Price rise: The original Premise on data
Latest News »
New Delhi/Mumbai: In March of 2013, onion prices in India began to climb steadily; by August, they had reached Rs.80 a kg in some cities, making regular headlines and prompting editorials referring to “the great onion crisis of 2010”. By late September, when the wholesale price index for onions touched an all-time high of 845.6, the new governor of the Reserve Bank of India (RBI), Raghuram Rajan, raised policy rates. Rajan later told reporters that food inflation was “worryingly high”.
In December, the issue turned political. When the results of state assembly elections were declared, the political cost of inflation finally sank in: voters had turned out in record numbers against the ruling Congress party at the centre, which they held responsible for untrammelled inflation. The Congress admitted that price rises had created widespread resentment, and identified reforms in vegetable wholesale markets in Congress-ruled states as a key priority ahead of national elections due in April-May, as did the Bharatiya Janata Party (BJP) government in Delhi back in 1998.
Unlike during previous price spikes, however, this time around someone from outside the government had been watching. Premise Data Corp., a fledgling data collection and analysis company based out of San Francisco, was tracking the numbers on the ground and comparing them to the official inflation index.
Premise, which officially launched in October, had funding from Google Ventures, Andreessen Horowitz and Harrison Metal, as well as the brains of both a computer scientist from Google, Joe Reisinger, and a quantitative analyst, David Soloff. Using data collectors armed with smartphones all over the country, capturing real-time instances of price rises in local markets, Premise had been observing the upward trend since February, and early on had spotted that something was awry.
“Our data systems picked up the price spikes in onions early in India: in May,” said Soloff, speaking from San Francisco. “We thought there was something wrong with the persistent price increases but official data a few months down the line confirmed those trends and we had a policy reaction (from RBI) as late as September.”
The lagging response from policy makers was something that Soloff had seen before; in fact, it had given him the idea for the company in the first place. Back in 2009, in the wake of the US financial crisis, Soloff remembered what he describes as a “meme, or a set of stock phrases that kept appearing in the press”. The message, Soloff said, was roughly: “What the hell is going on?”
“For me the moment that I started scratching my head was in late ’09, when people were talking about broad deflationary trends that seemed to persist, and yet I saw how my grocery bill was skyrocketing,” Soloff said. “I wondered what would happen when you replicated my experience across towns and cities and countries.”
The economy seemed to be changing faster than existing tools could measure. Part of the problem, Soloff said, was that the official measures of how the US economy was doing—the jobs and inflation reports—came out too late to be of much use to policymakers. “It seemed there was an increasing divide between how quickly things moved and an older, macro index that we were relying on. It’s not surprising there’s so much conjecture, it’s because the indicators themselves are artefacts of a former age.”
Soloff wasn’t the only one who felt the need for more current data. Hal Varian, Google’s chief economist, who had taken a role on Premise’s advisory board after Google Ventures invested in the company, has been making the same claim for some time.
“The buzz word these days is ‘now-casting’.” Varian said. “You need a picture of what’s happening to the economy now. If we think back to the recession of 2008, when the Obama economic team came in and was trying to develop a response to this big shock to the financial system, they were working with data that was six weeks old. So they really underestimated, not because they did anything wrong, but because they didn’t have the data,” Varian said. In the developing world, the data limitations were typically even more severe, constraining decision-making by households, governments, corporations, and financial institutions. If only, Soloff had thought in 2009, there was an alternative tracking mechanism that could provide data on price rises within days instead of weeks, and help people to understand why and how prices were fluctuating. For such a mechanism to be successful, two conditions would be necessary: the data had to be speedily acquired and processed, and it had to be verifiable. The marriage of cloud computing with smartphone technology provided Premise with just the right tools to devise a liquid inflation gauge, said Soloff. Premise’s data collectors, typically students, take photos of produce in stores and upload them onto a specially designed app. Premise claims to have a presence in 30 countries. In India, a country with a notoriously volatile food inflation rate, the system seems to be working.
“What you have now is the emergence of a real on-the-ground story. We are in our seventh or eighth month in India and we are able to spot trends now,” Soloff said. Premise is focusing on foods within the consumer price index and is building up an alternative gauge that uses the same weights as the official index but provides more liquid and detailed data.
The final goal is to develop alternatives to official gauges of economic activity that will provide real-time data to help consumers and investors judge the situation on the ground with far greater precision than they have done in the past. At the very least, Soloff says, Premise’s data manages to capture volatility—and often that is enough for people who may be agnostic about where an index is headed but are keenly interested in the fluctuations in its levels.
“Ultimately, the final product we would like to build is a transparent and liquid index of inflation and similar indices for specific consumer items. So, a producer who relies upon a particular item as his input can then hedge against the price rise effectively,” said Soloff.
Collecting the data
Initially, at least, Premise has to be specific about the kind of data on which it will focus. “We know we want a number of products, documented over a certain number of cities, a minimum number of data points and neighbourhood coverage maps,” said Soloff. “We farm out the location and documentation of these products in the wild via our mobile network. Every day up and down that chain we adjust. We ask (our data collectors) to use their commutes to collect information and everyday on their cellphones there is a dynamically adjusting list of specs. It might be the price of one kilo of a certain cut of lamb, quite specific, they are paid on a per data point basis.”
Premise’s team makes sure that the data collection points are well dispersed geographically and according to store type, Soloff said. The vast majority of the collectors work for less than two hours per day, capturing 30-40 price points in neighbourhood markets. The collectors are themselves interested in knowing how price trends are moving in their locality and can compare localities across their cities using Premise’s data maps, said Soloff. There are benefits for the shopkeepers too. “We are now providing shopkeepers with a report of where they stack up in their area”.
There are inherent difficulties with data provided by a workforce of non-professional data collectors, said Varian, but these seem to be overcome by the method by which they capture the prices. “If you want to have broader access at a lower cost, you are going to have a team of people who are not professionals, so that validation component has to make the model work,” he said. In Premise’s case, the validation is done by a computer, geo-stamping and time-stamping the images that flood in, making sure that each comes from the correct location and time. Without the smartphone, Varian said, such precision would not be possible.
“Within a day or two days, you have that information that you requested gathered by this crowd-sourced method and the computer has validated this data. In developed countries you can have a relationship with Walmart or Target and find these things out, but in countries that are much smaller there’s no supply-chain connection. You can send people to the store and then Proctor and Gamble back in Cincinnati knows how their product is being displayed in a little store in Vietnam and you couldn’t really do that before. Now it’s easy, the critical component is that device, the mobile phone.”
Premise’s ambitions are big and Soloff says the applications for its model are broader than simply tracking inflation. While the company’s major client is currently Bloomberg, the financial data service, it also serves a number of large financial, government and non-government institutions as well as consumer packaged goods companies. Premise is already working on alternative gauges to track proxies for consumption, productivity and demand.
“In the US our government spends over $300 million a year tracking two numbers—the jobs number and the inflation number,” he said. “Why wouldn’t we be a wholesale provider to them? Can we deliver 30% of that data responsibly? Is there a way to make inflation an investable asset? You need a transparent index that you can benchmark against.”
“I don’t see many investors opting for the kind of price data they are offering at this stage because it doesn’t offer you much help in predicting stock prices. Investors who make a broad call on an economy while investing will scarcely bother to know about granular price information, and may look only at the headline inflation numbers.”
It may interest more investors if information vendors (such as Premise) are able to provide accurate day-by-day data on real variables in the economy such as retail sales figures, said Mukherjea.
Investors in bond, currency and commodity markets may be more interested than stock market participants in the kind of price information that Premise provides. “For many commodity and forex traders, comparing prices of consumer staples across geographies can be very helpful in taking decisions,” said Soloff. “The bond market is developing in several emerging markets such as India and Brazil, and investors in these markets would want to have a reliable predictor of inflation.”
The data Premise offers could be of help to central banks, according to Ajay Shah, economist and professor at the National Institute of Public Finance and Policy (NIPFP).
“More than a bond trader or a currency trader, it is monetary policymakers who stand to gain the most from more detailed and faster price information,” said Shah. “Our central bank will increasingly become an inflation-targeting central bank, and hence any and every technique that helps the central bank gauge and forecast prices better is welcome.”
Varian pointed to one example of Premise targeting areas where government figures have been questionable. “The Economist stopped reporting the official price inflation numbers from Argentina because they said they were fiction and people from Premise and another group called The Billion Prices Project looked at online data and got a much, much clearer picture of what was going on than one could get from the official statistics, so, now, attempting to misrepresent or hide or contain that data is futile.”
Soloff is interested in systems. Previous to starting Premise with Reisinger he’d worked in tech companies in Northern California for more than a decade. Before that he was a quantitative analyst for banks building trading models. But he had a humanities background, having studied linguistics at Columbia University and history and economics at UCB (University of California, Berkeley). He then worked as a Latin and Greek bibliographer for a rare books dealer in Minneapolis, Minnesota. Throughout, he says, his liking for systems has driven his career choices.
“I like systems theory, I like machinery, the economy is just a big machine with all these tightly fitting parts,” he said. “Every linguistic system also has its own rules, I did a lot of Latin, Greek and Arabic and Aramaic, I think linguistics training helps people understand that when you alter the system the end results change.”
He is convinced that the applications for the data Premise will provide is broad-based, and will appeal not just to a broad class of investors but also to ordinary consumers, who may want to match their daily lived experience with what the government tells them about it in its official reports. The newest offering from Premise is a consumer app that will allow consumers to compare prices across stores and neighbourhoods in their cities; almost a price-centric version of Google Street View. As more and more people sign on to it, they can be persuaded to also upload price data for Premise, Soloff hopes, leading to a crowd-sourced, self-sufficient data network.
“The typical (data) structure is often one where demands for information live outside the supply line,” said Soloff. “Some trader in London might want to know about prices in Mumbai, even if people on the ground have no idea what happened. I don’t think that’s right. What happens when the people who contribute the information are the primary beneficiaries of it?”
“If I have the makings of an econometric map on my mobile I have a real incentive to contribute because I and my family and peers are directly benefiting from it. The network then becomes self-sufficient.”