Latest / Trumponomics / Big Data Goes Where Economies Fear to Tread

Transcript

Big Data Goes Where Economies Fear to Tread · Trumponomics

• 0:00
  Speaker 1: Big data. In much of the developed world, it can
• 0:04
  Speaker 1: refer to the analysis of vast quantities of information to
• 0:07
  Speaker 1: search for valuable trends. Yet in developing nations, where actual
• 0:12
  Speaker 1: economic data happens to be pretty scarce, using big data
• 0:17
  Speaker 1: is the latest way to truly understand what is happening
• 0:21
  Speaker 1: in a country's economy. And this just happens to be
• 0:24
  Speaker 1: pretty done interesting, from satellites the track oil storage facilities
• 0:29
  Speaker 1: in China to using cell phone data to monitor poverty
• 0:33
  Speaker 1: across Africa, where simply going where government figures don't welcome
• 0:46
  Speaker 1: to benchmark. I'm Scott Landman, an economics editor with Bloomberg
• 0:50
  Speaker 1: in Washington, and I'm Daniel Moss, global economics writer at
• 0:54
  Speaker 1: Bloomberg View in New York. So, Dan, we take it
• 0:58
  Speaker 1: for granted the existence of reliable economic data in our jobs.
• 1:02
  Speaker 1: I mean, I edit our u S economy coverage, we
• 1:05
  Speaker 1: get bountiful data unemployment, GDP, inflation. But that's not the
• 1:11
  Speaker 1: case everywhere, right, Corporations, governments, lender's investors, everyone's making decisions
• 1:20
  Speaker 1: based on assumptions about the economic trajectory. But what if
• 1:25
  Speaker 1: those assumptions are built on data that isn't reliable. Well,
• 1:29
  Speaker 1: that's what we're going to talk about today. Before we
• 1:31
  Speaker 1: bring in Joshua Blumenstock, a professor who does research based
• 1:35
  Speaker 1: on big data, I wanted to talk first with a
• 1:37
  Speaker 1: Bloomberg colleague who has taken a keen interest in this topic.
• 1:42
  Speaker 1: Jeff Karns is our China Economy editor, and he joins
• 1:46
  Speaker 1: us from Beijing. Jeff, it's pretty late night where you are,
• 1:50
  Speaker 1: so I'll say, Wan Chong, how hid? And I'll say, Jeff,
• 1:55
  Speaker 1: can you tell us a little bit about how people
• 1:57
  Speaker 1: are using unusual sources of to understand what's going on
• 2:01
  Speaker 1: in China's economy. Since you've spent a lot of time
• 2:04
  Speaker 1: on this topic, well, there's a lot going on. I
• 2:07
  Speaker 1: think that it's Uh, it's a really unique case because
• 2:10
  Speaker 1: not only is this the you know, the second largest
• 2:13
  Speaker 1: economy in the world, but it's also it's growing very
• 2:16
  Speaker 1: quickly and on pace to become the largest. But it's
• 2:19
  Speaker 1: so much less well understood than a lot of other
• 2:23
  Speaker 1: countries out there, and the depth and breadth of the
• 2:26
  Speaker 1: data are really not what folks like yourself, uh, and
• 2:30
  Speaker 1: a lot of institutional investors and economists might be used to,
• 2:34
  Speaker 1: and that makes it really a great proving ground for
• 2:39
  Speaker 1: new technologies to look at everything from consumers spending to
• 2:44
  Speaker 1: crude oil storage, to kind of get past the government
• 2:47
  Speaker 1: data which may be spotty or not trusted, and use
• 2:52
  Speaker 1: new and independent ways of looking at these things to
• 2:57
  Speaker 1: get past the curtain of that. And so that's something
• 2:59
  Speaker 1: that we've written a few things about and hope to
• 3:03
  Speaker 1: do more. We're really interested in the speed of things
• 3:07
  Speaker 1: that are being put out in real time, from consumer
• 3:11
  Speaker 1: credit card swipes to spending on mobile apps, and also
• 3:17
  Speaker 1: looking at the bigger picture of the commodities market for oil,
• 3:22
  Speaker 1: which is China's the biggest consumer of oil, and we're
• 3:26
  Speaker 1: really excited about all the new things that are coming out. So, Jeff,
• 3:29
  Speaker 1: if I understand what you're saying correctly, China's economy is
• 3:33
  Speaker 1: so big yet so little understood. It's almost like was
• 3:37
  Speaker 1: the fertile ground just waiting for this big data test
• 3:41
  Speaker 1: case and this is it. Well, it is a test case.
• 3:46
  Speaker 1: Is probably a good point. You reached an issue that
• 3:48
  Speaker 1: for economists, they like to have the long series going
• 3:52
  Speaker 1: back decades to look at very different economic eras across time.
• 3:56
  Speaker 1: But when you are going to look at these kind
• 3:58
  Speaker 1: of new things, whether they're it's data coming from satellites
• 4:02
  Speaker 1: or from Internet traffic or something like that. You're not
• 4:05
  Speaker 1: gonna be able to go back very far, so there
• 4:07
  Speaker 1: is a definite trade off. You'll get the speed and
• 4:09
  Speaker 1: immediacy of being able to do something very quickly, but
• 4:13
  Speaker 1: you may only be able to go back five years.
• 4:15
  Speaker 1: So there's a real trade off for what economists like
• 4:18
  Speaker 1: to have in terms of both the length and the
• 4:22
  Speaker 1: quality of the data. This isn't just in China, Jeff.
• 4:26
  Speaker 1: This is also something's happening in the rest of the world. Right,
• 4:30
  Speaker 1: that's right. The advanced and wealthier economies have generally pretty
• 4:34
  Speaker 1: good government data that's that's trusted and can complete and
• 4:38
  Speaker 1: measures things like wages and inflation and all of these
• 4:43
  Speaker 1: different things that have been independently measured and checked by
• 4:47
  Speaker 1: government specialists for form decades or more. And when you
• 4:52
  Speaker 1: when you get beyond that and you go to places
• 4:55
  Speaker 1: like Africa and see that there's maybe only in the
• 4:59
  Speaker 1: whole cluster of African nations, even only a handful of
• 5:02
  Speaker 1: countries that have this data, it maybe is more appealing
• 5:06
  Speaker 1: to turn to an alternative like satellites that can track
• 5:10
  Speaker 1: how much light the countries are throwing off at night
• 5:15
  Speaker 1: to kind of serve as a proxy for small countries
• 5:19
  Speaker 1: that may be producing very little economic data or not
• 5:21
  Speaker 1: at all. You know, Jeff, one of the things that's
• 5:23
  Speaker 1: really cool when I hear you talk about this is
• 5:27
  Speaker 1: too often people consider the study of economics to be
• 5:31
  Speaker 1: a marriage to dry statistics. Yet you're talking about satellites
• 5:35
  Speaker 1: flying over car parks, you're talking about credit cards, swipes,
• 5:41
  Speaker 1: you're talking about pretty hip things. Well, these are very hip,
• 5:46
  Speaker 1: they're also they're also very new. One of the things
• 5:48
  Speaker 1: that's interesting about China specifically is that there's a high
• 5:52
  Speaker 1: degree of centralization that probably people and most of the
• 5:55
  Speaker 1: rest of the world wouldn't really recognize. For bank cards,
• 5:59
  Speaker 1: there's really one company called Union Pay that handles everything,
• 6:02
  Speaker 1: and beyond that, there's two platforms, we Chat and Ali
• 6:06
  Speaker 1: Baba's Ali Pay, where people use their phones to pay,
• 6:09
  Speaker 1: and that's about it. You can kind of live on
• 6:11
  Speaker 1: those things forever. And these things are very centralized, and
• 6:16
  Speaker 1: so when you have all of the bank card information
• 6:20
  Speaker 1: going through one provider, you can drill into all kinds
• 6:24
  Speaker 1: of different parts of the economy like karaoke bars and
• 6:30
  Speaker 1: hot pot restaurants and see these things at one stop shop.
• 6:35
  Speaker 1: Last question, Jeff before we let you go to bed,
• 6:38
  Speaker 1: And we have to address this. Every time there's a
• 6:41
  Speaker 1: discussion about Chinese numbers, whether they're perceived as good numbers
• 6:46
  Speaker 1: or bad numbers, there's this cry, oh, they're fake. Everyone
• 6:51
  Speaker 1: knows they're made up. The government said, so, where does
• 6:54
  Speaker 1: all this come from? My understanding, as it comes from
• 6:58
  Speaker 1: a leaked cable based on a discussion between Prime Minister
• 7:03
  Speaker 1: Leak Cheng, then a provincial official and a US diplomat
• 7:07
  Speaker 1: ten years ago, where does this idea that the data
• 7:10
  Speaker 1: is fake come from? Well, that's one. That's one place,
• 7:15
  Speaker 1: and that that is something that that did gain some notoriety.
• 7:19
  Speaker 1: But it's interesting because what was being discussed in that
• 7:21
  Speaker 1: was things that we would think of as being very old,
• 7:24
  Speaker 1: like rail car volumes and electricity consumption. It's a really
• 7:29
  Speaker 1: different economy now and these new measures are kind of
• 7:34
  Speaker 1: providing a way to add a different layer of truth
• 7:37
  Speaker 1: to that. And also the question of the vacity of
• 7:42
  Speaker 1: the data hasn't gone away. The government said just this
• 7:45
  Speaker 1: year the two provinces had faked different economic indicators. They
• 7:50
  Speaker 1: didn't say exactly what or how or when or who,
• 7:52
  Speaker 1: but this is something that the government revealed on its own,
• 7:57
  Speaker 1: and so no less of authorities than President She and
• 8:00
  Speaker 1: King has said earlier this year that officials must be
• 8:03
  Speaker 1: frank and forth right with their reports. So when we
• 8:07
  Speaker 1: look at the big data, though it really lines up
• 8:09
  Speaker 1: a lot with the official numbers, we don't see a
• 8:11
  Speaker 1: big difference. We don't see that the government numbers are phony,
• 8:15
  Speaker 1: and it's really the economy is growing a lot less quickly.
• 8:19
  Speaker 1: It's a supplemental thing instead of being a totally different
• 8:22
  Speaker 1: story from fake data to virtually no data. Let's turn
• 8:26
  Speaker 1: to a true expert on this topic now. Joshua Blumenstock
• 8:31
  Speaker 1: is an assistant professor at the University of California Berkeley
• 8:35
  Speaker 1: School of Information and the director of the Data Intensive
• 8:38
  Speaker 1: Development Lab. He focuses on using novel data and methods
• 8:42
  Speaker 1: to better understand the causes and consequences of global poverty.
• 8:47
  Speaker 1: Josh thanks for joining us today. Thanks for having me. Josh,
• 8:50
  Speaker 1: Can you first just tell us a little bit about
• 8:52
  Speaker 1: yourself and how you came to this research specialty. So,
• 8:56
  Speaker 1: I was a liberal arts major in college, and I
• 8:59
  Speaker 1: guess I was fortunate to have spent a lot of
• 9:01
  Speaker 1: time abroad and growing up. I was always interested in understanding,
• 9:05
  Speaker 1: you know, why people were poor and thinking about ways
• 9:09
  Speaker 1: that might help people in difficult circumstances. But at the
• 9:13
  Speaker 1: same time, I always knew that my comparative advantage was
• 9:16
  Speaker 1: working with numbers and data and doing sort of computation.
• 9:20
  Speaker 1: So to be honest, I I sort of floundered around
• 9:22
  Speaker 1: a little bit because for a long time there wasn't
• 9:25
  Speaker 1: a real way to piece together data and quantitative modeling
• 9:28
  Speaker 1: with the problems that were really salient and developing countries.
• 9:32
  Speaker 1: But starting around ten or fifteen years ago, there tended
• 9:36
  Speaker 1: to be more and more sources of really large scale
• 9:39
  Speaker 1: big data coming from even the poorest countries in Sub
• 9:42
  Speaker 1: Saharan Africa, coming from mobile phone networks, coming from satellite
• 9:46
  Speaker 1: imagery and so forth, and so that really created an
• 9:49
  Speaker 1: opportunity for sort of quantitative wonks like me to think
• 9:53
  Speaker 1: about how those data could be used in development economics research.
• 9:57
  Speaker 1: And that's what I do. Now. The two big has
• 10:00
  Speaker 1: taken on rather sinister connotations in the US. What you're
• 10:05
  Speaker 1: saying is that, in terms of studying poverty and figuring
• 10:09
  Speaker 1: out the best way to alleviate that poverty, Big diighter
• 10:12
  Speaker 1: is a godsend. Yeah. I mean, you guys really hit
• 10:15
  Speaker 1: it on the head and the motivation. Just to pull
• 10:18
  Speaker 1: a couple examples. In Angola, the last census prior to
• 10:23
  Speaker 1: two thousand fourteen was in nineteen seventy and in between
• 10:26
  Speaker 1: those two census is the population of the country grew
• 10:29
  Speaker 1: from roughly six million to roughly twenty five million. And
• 10:33
  Speaker 1: so you can imagine that if a policymaker or a
• 10:36
  Speaker 1: researcher is relying on those data to understand the complexion
• 10:40
  Speaker 1: of the country, um, those numbers are just totally off.
• 10:43
  Speaker 1: I mean, when the jobs numbers in the US jump
• 10:47
  Speaker 1: by a quarter of a point, the it's sort of
• 10:49
  Speaker 1: the shock waves reverberate throughout the economy. This is this
• 10:52
  Speaker 1: is like a three d fifty percent change um that
• 10:55
  Speaker 1: is just not reflected in the official statistics. And so
• 10:58
  Speaker 1: I think the idea is, as Jeff was sort of
• 11:01
  Speaker 1: hitting on this too, not that big data would replace
• 11:05
  Speaker 1: these national official statistics, but when you don't have other information,
• 11:10
  Speaker 1: this can provide a supplement to fill in some of
• 11:12
  Speaker 1: the gaps. Can you tell us a little bit more
• 11:14
  Speaker 1: about how you used cell phone data in Africa to
• 11:19
  Speaker 1: analyze poverty or population or whatever it might be. I mean,
• 11:23
  Speaker 1: I just I find it really fascinating. With traditional economic data,
• 11:27
  Speaker 1: there's a lot of government surveying involved, but you know,
• 11:30
  Speaker 1: here you I guess you have vast troves of private
• 11:33
  Speaker 1: data that are are helping you find novel ways to
• 11:36
  Speaker 1: do research. Yeah, and and Jeff alluded to this as well.
• 11:40
  Speaker 1: This is not sort of totally new. People have been
• 11:42
  Speaker 1: using satellite data, in particular night light data. So this
• 11:46
  Speaker 1: is imagery taken by satellites when it's dark to measure
• 11:49
  Speaker 1: economic activity. And for a long time people have known that,
• 11:52
  Speaker 1: you know, cities and places that shine brightly tend to
• 11:55
  Speaker 1: be wealthier. The problem is those methods don't actually really
• 11:58
  Speaker 1: work that well in Africa because there are vast areas
• 12:02
  Speaker 1: of the continent that don't have much light that's detectable
• 12:05
  Speaker 1: from outer space at night. That doesn't mean they're abjectly poor,
• 12:09
  Speaker 1: It just means that they're not They don't have street
• 12:11
  Speaker 1: lamps and the sort of things that satellites pick up.
• 12:13
  Speaker 1: And so over the last several years, myself and other
• 12:17
  Speaker 1: researchers have been looking to all for alternatives to night
• 12:20
  Speaker 1: lights data. And so some of the work I've done
• 12:22
  Speaker 1: shows that looking at the activity of mobile phone calls
• 12:26
  Speaker 1: within a country, and these are fairly precisely located geographically
• 12:30
  Speaker 1: because of the physical layout of the mobile phone network.
• 12:34
  Speaker 1: That that sort of information contains signs that allow you
• 12:37
  Speaker 1: to measure with fairly high accuracy subregional poverty and wealth.
• 12:43
  Speaker 1: More recently people have used daytime satellite imagery, the sort
• 12:46
  Speaker 1: of images you see when you look on Google Maps
• 12:49
  Speaker 1: and you see pictures of your house that you know,
• 12:51
  Speaker 1: and when you do that in Africa and now the
• 12:53
  Speaker 1: whole world is is imaged fairly frequently. Um, there are
• 12:58
  Speaker 1: also correlates of poverty, and you can take fancy machine
• 13:01
  Speaker 1: learning algorithms and apply them to the raw data and
• 13:04
  Speaker 1: spit out on the other end something like a best
• 13:07
  Speaker 1: guess of whether a village is below the poverty line
• 13:10
  Speaker 1: or above the poverty line. So that's the poverty identification.
• 13:14
  Speaker 1: What about the next step, the alleviation. Does Big Day
• 13:17
  Speaker 1: to help with that as well? That's a great question,
• 13:21
  Speaker 1: and I think, to be honest, the answer is not yet,
• 13:24
  Speaker 1: at least not really. I think right now this is
• 13:27
  Speaker 1: really the test balloon phase of this sort of work.
• 13:30
  Speaker 1: I mean, these data have really been around for only
• 13:33
  Speaker 1: the last five years, and right now the research community
• 13:36
  Speaker 1: is working with them very actively, trying to figure out
• 13:39
  Speaker 1: basic things like what can you measure accurately and doing
• 13:43
  Speaker 1: the sort of things that the research community should do,
• 13:45
  Speaker 1: like validate those measurements to make sure they're actually accurate.
• 13:49
  Speaker 1: And I think, to be honest, it will take a
• 13:50
  Speaker 1: few years before these tools migrate into the policy community.
• 13:54
  Speaker 1: I think already you see collaborations forming between people like
• 13:59
  Speaker 1: myself and local governments. And I work in places like
• 14:03
  Speaker 1: Afghanistan and Malawi and Rwanda and have had many conversations
• 14:07
  Speaker 1: and discussions and are fleshing out sort of joint cooperative
• 14:11
  Speaker 1: projects with these governments. But not yet have you seen
• 14:15
  Speaker 1: active use of these methods to like totally change how
• 14:18
  Speaker 1: government policies being made well. That brings us to another question,
• 14:22
  Speaker 1: which is, you know, I spend time in China. Chinese
• 14:25
  Speaker 1: government likes to keep a very tight lid on the
• 14:28
  Speaker 1: kind of statistics that are out there. There's even been
• 14:31
  Speaker 1: some private numbers that have kind of disappeared over the
• 14:34
  Speaker 1: years when they maybe cut a little too close to
• 14:36
  Speaker 1: the bone. You know, in Africa, some of these governments
• 14:39
  Speaker 1: aren't known for being well functioning democracies for that matter. Josh,
• 14:44
  Speaker 1: you are these governments welcoming of studies like the ones
• 14:49
  Speaker 1: that you're doing. I think the answer is always going
• 14:53
  Speaker 1: to be it depends by and large, the people that
• 14:56
  Speaker 1: I've encountered are actually very receptive to these sort of
• 15:00
  Speaker 1: methods and measurements, at least the people that matter are.
• 15:03
  Speaker 1: For instance, we're working with the government of Afghanistan and
• 15:06
  Speaker 1: they believe that they've never really had an accurate account
• 15:09
  Speaker 1: of the number of people in the country, and for
• 15:12
  Speaker 1: the top level bureaucrats, this is really something that that
• 15:15
  Speaker 1: could be valuable for them, I think at a at
• 15:17
  Speaker 1: a higher level, Um, and you guys were touching on
• 15:20
  Speaker 1: this earlier. These data also provide some sort of force
• 15:24
  Speaker 1: transparency in a sense. Um. So, speaking of China, you know,
• 15:27
  Speaker 1: there's studies that use satellite imagery to measure air pollution,
• 15:31
  Speaker 1: and what you can see is that the satellite based
• 15:34
  Speaker 1: measures of air pollution, which by most accounts should be
• 15:37
  Speaker 1: fairly objective, systematically diverge from what have been reported in
• 15:41
  Speaker 1: the local provinces. And so I guess the thought is
• 15:44
  Speaker 1: that when people know that there are these more objective
• 15:47
  Speaker 1: measurements out there, it creates a little bit more of
• 15:51
  Speaker 1: a strategic tension when people want to systematically fudge those numbers. Jeff,
• 15:55
  Speaker 1: how do you see it from China? Does that make
• 15:57
  Speaker 1: sense from from your perspective of I mean, what have
• 16:00
  Speaker 1: you seen happen just in the last couple of years
• 16:02
  Speaker 1: with these kinds of private statistics, well as private statistics
• 16:06
  Speaker 1: there you're your observation about them being subject to higher
• 16:11
  Speaker 1: forces is one that's noted. There have been some new
• 16:15
  Speaker 1: and exciting numbers that have come out, and not all
• 16:18
  Speaker 1: of them are fill around. And as Dan mentioned earlier,
• 16:22
  Speaker 1: there's also been questions about the government. But the thing
• 16:26
  Speaker 1: about China is that it is so vast and there's
• 16:29
  Speaker 1: so much that this stuff is going to come together
• 16:32
  Speaker 1: in some kind of colorescent form. The three big technology
• 16:37
  Speaker 1: companies Buy Do, Ali, Baba, and ten Cents, the last
• 16:40
  Speaker 1: of which operates we Chat, the ubiquitous messaging and financial
• 16:44
  Speaker 1: platform increasingly are spitting out more and more ways to
• 16:50
  Speaker 1: look at the consumer and mobility and all kinds of
• 16:54
  Speaker 1: things that no one's ever been able to look at before,
• 16:57
  Speaker 1: and so the top down control of and kind of
• 17:00
  Speaker 1: cuts both ways. And these companies, with their total market
• 17:05
  Speaker 1: share of of tracking pretty much everybody, uh cast a
• 17:11
  Speaker 1: very wide net. And when you have a platform like
• 17:14
  Speaker 1: ali Baba that is so deeply enmeshed in a consumer
• 17:20
  Speaker 1: e commerce world like this that's bigger than anything that's
• 17:25
  Speaker 1: ever existed, you contract things very closely and start seeing
• 17:29
  Speaker 1: things over time that you never even would have expected.
• 17:33
  Speaker 1: And then you layer on top of that other things
• 17:36
  Speaker 1: like location, which the Bidy search engine knows where you are,
• 17:40
  Speaker 1: the way they can track who's going to shopping malls
• 17:43
  Speaker 1: and office parks and tourist sites and things like that
• 17:48
  Speaker 1: in real time, and they can create employment indexes from this,
• 17:52
  Speaker 1: where that's a great data point in a place where
• 17:56
  Speaker 1: the official unemployment rate nobody really looks at because it's
• 17:59
  Speaker 1: not reliable. Well, definitely a space to watch in the future.
• 18:04
  Speaker 1: Josh bloomin Stock Jeff Kerns, thank you so much for
• 18:06
  Speaker 1: joining us today on the podcast. Thank you guys. Thanks guys.
• 18:13
  Speaker 1: Benchmark will be back next week and until then, you
• 18:15
  Speaker 1: can find us on the Bloomberg terminal, Bloomberg dot com,
• 18:18
  Speaker 1: or Bloomberg App, as well as on Apple Podcasts, pocketcasts,
• 18:22
  Speaker 1: and Stitcher. While you're there, take a minute to rate
• 18:24
  Speaker 1: and review the show. So more listeners can find us
• 18:27
  Speaker 1: and let us know what you thought of the show.
• 18:29
  Speaker 1: You can follow me on Twitter at at scott Landman.
• 18:32
  Speaker 1: Dan You're at most Underschool Echo, Jeff you are at
• 18:37
  Speaker 1: Jeff Kern j E F F K E A r
• 18:40
  Speaker 1: n F and Josh you are at J bloomin stock
• 18:45
  Speaker 1: J B l U M E N S t O
• 18:46
  Speaker 1: c K. Benchmark is produced by Sarah Patterson. The head
• 18:50
  Speaker 1: of Bloomberg Podcasts is Francesca Levy. Thanks for listening, See
• 18:54
  Speaker 1: you next time.