledged the risks associated with shadow banking by stating in October 2013: “. . . we are soberly aware of potential problems and challenges from falling demand, overcapacity, local debts and shadow banking, and we are paying close attention to possible impacts coming from the outside.” Government officials are attempting to place controls and workable regulations on shadow banking as evidenced by the actions of The China Banking Regulator Commission (CBRC).

The CBRC is trying to determine if banks have been using shadow banking products in order to cover up loans to money-losing “zombie firms” or businesses in government-restricted industries. In the first three months of 2017, the CBRC made fines of 190 million yuan, or 25.22 million euros, from 485 investigations. The CBRC seeks to construct a firewall between Chinese banks and trust companies in order that trust companies will incur market risks themselves, including business default risks that may occur in economic recessions.

President Xi Jingping appointed Guo Shuqing as CBRC chairman in 2016, who is highly regarded as a knowledgeable technocrat and former Shandong Province governor, China’s third-largest provincial economy. Guo has been aggressive as CBRC chair by instituting new rules and regulations in China’s banking industry.

Whether China’s bankers like it or not, they now face stricter lending standards, monthly lending ceilings, and newly imposed disclosure requirements regarding off-balance sheet assets. Guo has even ordered financial institutions selling wealth management products to clients that they cannot make payments on investments losing money or keep them on their balance sheet.

In February 2017, rules were drafted preventing banks from investing in non-standard credit assets to reduce risk in shadow banking while still allowing continued operation. While these and other government instituted measures are steps in the right direction, more action is needed and soon.

Can Financial Disaster Be Averted?

China’s biggest challenge is creating a growing, sustainable economy allowing businesses to prosper and individual investors to reap the rewards for their risk. However, unless immediate and necessary changes are made to China’s banking system, it will be difficult for any type of long or short-term economic stability. Banking liberalization policies will be difficult to implement if the central government insists on maintaining authoritarian control and keeping politics and favoritism at the top of its agenda.

New Solar Tech Produces 50% More Energy Than Silicon Cells

Researchers have created a concentrating photovoltaic (CPV) system with embedded microtracking that is capable of producing 50 percent more energy per day than the standard silicon solar cells.

“Solar cells used to be expensive, but now they’re getting really cheap,” says Chris Giebink, an assistant professor of electrical engineering at Penn State.

“As a result, the solar cell is no longer the dominant cost of the energy it produces. The majority of the cost increasingly lies in everything else—the inverter, installation labor, permitting fees, etc.—all the stuff we used to neglect,” he says.

This changing economic landscape has put a premium on high efficiency. In contrast to silicon solar panels, which currently dominate the market at 15 to 20 percent efficiency, concentrating photovoltaics focus sunlight onto smaller, but much more efficient solar cells like those used on satellites, to enable overall efficiencies of 35 to 40 percent.

Current CPV systems are large—the size of billboards—and have to rotate to track the sun during the day. These systems work well in open fields with abundant space and lots of direct sun.

“What we’re trying to do is create a high-efficiency CPV system in the form factor of a traditional silicon solar panel,” says Giebink.

To do this, the researchers embed tiny multi-junction solar cells, roughly half a millimeter square, into a sheet of glass that slides between a pair of plastic lenslet arrays. The whole arrangement is about two centimeters thick and tracking is done by sliding the sheet of solar cells laterally between the lenslet array while the panel remains fixed on the roof.

An entire day’s worth of tracking requires about one centimeter of movement, which is practically imperceptible.

“Our goal in these recent experiments was to demonstrate the technical feasibility of such a system,” says Giebink. “We put together a prototype with a single microcell and a pair of lenses that concentrated sunlight more than 600 times, took it outdoors, and had it automatically track the sun over the course of an entire day.”

Because the team needed to know exactly how much direct and diffuse sunlight there was during the test, they set up at the Russell E. Larson Agricultural Research Center at Penn State where there is a National Oceanic and Atmospheric Administration Surface Radiation monitoring site. Researchers worked to test the system over two sunny days from dawn to dusk right alongside a commercial silicon solar cell.

The CPV system reached 30 percent efficiency, in contrast to the 17 percent efficiency of the silicon cell. All together over the entire day, the CPV system produced 54 percent more energy than the silicon and could have reached 73 percent if microcell heating from the intense sunlight were avoided.

According to Giebink, this embedded tracking CPV technology would be perfect for places with lots of direct sunlight, such as the southwestern United States or Australia.

Giebink notes that major challenges still lie ahead in scaling the system to larger areas and proving that it can operate reliably over the long term, but he remains optimistic.

“With the right engineering, we’re looking at a step-change in efficiency that could be useful in applications ranging from rooftops to electric vehicles—really anywhere it’s important to generate a lot of solar power from a limited area.”

A paper summarizing the research appears in Nature Energy. Additional researchers who contributed to this work are from Penn State, Semprius Inc., and the University of Illinois at Urbana-Champaign.

The Advanced Research Projects Agency-Energy and the National Science Foundation supported this work

Could bond funds break the market?

The Bank of England worries about a sell-off

GOOD generals know that the next war will be fought with different weapons and tactics from the last. Similarly, financial regulators are right to worry that the next crisis may not resemble the credit crunch of 2007-08.

The last crisis arose from the interaction between the market for mortgage-backed securities and the banking system. As investors became unsure of the banks’ exposure to bad debts, they cut back on their lending to the sector, causing a liquidity squeeze. Since then, central banks have insisted that commercial banks improve their capital ratios to ensure they are less vulnerable.

Might the next crisis originate not in the banking system, but in the bond market? That is the subject of a new paper* from the Bank of England. The worry centres on the “liquidity mismatch” between mutual funds, which offer instant redemption to their clients, and the corporate-bond market, where many securities may be hard to trade in a crisis. The danger is that forced selling, to return money to investors, leads to big falls in bond prices, creating a feedback loop.

If that concern seems fanciful, think back to the summer of 2016, when British mutual property funds had to suspend redemptions in the wake of the EU referendum vote. Fund managers simply could not sell properties fast enough to pay off their investors.

The corporate-bond market is a particular concern because it is much less liquid than the equities market. That liquidity has fallen in recent years, because banks have become less willing to act as market-makers. This reluctance is rooted in the regulations imposed after the last crisis, which require banks to hold more capital.

The Bank of England’s study focused on European mutual funds that own investment-grade bonds (the safest category). Since 2005, the worst month for redemptions in this sector occurred in October 2008, when outflows reached the equivalent of 1% of assets under management each week. The sell-off was accompanied by a rise in bond spreads—the gap between the yield on investment-grade bonds and that on government debt—of around a percentage point.

Some of that increase was obviously caused by a deterioration in the economy—investors realised that bond issuers were more likely to default. But the bank reckons that around half the shift was the result of a decline in liquidity. In other words, bond investors demanded a higher yield to compensate them for the difficulty they might face in selling their holdings.

The bank reckons that, if a 1% outflow of mutual-fund assets happened today, then European investment-grade spreads would rise, for liquidity reasons alone, by around four-tenths of a percentage point. That may not sound much, but it is around a third of the average spread since 2000.

What if the sell-off is greater than it was in 2008? After all, near-zero rates on cash must have pushed a lot of investors into corporate-bond funds in recent years. Some of those investors may be using bond funds as “rainy day” money and will thus be reluctant to sit tight if their savings are losing value.

Others could step in to buy the bonds. Long-term holders like pension funds and insurance companies are obvious candidates to do so, although they tend to be slow to react. Hedge funds are more nimble bargain-hunters but they often depend on financing from the banks, and that may not be available in a crisis.

Finally, the banks themselves could step in, but they face capital charges on their market-making activities. The moment could come, the bank suggests, when “dealers reach the limit of their capacity to absorb those asset sales”. This would be the “market-breaking point”. And that stage could be reached when redemptions equal 1.3% of net assets of corporate-bond funds—in other words, only 30% higher than during the 2008 crisis.

A sell-off in corporate bonds ought not to be as damaging as the mortgage-related crisis of 2008. Investors don’t tend to use borrowed money to buy such bonds, and the big asset-management companies don’t back funds with their own capital. Corporate bonds also comprise only a small part of most portfolios. But it could still be traumatic if bond funds need to be suspended. That could undermine retail investors’ confidence in the liquidity of the mutual funds on which many depend for their retirement income. The bank is right to be alert to the risks.

South-East Asia’s future looks prosperous but illiberal

Democracy is losing ground even as the region grows richer

THE young woman with the microphone cajoles, hectors and wheedles customers with the breathless enthusiasm of a livestock auctioneer at a county fair. She is standing behind a table stacked high with blue jeans; most of the milling crowd is dressed in lungyis, Myanmar’s skirt-like national dress. The fancy mall around them is anchored by a huge department store, dotted with banks and mobile-phone stalls and topped by a cinema and video arcade.

Myanmar has been growing so fast—by an average of 7.5% a year for the past five years—that the boom is reverberating in Mae Sot, just across the border in Thailand. Two years ago, says a longtime resident, the site of the mall was a swamp, and Mae Sot was a poky little border town with two small grocery stores. Today huge supermarkets, car dealers, electronics outlets and farm-equipment showrooms line the wide new road from the border into town, patronised by a steady stream of Burmese shoppers. Skeletons of future apartment blocks loom; the Thai government is building a new international airport. The Asian Development Bank (ADB) forecasts that Myanmar’s growth will hit 8% next year.

The region is full of such stories. Cambodia, Laos, the Philippines and Vietnam have been growing only slightly more slowly. Overall, the ten countries of the Association of South-East Asian Nations (ASEAN) grew at an annual rate of 5% over the past five years: not quite as fast as China or India, but much faster than Europe, Japan or America. The region’s 625m-odd people are growing richer and better educated; they will live longer, healthier and more prosperous lives than their parents. Of course, plenty of poverty remains—most people in Myanmar are still subsistence farmers—but the region’s economic trends are promising.

Back from the red

It was not always obvious that the South-East Asian economies would do so well. Only a generation ago Myanmar was cut off from the world by despotic generals; Cambodia’s 25-year-old civil war was still sputtering; and Vietnam was only just beginning to experiment with some timid market reforms. The wealthier countries in the region, meanwhile, had seen their economies, and the underlying models of growth, shattered by the Asian financial crisis of 1997.

The crisis proved salutary. Indonesia, the Philippines and Thailand all adopted sounder macroeconomic policies and made some effort to curb the cronyism that had accompanied earlier growth. Nominally communist Laos and Vietnam and autarkic Myanmar all embraced free markets, up to a point. The days of nationalisation and central planning seem to be over. In much of the region inefficient and coddled state-owned businesses endure, and rent-seeking, corruption and protectionism are all more common than they should be. But across South-East Asia, liberal economics has won the argument.

Politically, however, the region is moving in the opposite direction. The Asian crisis may have brought huge economic hardship, but it did at least unseat Suharto, Indonesia’s strongman of 32 years, and instigate political reforms elsewhere. In the years that followed, imperfect democracies in Malaysia, the Philippines and Thailand appeared to be gaining strength. And Myanmar, after years of isolation and repression, embarked on an unexpected transition to democracy.

Advertisement

But hoped-for openings never came in Laos and Vietnam, where the Communist Party has always been nakedly repressive. Singapore remains an illiberal, albeit effective, technocracy. The leaders of Malaysia and Cambodia, Najib Razak and Hun Sen, have proved depressingly adept at locking up critics and persecuting opponents. Cambodia’s most prominent opposition politician, Sam Rainsy, lives in exile to avoid imprisonment for a spurious conviction for defamation. Opposition figures in Malaysia find themselves in court on charges as varied as corruption and sodomy.

The junta that seized power in Thailand three years ago promises an election next year. Even in the unlikely event that it is free and fair, the constitution—which the army wrote and the new king signed in May—creates a junta-led Senate, imposes the generals’ 20-year plan on the country and provides ample grounds to remove any elected leader whom the army finds lacking. All this is designed to prevent voters from electing the “wrong” leaders, in the army’s view, as they have done at every opportunity over the past 15 years.

Democratic institutions are not yet quite that weak in the region’s two biggest countries, Indonesia and the Philippines, but in both liberals have more cause for fear than hope. Filipino voters, justifiably frustrated by the way that a few prominent families dominate politics, and by how recent economic growth has failed to reduce the high poverty rate, elected Rodrigo Duterte as president last year. Alone among the five candidates, he seemed to care about ordinary people; his brutal anti-drug campaign has appalled foreigners but is popular at home.

Mr Duterte reminisces fondly about the dictatorship of Ferdinand Marcos and seems to crave dictatorial power himself. He has declared martial law on the southern island of Mindanao (see Banyan), and often muses about doing the same nationally. He veers between indifference and hostility to troublesome principles such as due process, the separation of powers and the rule of law—all of which need shoring up, not weakening.

An election for governor of Jakarta in April, meanwhile, has harmed Indonesia’s reputation for religious tolerance (see next story). Islamist agitators campaigned against the Christian incumbent, Basuki Tjahaja Purnama, falsely claiming that he had insulted the Koran. Anies Baswedan, one of his rivals, embraced their shameless attempt to stir up sectarian tension, and won. Prabowo Subianto, a tub-thumping nationalist who lost the presidential election in 2014, backed Mr Baswedan. The fear is that Mr Prabowo, inspired by Mr Baswedan’s success, will try to foster similar divisions at the national level.

But it is Myanmar that most encapsulates the region’s democratic reversal. When the army ceded power last year to Aung San Suu Kyi, its Nobel-prize-winning opponent of 30 years, expectations were astronomically high, even though the constitution the generals had written severely limited her powers. That has made her government’s craven and repressive acts all the more bewildering. It has charged more reporters with defamation than did her military-backed predecessor. She has been shamefully silent about the continuing persecution of the Rohingya, a Muslim minority, not even admitting, let alone trying to stop, the army’s well-documented campaign of rape, murder and destruction against Rohingya villages. It does not help that since Donald Trump became president, America, long the loudest champion of liberal values in the region, has more or less let the subject drop.

Together, technology and teachers can revamp schools

How the science of learning can get the best out of edtech

IN 1953 B.F. Skinner visited his daughter’s maths class. The Harvard psychologist found every pupil learning the same topic in the same way at the same speed. A few days later he built his first “teaching machine”, which let children tackle questions at their own pace. By the mid-1960s similar gizmos were being flogged by door-to-door salesmen. Within a few years, though, enthusiasm for them had fizzled out.

Since then education technology (edtech) has repeated the cycle of hype and flop, even as computers have reshaped almost every other part of life. One reason is the conservatism of teachers and their unions. But another is that the brain-stretching potential of edtech has remained unproven.

Today, however, Skinner’s heirs are forcing the sceptics to think again. Backed by billionaire techies such as Mark Zuckerberg and Bill Gates, schools around the world are using new software to “personalise” learning. This could help hundreds of millions of children stuck in dismal classes—but only if edtech boosters can resist the temptation to revive harmful ideas about how children learn. To succeed, edtech must be at the service of teaching, not the other way around.

Pencils down

The conventional model of schooling emerged in Prussia in the 18th century. Alternatives have so far failed to teach as many children as efficiently. Classrooms, hierarchical year-groups, standardised curriculums and fixed timetables are still the norm for most of the world’s nearly 1.5bn schoolchildren.

Too many do not reach their potential. In poor countries only a quarter of secondary schoolchildren acquire at least a basic knowledge of maths, reading and science. Even in the mostly rich countries of the OECD about 30% of teenagers fail to reach proficiency in at least one of these subjects.

That share has remained almost unchanged over the past 15 years, during which billions have been spent on IT in schools. By 2012 there was one computer for every two pupils in several rich countries. Australia had more computers than pupils. Handled poorly, devices can distract. A Portuguese study from 2010 found that schools with slow broadband and a ban on sites such as YouTube had better results than high-tech ones.

What matters is how edtech is used. One way it can help is through bespoke instruction. Ever since Philip II of Macedon hired Aristotle to prepare his son Alexander for Greatness, rich parents have paid for tutors. Reformers from São Paulo to Stockholm think that edtech can put individual attention within reach of all pupils. American schools are embracing the model most readily. A third of pupils are in a school district that has pledged to introduce “personalised, digital learning”. The methods of groups like Summit Public Schools, whose software was written for nothing by Facebook engineers, are being copied by hundreds of schools.

Advertisement

In India, where about half of children leave primary school unable to read a simple text, the curriculum goes over many pupils’ heads. “Adaptive” software such as Mindspark can work out what a child knows and pose questions accordingly. A recent paper found that Indian children using Mindspark after school made some of the largest gains in maths and reading of any education study in poor countries.

The other way edtech can aid learning is by making schools more productive. In California schools are using software to overhaul the conventional model. Instead of textbooks, pupils have “playlists”, which they use to access online lessons and take tests. The software assesses children’s progress, lightening teachers’ marking load and giving them insight on their pupils. Saved teachers’ time is allocated to other tasks, such as fostering pupils’ social skills or one-on-one tuition. A study in 2015 suggested that children in early adopters of this model score better in tests than their peers at other schools.

Pay attention at the back

Such innovation is welcome. But making the best of edtech means getting several things right. First, “personalised learning” must follow the evidence on how children learn. It must not be an excuse to revive pseudoscientific ideas such as “learning styles”: the theory that each child has a particular way of taking in information. Such nonsense leads to schemes like Brain Gym, an “educational kinesiology” programme once backed by the British government, which claimed that some pupils should stretch, bend and emit an “energy yawn” while doing their sums.

A less consequential falsehood is that technology means children do not need to learn facts or learn from a teacher—instead they can just use Google. Some educationalists go further, arguing that facts get in the way of skills such as creativity and critical thinking. The opposite is true. A memory crammed with knowledge enables these talents. William Shakespeare was drilled in Latin phrases and grammatical rules and yet he penned a few decent plays. In 2015 a vast study of 1,200 education meta-analyses found that, of the 20 most effective ways of boosting learning, nearly all relied on the craft of a teacher.

The second imperative is to make sure that edtech narrows, rather than widens, inequalities in education. Here there are grounds for optimism. Some of the pioneering schools are private ones in Silicon Valley. But many more are run by charter-school groups teaching mostly poor pupils, such as Rocketship and Achievement First—or Summit, where 99% of graduating pupils go on to university and laggards make the most progress relative to their peers in normal classes. A similar pattern can be observed outside America. In studies of edtech in India by J-PAL, a research group, the biggest beneficiaries are children using software to receive remedial education.

Third, the potential for edtech will be realised only if teachers embrace it. They are right to ask for evidence that products work. But scepticism should not turn into Luddism. A good model is São Paulo, where teachers have welcomed Geekie, an adaptive-software company, into public schools.

In 1984 Skinner called opposition to technology the “shame” of education. Given what edtech promises today, closed-mindedness has no place in the classroom.

Where America Gets Its Electricity

California warms to solar. Texas leans into wind. And the East Coast is switching to natural gas.

The way the U.S. generates electricity has changed a lot over the past decade.

The Big Switch

Net electricity generation by source, trailing 12 months

This chart ran with the electrical utility industry's seeming conviction that its reliance on natural gas and renewables is just going to keep growing, regardless of the Donald Trump administration's efforts to engineer a coal resurgence.

The chart only included the top five sources of electrical power in the U.S. 

1 Perhaps surprisingly, solar power doesn't make the cut. Even when you include the Energy Information Administration's estimate of the power generated by rooftop panels and other small-scale solar, it's still in seventh place behind biomass (burning wood, mainly).

Solar is gaining fast, though, and in a few places, it has already arrived. In sunny, renewables-obsessed California, for example, solar recently nosed out nuclear for third place among large-scale electricity sources:

Solar and Wind on the Rise in California

Net electricity generation by source, trailing 12 months

With the estimated contribution of small-scale solar added in, the sun is now California's No. 2 source of electricity. Coal has not been a significant part of the state's electricity-generating mix for a long, long time, although that's somewhat misleading given that California imports coal-generated power from its neighbors. The state is heavily reliant on natural gas generation, but its natural gas use has been declining lately thanks in part to the growth in solar and wind and in part to a very wet winter that's been great for hydropower generation. Also, geothermal energy has long been a major source of electricity in California, and was only passed by solar and wind in 2015.

California's energy mix, and the changes in it in recent years, is very different from those of the nation as a whole. Now, we all know California can be weird. But it turns out that most states' electricity-generation charts don't look much like the national one. Consider Texas, the second-most-populous state after California (and, thanks in part to its ridiculously hot summers, a much bigger electricity user than California):

Wind Gains in Texas

Net electricity generation by source, trailing 12 months

Texas has seen a bit of the natural gas gains and coal declines that the rest of the country has -- with a marked if probably temporary reversal in recent months driven by higher natural gas prices. But the most significant change in its electricity picture has been the rise of wind power. This hasn't resulted in big drops in generation from other sources, though, because the fast-growing state has seen overall electrical generation increase by more than 20 percent since 2007.

Then there's Pennsylvania, a slow-growing longtime coal state that quite suddenly became the nation's second-biggest producer of natural gas:

Pennsylvania Trades Coal for Gas

Net electricity generation by source, trailing 12 months

Similar coal-to-gas switches can be seen in generation data from states up and down the Eastern Seaboard. In much of the Midwest, though, the changes have been more muted:

Slow Change in Wisconsin

Net electricity generation by source, trailing 12 months

What's apparent here is that the U.S. doesn't have a national market for electricity. It has lots of different regional and state markets, with energy mixes determined by the availability of fuels and other generation sources, and by the priorities of state and local politicians. Yes, national energy and environmental policies matter. But they're far from the only things that matter.