1 00:00:00,000 --> 00:00:04,200 The following is part of Cornell Contemporary China Initiative Lecture Series 2 00:00:04,200 --> 00:00:08,420 under the Cornell East Asia Program. The arguments and viewpoints of this talk 3 00:00:08,420 --> 00:00:11,220 belong solely to the speaker. We hope you enjoy. 4 00:00:11,220 --> 00:00:16,640 Professor Lovely finished her PhD in economics - 5 00:00:16,640 --> 00:00:18,260 Economics 6 00:00:18,260 --> 00:00:22,980 In economics in 1989 at the University of Michigan 7 00:00:22,980 --> 00:00:30,090 and joined the faculty in Syracuse soon after where she's still at and is also 8 00:00:30,090 --> 00:00:33,780 chairing their International Relations program. 9 00:00:33,780 --> 00:00:40,320 She's also an Eggers scholar and is going to present to us today on some of 10 00:00:40,320 --> 00:00:46,739 her work on Chinese economy obviously but specifically labor markets. 11 00:00:46,740 --> 00:00:52,520 Let's all thank Professor Mary for coming down today and welcome her. 12 00:00:52,520 --> 00:00:56,520 You should realize how lucky you are. It's amazing that you have this many 13 00:00:56,520 --> 00:01:04,470 speakers in in a year, I mean it's almost 16, and I am the lead batter for the 14 00:01:04,470 --> 00:01:12,600 power women set - yeah series - so if you look at the CCI schedule it's gonna be 15 00:01:12,600 --> 00:01:17,850 all women this semester so it's gonna be an awesome semester. Gotta send some 16 00:01:17,850 --> 00:01:22,500 students down here to attend some of these. So what I'm gonna be talking 17 00:01:22,500 --> 00:01:28,140 about today I think is a fun paper and I'm very happy to present it and I think 18 00:01:28,140 --> 00:01:32,610 it showcases a bunch of different things that we can do in economics but it's 19 00:01:32,610 --> 00:01:41,640 accessible to all disciplines and, or at least I hope it is, and I think shows 20 00:01:41,640 --> 00:01:46,920 you the kinds of data that we're using, the kinds of questions that we're asking, 21 00:01:46,920 --> 00:01:52,079 and how economics can reveal some of the changes that are happening in 22 00:01:52,079 --> 00:01:56,219 contemporary China. I know that this is an interdisciplinary series and you're 23 00:01:56,219 --> 00:02:00,509 going to have a wide variety of speakers coming and that's really exciting. 24 00:02:00,509 --> 00:02:04,890 I run an interdisciplinary program myself, International Relations is an 25 00:02:04,890 --> 00:02:10,220 interdisciplinary program. I am the first member of my tribe, the econ tribe, 26 00:02:10,220 --> 00:02:14,120 to lead the International Relations program and it has been 27 00:02:14,120 --> 00:02:19,280 a great experience for me and a great learning experience to find out 28 00:02:19,280 --> 00:02:23,150 how much the other disciplines contribute to my own understanding of 29 00:02:23,150 --> 00:02:27,900 what's happening in contemporary China, so I look forward to your questions at the end. 30 00:02:27,900 --> 00:02:33,190 So here's my title "Economic Integration and Spatial Wage Variation: 31 00:02:33,190 --> 00:02:41,320 How valuable is market access in China?" And this is joint work with two of my 32 00:02:41,320 --> 00:02:49,040 graduate students. Yang Liang is my own PhD student and Hongsheng Zhang is a visitor. 33 00:02:49,040 --> 00:02:54,820 China has a very active China state scholarship program where advanced 34 00:02:54,829 --> 00:02:58,370 graduate students will come to the United States to study, and I've been 35 00:02:58,370 --> 00:03:01,980 very fortunate to have several of these young scholars with me. 36 00:03:01,980 --> 00:03:08,680 So our team put together this paper. What are the broad questions that this paper 37 00:03:08,690 --> 00:03:13,940 really speaks to? And the first one is whether we observe systematic spatial 38 00:03:13,940 --> 00:03:19,310 patterns in China's nominal wages. And nominal wages are the wages that are 39 00:03:19,310 --> 00:03:23,140 paid by the employer. Real wages are what you can think of as your 40 00:03:23,140 --> 00:03:27,420 consumption bundle because the difference between those of course is local prices. 41 00:03:27,420 --> 00:03:30,640 We all know that if you get a job offer in New York City and they 42 00:03:30,650 --> 00:03:35,030 offer you $50,000 you may be very flattered. Start to look at rents in New 43 00:03:35,030 --> 00:03:39,739 York City you're gonna realize that your living situation might even be, 44 00:03:39,739 --> 00:03:43,340 you know, you may feel poorer than you did as a graduate student or an 45 00:03:43,340 --> 00:03:46,699 undergraduate student which is not good. So you have to really take account of 46 00:03:46,700 --> 00:03:51,520 the local price index. Here we're not looking at that. We're looking at nominal wages, 47 00:03:51,520 --> 00:03:55,040 the wages that employers can pay, and this is a question that a 48 00:03:55,040 --> 00:03:59,569 lot of economists have asked over time. So for example why are wages higher in 49 00:03:59,569 --> 00:04:03,290 New York City than they are in say Syracuse, and a lot of people will answer 50 00:04:03,290 --> 00:04:07,840 well because rents cost so much in New York, but of course that's circular thinking. 51 00:04:07,840 --> 00:04:11,440 Employers wouldn't be able to pay wages and rents wouldn't be able to 52 00:04:11,440 --> 00:04:15,320 be so high, people weren't more productive in these major cities. 53 00:04:15,320 --> 00:04:21,820 So that has formed a lot of thinking both in urban economics and the literature that 54 00:04:21,820 --> 00:04:24,900 I'm drawing from today which is called the new economic geography. 55 00:04:24,900 --> 00:04:30,539 So I'll explain more as I go along but that's the basic area that 56 00:04:30,539 --> 00:04:37,770 we're gonna be working in. Now we do observe spatial variation in wages and 57 00:04:37,770 --> 00:04:43,590 what I want to ask is can we explain these in parts by differential access to 58 00:04:43,590 --> 00:04:49,760 consumer or supplier markets, and I will fill in the details on that as I move forward. 59 00:04:49,760 --> 00:04:52,340 What I'm really gonna be interested in is whether this 60 00:04:52,349 --> 00:04:56,909 relationship has changed over time in China and whether it's contributed to 61 00:04:56,909 --> 00:05:05,219 increasing the spatial inequality in China. So do these differences become 62 00:05:05,220 --> 00:05:10,040 more pronounced over time and are they contributing to increased inequality? 63 00:05:11,820 --> 00:05:16,560 So one of the interesting things in looking at Chinese wages is that they 64 00:05:16,560 --> 00:05:23,159 look more Western over time and that is both for good and for bad. The good part 65 00:05:23,160 --> 00:05:27,640 is of course that we just rose everywhere in China over the last 20 years, 66 00:05:27,640 --> 00:05:33,200 and that is true in rural areas and in urban areas, in eastern China and western China, 67 00:05:33,210 --> 00:05:39,040 for skilled workers and for unskilled workers. Many people point out that the 68 00:05:39,040 --> 00:05:42,599 transition of China to a market economy contributed to a tremendous 69 00:05:42,599 --> 00:05:49,110 decrease in the number of people living in abject poverty on Earth and that 70 00:05:49,110 --> 00:05:52,540 that is a major accomplishment, something we should all celebrate. 71 00:05:52,540 --> 00:05:58,860 Along with that came some changes in the rewards to individual people, so the rewards for 72 00:05:58,860 --> 00:06:04,919 so-called productive characteristics increase. So the college premium, that is 73 00:06:04,919 --> 00:06:09,629 the increase in your wage that you get for finishing college as opposed just to 74 00:06:09,629 --> 00:06:16,880 high school, increased. It was estimated to be 39% in 1995 and it was 88% 75 00:06:16,880 --> 00:06:22,860 by 2002 so you'd get 88% more for going to college. Now that of 76 00:06:22,860 --> 00:06:27,520 course mimics what we see in the West where there's a premium for going to college, 77 00:06:27,520 --> 00:06:30,000 and in fact one of the reasons why people do go to college - 78 00:06:30,000 --> 00:06:34,680 if I go to college and make this investment I will earn a higher wage - and we see 79 00:06:34,680 --> 00:06:42,800 that increasing wage premium in China. Not so happy is the gender wage gap. 80 00:06:42,800 --> 00:06:50,240 Wages of male workers relative to similar females were 11% higher in 1995, so not 81 00:06:50,249 --> 00:06:58,559 much of a wage gap, but it was 17% by 2002 and 30% by some estimates in 2007. 82 00:06:58,560 --> 00:07:03,660 So we're really seeing a big increase in gender wage inequality in China. 83 00:07:03,660 --> 00:07:07,940 I haven't studied that but I think it's a fascinating issue to study, of course we 84 00:07:07,940 --> 00:07:13,620 see it in U.S. wages and European wages and we don't really understand why it exists. 85 00:07:13,620 --> 00:07:17,320 We can control for lots of different things things and we still can't make it go away. 86 00:07:17,320 --> 00:07:21,780 We can make it smaller but we don't really know why this wage gap persists. 87 00:07:21,780 --> 00:07:24,029 One would think that if two people were equally productive 88 00:07:24,029 --> 00:07:29,460 competition in the labor market would drive those wage gaps to zero, but they haven't. 89 00:07:29,460 --> 00:07:33,580 And here we see that in China, actually as China has become more of 90 00:07:33,580 --> 00:07:37,940 a market economy, those wage gaps have actually increased. 91 00:07:37,949 --> 00:07:42,629 Lastly we see that echoes of the planned economy have faded over time and in 92 00:07:42,629 --> 00:07:47,189 particular earnings differentials between state and private sector 93 00:07:47,189 --> 00:07:53,550 decreased among urban workers. We'll be able to look at that question here, 94 00:07:53,550 --> 00:07:56,759 by the end we're not gonna see any significant difference. 95 00:07:56,760 --> 00:07:59,920 I take that back let me look at the regression results for sure, 96 00:07:59,920 --> 00:08:09,000 but absolutely the gap has decreased. Now what we're looking at is monetary compensation, 97 00:08:09,000 --> 00:08:11,639 there may be other forms of compensation for working in a 98 00:08:11,640 --> 00:08:15,440 state-owned enterprise that remain, and we're not going to capture those. 99 00:08:15,440 --> 00:08:22,700 It could be expectation of more stable employment or prestige or other factors, 100 00:08:22,709 --> 00:08:26,579 connections that you feel will be important professionally or personally, 101 00:08:26,580 --> 00:08:31,100 and those are not going to be captured by our measurements. 102 00:08:32,220 --> 00:08:37,680 So what is spatial wage variation? I didn't really define it before. I'm going to define it 103 00:08:37,680 --> 00:08:42,479 as variation in nominal wages that's not explained by differences in an 104 00:08:42,480 --> 00:08:47,420 individual workers characteristics, but rather it's tied to their location. 105 00:08:47,420 --> 00:08:52,520 So we take you, you are the same person you always were, and we move you to a 106 00:08:52,529 --> 00:08:58,320 different location and find out that you're worth more. Your nominal wage goes up. 107 00:08:58,320 --> 00:09:00,640 So that's spatial wage variation. 108 00:09:00,640 --> 00:09:05,140 And is there evidence of spatial wage variation in China, and the answer is yes. 109 00:09:05,140 --> 00:09:11,840 There's lots of different measures in the literature but what we did is just we 110 00:09:11,850 --> 00:09:17,209 ran a regression on individual wages of workers and we include a set of 111 00:09:17,209 --> 00:09:20,939 provincial dummy variables, or fixed effects, basically looking at what's the 112 00:09:20,939 --> 00:09:27,300 average within this province, and we find that it explains about 12% of the 113 00:09:27,300 --> 00:09:34,120 variation in wages in a sample of workers in 1995, 14% in 2002, 114 00:09:34,120 --> 00:09:41,399 and almost 13% in 2007, so pretty much has a stable effect on 115 00:09:41,399 --> 00:09:48,569 the variation in wages. So this says that spatial variation on wages mattered in 116 00:09:48,569 --> 00:09:58,559 1995 and it continues to matter in 2007, and it explains about 12-15% 117 00:09:58,559 --> 00:10:04,529 of the variation in individual wages. Now there's similar findings in the 118 00:10:04,529 --> 00:10:09,990 literature on urban income equality by people who have studied individual 119 00:10:09,990 --> 00:10:13,920 survey data or information on wages for individuals where they're able to 120 00:10:13,920 --> 00:10:18,620 control for individual workers' characteristics, such as education 121 00:10:18,620 --> 00:10:24,600 so important experience, and between 12 and 20% of urban wage inequality is 122 00:10:24,600 --> 00:10:31,679 attributed to differences in mean provincial incomes. So spatial variation 123 00:10:31,680 --> 00:10:36,560 in wages is important to China and understanding wage variation, 124 00:10:36,560 --> 00:10:41,640 which of course is a key component in understanding income inequality. 125 00:10:41,640 --> 00:10:49,820 So one possibility is that market access, which I'm going to define now, is a reason for 126 00:10:49,829 --> 00:10:57,700 these wage differences, and if that's true then these wage differences should be systematically related 127 00:10:57,700 --> 00:11:02,700 to what we think of as market access of a city; how good is the location. 128 00:11:02,700 --> 00:11:07,220 Now we might think wages in New York are high for a lot of different reasons - 129 00:11:07,220 --> 00:11:12,840 New York City that is. One might be that you have a lot of other creative people 130 00:11:12,840 --> 00:11:17,400 there so there's a lot of spillovers. Another might be that originally 131 00:11:17,400 --> 00:11:21,660 New York was chosen for its excellent location and that it was near to a lot of 132 00:11:21,670 --> 00:11:29,920 markets both within the US and overseas. So what we have in the literature 133 00:11:29,920 --> 00:11:34,440 especially stemming from Paul Krugman, who now you may know as a famous 134 00:11:34,440 --> 00:11:38,400 New York Times columnist and many people think of him as a political 135 00:11:38,400 --> 00:11:43,460 commentator but of course he did some very important fundamental work in economics - lots, 136 00:11:43,460 --> 00:11:47,320 that's why he won the Nobel Prize in Economics, so this is that Paul Krugman. 137 00:11:47,320 --> 00:11:52,030 Krugman and Venables showed theoretically that 138 00:11:52,030 --> 00:11:56,350 when we have scale economies at the level of the firm, that means bigger 139 00:11:56,350 --> 00:12:04,510 firms are more productive, and trade is costly, local economic activity is 140 00:12:04,510 --> 00:12:10,270 influenced by proximity to external markets. Workers are more productive in 141 00:12:10,270 --> 00:12:15,310 locations that have good locations. Now what does that mean? So the basic idea is 142 00:12:15,310 --> 00:12:18,910 that firms wanna concentrate production because they have economies 143 00:12:18,910 --> 00:12:23,140 of scale at the firm level. What that means is they have to build a 144 00:12:23,140 --> 00:12:32,160 plant or they have to build a set of other services, say accounting and other services, 145 00:12:32,160 --> 00:12:36,390 that they're going to use to provide to the rest of the 146 00:12:36,390 --> 00:12:42,540 company and because of that they want to concentrate production in one location. 147 00:12:42,540 --> 00:12:48,700 Now where are you going to choose to put that activity? The theory tells us that 148 00:12:48,700 --> 00:12:54,960 you're going to want to put that plant or business in the location that 149 00:12:54,960 --> 00:13:01,200 minimizes the cost of getting your goods to consumers. So if we think of 150 00:13:01,200 --> 00:13:06,080 why do we have Chicago where Chicago is, of course it was on the Great Lakes 151 00:13:06,080 --> 00:13:11,680 and it could access Midwest supplies and also send 152 00:13:11,680 --> 00:13:17,420 goods east through the Great Lakes and then down into the Erie Canal, a long time ago. 153 00:13:17,420 --> 00:13:19,480 So if we think about all the great cities in the world 154 00:13:19,480 --> 00:13:23,320 a lot of them have these great locations. So what we're really saying is that 155 00:13:23,320 --> 00:13:28,870 there should be this systemic relationship between this idea of having 156 00:13:28,870 --> 00:13:33,660 a good location because of proximity to consumer markets and suppliers, 157 00:13:33,660 --> 00:13:39,000 and the wage that workers - that firms are able to pay in that location. 158 00:13:43,960 --> 00:13:50,360 Now this theory is quite rich and there are other propositions that flow from the theory. 159 00:13:50,370 --> 00:13:55,720 One is that we will have trade induced agglomeration. What that means is that in 160 00:13:55,720 --> 00:14:00,520 these locations you'll have a lot of businesses locating so you'll get 161 00:14:00,520 --> 00:14:04,090 what economists call agglomeration, which is just a lot of firms being 162 00:14:04,090 --> 00:14:09,280 together in one space. So that's not a different explanation, it's sort of part 163 00:14:09,280 --> 00:14:15,550 of the explanation. We'll get factor inflows - workers will come into the area to 164 00:14:15,550 --> 00:14:20,740 take advantage of the higher wages and the jobs that are created. We'll get 165 00:14:20,740 --> 00:14:25,480 something that's called the home market effect. If the people in that area have a 166 00:14:25,480 --> 00:14:31,180 particular preference for goods those goods will then be produced in 167 00:14:31,180 --> 00:14:34,180 large quantities, because they're produced in large quantities with scale 168 00:14:34,180 --> 00:14:38,980 economies will be very price competitive and will tend to export those goods. 169 00:14:38,980 --> 00:14:45,700 So we can think of differences in the types of cars for example or different 170 00:14:45,700 --> 00:14:51,280 types of foods that people like in local areas, that would explain why some areas export those, 171 00:14:51,280 --> 00:14:55,700 so we'll get a home market effect. And we'll also have shock sensitivity 172 00:14:55,700 --> 00:14:59,140 and what shock sensitivity means is that if there's some reason why 173 00:14:59,140 --> 00:15:02,940 the area is able to grow faster than other regions for a short period of time, 174 00:15:02,940 --> 00:15:08,230 that will not be reduced. Because firms will come in, it will support the 175 00:15:08,230 --> 00:15:11,950 higher productivity. And so you can get these situations where the government 176 00:15:11,950 --> 00:15:17,850 perhaps can intervene, create a situation where firms move in, 177 00:15:17,850 --> 00:15:22,680 and then the firms will stay. So we have a variety of different results that come 178 00:15:22,680 --> 00:15:27,240 from the theory. Today I'm only going to be looking at the idea that wages should 179 00:15:27,240 --> 00:15:32,730 be higher in areas that have better access to consumer and producer and 180 00:15:32,730 --> 00:15:38,759 supplier markets. Now what is market access? Market access is going to be a 181 00:15:38,759 --> 00:15:44,850 characteristic of the workers' location, so New York City, or Beijing, 182 00:15:44,850 --> 00:15:51,089 more appropriately, or Shanghai or Shenzhen. So market access, sometimes also 183 00:15:51,089 --> 00:15:55,589 called real market potential by certain authors, is the characteristic of the 184 00:15:55,589 --> 00:16:00,820 workers' location and the concept is drawn from new economic geography. 185 00:16:00,820 --> 00:16:07,200 How close is this location to potential consumers and suppliers? 186 00:16:07,220 --> 00:16:14,279 Now each potential market for a firm's exports has a sales capacity. You might think you 187 00:16:14,279 --> 00:16:19,050 want to be close to big markets. Well not just big, if you think about it 188 00:16:19,050 --> 00:16:23,220 for a minute, you'd also like to be next to big rich markets. 189 00:16:23,220 --> 00:16:28,640 But new economic geography tells us that it's more than just being near big rich markets. 190 00:16:28,640 --> 00:16:34,399 In fact the idea that locations that were near big rich markets has a 191 00:16:34,400 --> 00:16:39,500 intellectual history that goes back probably to the 1930s, maybe even before. 192 00:16:39,500 --> 00:16:46,860 Urban geographers, regional economists, all thought of this idea of location 193 00:16:46,860 --> 00:16:50,700 and access to markets. But they really would just look 194 00:16:50,700 --> 00:16:56,600 at GDP or earning power, expenditures in a market. New economic geography 195 00:16:56,600 --> 00:17:02,910 incorporates ideas from industrial organization and they say it's not just 196 00:17:02,910 --> 00:17:07,020 being close to big rich markets, it's also how contestable those markets is - 197 00:17:07,020 --> 00:17:11,130 can I get my product into those markets. So they wanna take into account the 198 00:17:11,130 --> 00:17:17,610 number of competitors that I will face in that market and the prices 199 00:17:17,610 --> 00:17:23,480 that those competitors charge, so how aggressive are they in locating there. 200 00:17:23,480 --> 00:17:27,500 So that means that we can't just use the GDP of all the places that are 201 00:17:27,500 --> 00:17:32,940 surrounding Shanghai or New York City to measure the market access 202 00:17:32,940 --> 00:17:36,860 or the market potential of locating my firm there, I have to also 203 00:17:36,860 --> 00:17:41,740 consider how aggressive competition is going to be in those markets. 204 00:17:41,740 --> 00:17:47,220 So that's something that's different than the older literature. So the market access 205 00:17:47,220 --> 00:17:52,520 of a city, city j, is measured as the distance weighted sum of the capacity of 206 00:17:52,520 --> 00:17:59,100 all locations into which firms in city j might sell. So I'm going to weight this 207 00:17:59,100 --> 00:18:05,490 idea of these rich contestable markets to characterize the market 208 00:18:05,490 --> 00:18:11,540 access of city j. Now the challenge is operationalizing this concept, 209 00:18:11,540 --> 00:18:16,320 like all things we may have a nice theory and then we have data and the two sometimes 210 00:18:16,320 --> 00:18:22,740 don't have this nice matching. In this case the theory leads 211 00:18:22,740 --> 00:18:27,120 us naturally to what's called a gravity equation, and I'll talk about that in a minute 212 00:18:27,120 --> 00:18:30,450 but we're actually able to recover some parameters that look a lot 213 00:18:30,450 --> 00:18:33,570 like what the theory is talking about, and I'll talk about that in a minute. 214 00:18:33,570 --> 00:18:38,910 And also we're able to actually estimate the weighting term so we want to weight 215 00:18:38,910 --> 00:18:44,400 a market, discount a market that's farther away more than a discount that's 216 00:18:44,400 --> 00:18:50,190 closer because, among other things, I have to spend more money to get my goods to 217 00:18:50,190 --> 00:18:55,300 the faraway market. So how are we going to do that? 218 00:18:55,300 --> 00:18:59,280 I'll go and tell you that in a minute, but first I'm going to try to convince you that this 219 00:18:59,280 --> 00:19:02,900 market access measure, which I'm going to create in a minute, actually matters. 220 00:19:02,900 --> 00:19:07,720 So there was quite a bit of evidence from developed countries that appeared 221 00:19:07,730 --> 00:19:15,750 after 2000, so Redding's, Steve Redding, and Venables in 2004 published a paper 222 00:19:15,750 --> 00:19:20,310 where they find that the GDP per capita is higher in locations with better 223 00:19:20,310 --> 00:19:25,140 market access, and they used a very large sample of countries - 110 countries. 224 00:19:25,140 --> 00:19:29,840 It's mostly developed countries but they include some developing or emerging 225 00:19:29,850 --> 00:19:33,960 market countries in that sample obviously to get to 110, and they did it 226 00:19:33,960 --> 00:19:40,660 for the period 1992 to 1996. So people seem to be richer in areas that have 227 00:19:40,660 --> 00:19:43,670 better market access. 228 00:19:43,770 --> 00:19:49,170 Head and Meyer, 2006, they investigated the dispersion of average wages, mean wages, 229 00:19:49,170 --> 00:19:55,260 across 15 European regions for 1995 to 2000, and they find this 230 00:19:55,260 --> 00:20:00,780 relationship that wages, nominal wages, are higher in locations, regions of Europe, 231 00:20:00,780 --> 00:20:06,100 that have better market access. Now interestingly enough in 2010 232 00:20:06,100 --> 00:20:11,700 Laura Hering and Sandra Poncet published a paper where they use data from China and 233 00:20:11,700 --> 00:20:14,980 it's really the first evidence we have from a developing country, 234 00:20:14,980 --> 00:20:19,020 so I'll explain in the next slide, their paper also had some other fine characteristics 235 00:20:19,020 --> 00:20:24,690 that we're going to continue to use in our study. So they showed that 236 00:20:24,690 --> 00:20:28,679 individual wages, that is wages for individual workers and controlling for 237 00:20:28,679 --> 00:20:33,179 the characteristics of these workers, is higher in Chinese cities with better 238 00:20:33,179 --> 00:20:40,050 market access. They only had data from 1995 and as we know a lot of things have 239 00:20:40,050 --> 00:20:44,180 happened in China since 1995 and that's really going to be our opening because 240 00:20:44,180 --> 00:20:49,800 'cause we're gonna revisit this relationship. They found that the relationship between 241 00:20:49,800 --> 00:20:55,420 market access and nominal wages did hold for people but not for everybody. 242 00:20:55,420 --> 00:20:58,410 It didn't hold for people who were working in state firms, state-owned 243 00:20:58,410 --> 00:21:02,790 enterprises, and it didn't hold for unskilled workers. So basically they 244 00:21:02,790 --> 00:21:06,750 found the relationship for skilled workers working in private and 245 00:21:06,750 --> 00:21:10,440 particularly in foreign firms. Now you might say well there really weren't, 246 00:21:10,440 --> 00:21:13,440 in the sample that they used, there really weren't a lot of workers working 247 00:21:13,440 --> 00:21:17,040 in the foreign invested enterprises so maybe this was just some kind of 248 00:21:17,040 --> 00:21:20,910 other premium that they weren't able to control for. So we're gonna revisit 249 00:21:20,910 --> 00:21:27,960 that and in particular we want to ask what happened after 1995. So their paper 250 00:21:27,960 --> 00:21:31,679 was one of the first, in fact probably the first paper, to use individual survey 251 00:21:31,680 --> 00:21:36,860 data from a developing country to study this relationship between geography and wages. 252 00:21:39,540 --> 00:21:47,420 Unlike the work by, go back here, Redding and Venables and Head and Meyer 253 00:21:47,420 --> 00:21:50,560 you'll see that they're both really using just average measures - 254 00:21:50,560 --> 00:21:58,080 GDP per capita or the average wage in a region. Unlike them Hering and Poncet 255 00:21:58,080 --> 00:22:02,520 were able to use survey data from the Chinese Household Income Project 256 00:22:02,520 --> 00:22:11,720 to look at each wage and control for the workers' characteristics - gender, 257 00:22:11,720 --> 00:22:17,780 age, which we might think is associated with experience on the job, education, 258 00:22:17,780 --> 00:22:23,570 region of the country - so a lot of things that you think would matter for the wage. 259 00:22:23,570 --> 00:22:30,240 And they then use this same regression model to say well is market 260 00:22:30,240 --> 00:22:34,080 access of the location another determinant of the wage, and they 261 00:22:34,080 --> 00:22:37,230 estimate what we're going to call the elasticity of the local wage with 262 00:22:37,230 --> 00:22:43,410 respect to the local market access. So how much does your wage increase when 263 00:22:43,410 --> 00:22:49,320 you move to a city that has a better location? So I've said before this 264 00:22:49,320 --> 00:22:52,500 innovation is important because there's so many things that can determine a wage 265 00:22:52,500 --> 00:22:56,970 and we may find this relationship between average wages in location simply 266 00:22:56,970 --> 00:23:03,930 because workers like to go to places by the sea or something else that says that 267 00:23:03,930 --> 00:23:08,550 better workers are going to those locations. It's when we are able to 268 00:23:08,550 --> 00:23:13,320 control for individual characteristics we're able to get a better test of the theory. 269 00:23:15,340 --> 00:23:19,520 The other thing is that they also use a better wage that is closer to the 270 00:23:19,520 --> 00:23:31,440 theoretically based measure to measure the idea of market access of a location. 271 00:23:32,740 --> 00:23:37,400 So let me try to explain that a little more. What we're going to do is 272 00:23:37,410 --> 00:23:42,150 we're going to extend their study using three waves of the same survey that they 273 00:23:42,150 --> 00:23:45,940 used for their study, which is called the Chinese Household Income Project. 274 00:23:45,940 --> 00:23:56,240 This is data that is widely available so you can access it. And they estimate the value, 275 00:23:56,240 --> 00:24:00,180 we're gonna estimate the value of market access as reflected in local 276 00:24:00,180 --> 00:24:07,740 nominal wages for three periods, 1995, 2002, and 2007. We're going to use methods 277 00:24:07,740 --> 00:24:14,010 that come from new economic geography to measure the capacity of the markets into 278 00:24:14,010 --> 00:24:18,780 which a location might sell, so we're going to use methods to actually capture 279 00:24:18,780 --> 00:24:25,110 that idea of largeness, richness, and also contestability. We're going to test 280 00:24:25,110 --> 00:24:31,110 whether wages have become more or less influenced by market access of the 281 00:24:31,110 --> 00:24:36,330 workers' location over time. So in their paper Hering and Poncet begin by talking 282 00:24:36,330 --> 00:24:42,400 about all the incredible reforms that are undertaken in China after 1995. 283 00:24:42,400 --> 00:24:48,260 We know that the changes to the labor market were really quite startling, 284 00:24:48,260 --> 00:24:56,660 particularly the decrease or downsizing of state-owned enterprises and the basic 285 00:24:56,870 --> 00:25:01,620 forcing of people out into the private labor markets that happened in Chinese 286 00:25:01,620 --> 00:25:07,110 cities in the second half of the 1990s. It's amazing how little the West really 287 00:25:07,110 --> 00:25:11,220 knows about this period in China because these changes were really seismic, 288 00:25:11,220 --> 00:25:19,060 they were enormous, and it also led to tremendous growth in the Chinese private sector. 289 00:25:19,060 --> 00:25:24,360 So these are tremendous changes, we also know that there are other things 290 00:25:24,360 --> 00:25:30,860 that happened. China began to integrate more internally, a decrease in barriers 291 00:25:30,860 --> 00:25:38,220 across provinces to trade. China increased its integration into the world economy, 292 00:25:38,220 --> 00:25:43,920 particularly by joining the WTO. So these are really big changes and 293 00:25:43,920 --> 00:25:47,970 Hering and Poncet speculate that market access is going to play a more 294 00:25:47,970 --> 00:25:52,580 important role in determining wages moving forward but they are unable to 295 00:25:52,580 --> 00:25:56,700 test it because they only use the 1995 sample. 296 00:25:58,380 --> 00:26:03,920 So we're gonna reconsider their results. First of all is market access more or less important? 297 00:26:03,930 --> 00:26:09,990 Secondly over time do we see all types of workers being influenced, both skill 298 00:26:09,990 --> 00:26:13,700 and unskilled, those in the private sector, those in the state sector, 299 00:26:13,700 --> 00:26:17,160 and we're going to be able to address a couple of technical issues 300 00:26:17,160 --> 00:26:21,060 that may have plagued their study, in particular issues of being unable to 301 00:26:21,060 --> 00:26:27,380 control for individual ability. So I will show you that as we move along. 302 00:26:27,380 --> 00:26:29,820 Now first of all you might want to say well what kind of 303 00:26:29,820 --> 00:26:34,110 predictions would we expect. I sort of raise the idea that over time we might 304 00:26:34,110 --> 00:26:39,050 expect to see market access being more important to wages because Chinese 305 00:26:39,050 --> 00:26:43,020 cities are becoming more integrated with other cities in China and more 306 00:26:43,020 --> 00:26:47,730 integrated with the world. Importantly they're also becoming, labor markets are 307 00:26:47,730 --> 00:26:51,390 looking more like Western labor markets over time, that is people are 308 00:26:51,390 --> 00:26:55,380 being hired and fired, they have more people who are not in the state sector, 309 00:26:55,380 --> 00:27:01,070 don't have permanent protections, aren't getting a variety of other types of 310 00:27:01,070 --> 00:27:04,770 benefits from being in the state sector and therefore a more likely to switch 311 00:27:04,770 --> 00:27:11,160 jobs voluntarily. So here's a couple of predictions. After 1995 China reduced 312 00:27:11,160 --> 00:27:17,600 barriers to interprovincial trade so the so-called interior border costs will fall, 313 00:27:17,600 --> 00:27:25,320 raising the value of interior proximity. One of the big questions at 314 00:27:25,320 --> 00:27:31,320 the beginning of this century, so around - between 2000 and 2005 say, was how 315 00:27:31,320 --> 00:27:34,800 fragmented were Chinese markets internally? And in fact we still have some people 316 00:27:34,800 --> 00:27:39,140 working on this issue. And there was quite a bit of controversy. 317 00:27:39,140 --> 00:27:42,900 Some people said there were lots, some people said there was little, and we know there 318 00:27:42,900 --> 00:27:46,260 are still administrative rules, particularly automobile sector, 319 00:27:46,260 --> 00:27:53,660 that fragment markets. My favorite one, and you don't correct me if I'm wrong, is about 320 00:27:53,670 --> 00:27:57,030 the cost of registering your car. Do they still have differential 321 00:27:57,030 --> 00:28:01,410 registration rates by where your car is made? Yeah so this is fun for 322 00:28:01,410 --> 00:28:06,080 especially for Americans just seems like just seems like "whaaaat," 323 00:28:06,080 --> 00:28:11,040 so if your car is made in Shanghai and you want to register in Beijing you're gonna 324 00:28:11,040 --> 00:28:15,210 pay one price which is higher than you would pay if you're registering a 325 00:28:15,210 --> 00:28:19,650 car in Beijing that was made in Beijing. So to an American that's like "whaaaat" 326 00:28:19,650 --> 00:28:24,780 because we just take a car we go across the city or we go from California to 327 00:28:24,780 --> 00:28:28,620 Ithaca, New York to go to college and we just pay whatever everybody else pays, 328 00:28:28,620 --> 00:28:33,030 right, no matter where our car was made. So there were these barriers that the 329 00:28:33,030 --> 00:28:35,350 West was kind of surprised to discover 330 00:28:35,350 --> 00:28:40,960 existed but the estimates on them that in terms of whether prices were the same 331 00:28:40,960 --> 00:28:44,860 across different locations in some cases show that they were pretty, 332 00:28:44,860 --> 00:28:50,920 these barriers really did influence living costs and living standards in different areas of China. 333 00:28:50,920 --> 00:28:54,429 So as these barriers fall what does that 334 00:28:54,429 --> 00:29:00,059 mean for market access? What I am interested in though is that if you are say 335 00:29:00,060 --> 00:29:08,380 Chengdu and you're inside China, interior, but you have a pretty good location, 336 00:29:08,380 --> 00:29:11,950 you're near a lot of other interior cities, when your ability to 337 00:29:11,950 --> 00:29:17,320 reach those other cities is freed you should have a better location. 338 00:29:17,320 --> 00:29:24,520 So interior borders really privileged the coastal cities versus the interior cities. 339 00:29:24,520 --> 00:29:28,480 So what we should be seeing over time is that interior cities become 340 00:29:28,480 --> 00:29:34,248 relatively more attractive for firms to locate and higher wages will appear there. 341 00:29:34,248 --> 00:29:36,860 The second prediction we have is that 342 00:29:36,860 --> 00:29:41,140 after 1995 China reduced trade and investment barriers and in particular 343 00:29:41,140 --> 00:29:44,950 they ended what was the so-called export enclaves. If you think about 344 00:29:44,950 --> 00:29:49,539 how China developed it opened up special economic zones, coastal cities that 345 00:29:49,539 --> 00:29:55,630 really try to contain the influence of both foreign direct investment and the 346 00:29:55,630 --> 00:30:02,000 so-called export sector. Over time of course those barriers have decreased, 347 00:30:02,000 --> 00:30:07,539 FDI has penetrated way into the heartland of the country, and labor flowed 348 00:30:07,539 --> 00:30:11,409 more freely back and forth. That means that we should see a stronger impact of 349 00:30:11,409 --> 00:30:17,559 these market forces on all wages because they're able to be reflected generally 350 00:30:17,560 --> 00:30:23,280 across the economy. So what we predict is that the elasticity of the local wage 351 00:30:23,280 --> 00:30:27,300 with respect to market access will rise, in other words that location 352 00:30:27,309 --> 00:30:33,970 will become more valuable over time. And lastly is the idea that after 1995 China 353 00:30:33,970 --> 00:30:39,120 really liberalized the wage setting, particularly for state enterprises. 354 00:30:39,120 --> 00:30:46,020 Wages are largely freed by 2000 and what that means is that all wages should begin to 355 00:30:46,030 --> 00:30:49,310 fully reflect the so-called market forces 356 00:30:49,310 --> 00:30:55,260 for better or for worse. Prior to this Chinese wages were set by administrative 357 00:30:55,260 --> 00:30:58,660 formula that had some local market 358 00:30:58,660 --> 00:31:05,140 adjustment but otherwise were set by a formula 359 00:31:05,140 --> 00:31:10,240 so they did not completely reflect and, in fact originally in the early 1990s, 360 00:31:10,240 --> 00:31:16,820 did not reflect the local market conditions at all or very little. So we should see 361 00:31:16,820 --> 00:31:24,950 in our data that contrary to what Hering and Poncet found in 1995, by 2002 and 2007 362 00:31:24,950 --> 00:31:29,090 we should see market access mattering not just for private firms or for 363 00:31:29,090 --> 00:31:32,520 foreign invested enterprises but for all workers. 364 00:31:34,760 --> 00:31:36,840 So what's our basic strategy 365 00:31:36,840 --> 00:31:42,140 and this is mainly an empirical paper and we're going to pull together data 366 00:31:42,140 --> 00:31:46,380 from a lot of different sources to carry out the following activities. 367 00:31:46,380 --> 00:31:51,395 The first thing we're gonna do is we're gonna estimate a gravity model of trade flows. 368 00:31:51,400 --> 00:31:54,200 I don't know if you've ever heard about gravity models before, 369 00:31:54,200 --> 00:31:58,820 they're quite familiar in economics and in geography, the idea 370 00:31:58,820 --> 00:32:04,670 basically is that we can explain trade between two regions as a bunch of pretty 371 00:32:04,670 --> 00:32:11,300 standard characteristics including their weight, which would be their GDP, and their location. 372 00:32:11,300 --> 00:32:14,840 We can amend that for a lot of other things like whether the 373 00:32:14,840 --> 00:32:20,120 regions share the same language or a same colonial past that would have 374 00:32:20,120 --> 00:32:26,470 established certain trading relationships. International economists 375 00:32:26,470 --> 00:32:31,700 estimate these things all the time but we're going to use it for our own 376 00:32:31,700 --> 00:32:35,630 purposes here coming out of a new economic geography which is we're going 377 00:32:35,630 --> 00:32:41,200 to use them to recover kind of the average weight of each possible market, 378 00:32:41,200 --> 00:32:45,290 and we're going to calculate another parameter. So we're gonna get from 379 00:32:45,290 --> 00:32:49,760 there an estimate of the weight of each market. That is gonna be our measure 380 00:32:49,760 --> 00:32:54,940 of that market's capacity. So if I'm New York and I'm close to Philadelphia 381 00:32:54,940 --> 00:32:59,740 I'm gonna measure the weight of Philadelphia by the average amount of 382 00:32:59,750 --> 00:33:03,340 sales that everyone makes to Philadelphia. That's really what the 383 00:33:03,340 --> 00:33:08,100 regression is going to pick up. I'm also gonna be able to estimate in that 384 00:33:08,100 --> 00:33:16,000 regression how much distance diminishes trade, so I'm gonna get this distance 385 00:33:16,000 --> 00:33:20,560 discounting parameter. So one of the problems in the old literature going 386 00:33:20,560 --> 00:33:25,240 back to the 1950s urban planning and geography literature was that we knew 387 00:33:25,240 --> 00:33:30,250 that a city had an advantage if it was close to a lot of big rich markets but we 388 00:33:30,250 --> 00:33:35,100 didn't really know how to discount faraway markets compared to close markets. 389 00:33:35,100 --> 00:33:40,560 So we would try a whole bunch of things - divide it, just put in the GDP, 390 00:33:40,560 --> 00:33:47,460 or divide it by distance, or distance taken to some parameter, 1, 1.1, 1.2 - 391 00:33:47,460 --> 00:33:50,640 we do a lot of these things and see if it mattered to our results. 392 00:33:50,640 --> 00:33:55,380 By estimating the gravity model we're gonna see on average how much is each 393 00:33:55,390 --> 00:34:01,780 additional kilometer of distance between city i and city j reduce trade 394 00:34:01,780 --> 00:34:07,360 and we're gonna be able to use that in calculating the market access of each city. 395 00:34:07,360 --> 00:34:10,090 So just to recap the gravity models are 396 00:34:10,090 --> 00:34:13,320 gonna give us two really important pieces of information. 397 00:34:13,320 --> 00:34:19,080 First, the weight of each possible market to which I might sell from my location, 398 00:34:19,080 --> 00:34:30,000 and the discount parameter that I need to weight cities that are closer versus far away. 399 00:34:30,000 --> 00:34:33,300 The second step we're gonna do is we're going to turn to the 400 00:34:33,300 --> 00:34:37,800 Chinese Household Income Project and take their survey data and it gives data on 401 00:34:37,810 --> 00:34:43,179 individuals and I'm gonna know what their nominal wage is and I'm going to 402 00:34:43,180 --> 00:34:46,840 know their individual characteristics as well as their location. 403 00:34:46,840 --> 00:34:54,220 Of course location is a key variable for me. And I'm going to run what's called a Mincer 404 00:34:54,220 --> 00:34:58,900 wage regression, it just says the log of the wage is a function of a whole bunch 405 00:34:58,900 --> 00:35:04,380 of things - your education, your gender your age, your region I'm gonna 406 00:35:04,380 --> 00:35:11,140 include region - to sector and then I'm gonna add the market access of your city, 407 00:35:11,140 --> 00:35:14,180 so we know what city you're in and I can add that through regression, 408 00:35:14,180 --> 00:35:16,620 now I'm gonna see if that's a significant determinant 409 00:35:16,630 --> 00:35:21,940 of the wages of people who live in that city. And then lastly I'm going to 410 00:35:21,940 --> 00:35:26,950 pull some other information from the Chinese Household Income Project surveys 411 00:35:26,950 --> 00:35:31,600 and from other sources to control for other possible explanations for the 412 00:35:31,600 --> 00:35:36,100 relationship that I find, so-called robustness checks, including selection of 413 00:35:36,100 --> 00:35:41,160 better people into better cities. That's something that's oftentimes mentioned. 414 00:35:41,160 --> 00:35:45,880 So wages in Beijing are higher because people in Beijing have access to Tsinghua 415 00:35:45,880 --> 00:35:50,200 or Beida so that's why this is happening. And so I'm gonna show you 416 00:35:50,200 --> 00:35:53,770 some of the things we do to try to guard against that. Or there are regional 417 00:35:53,770 --> 00:35:57,880 characteristics or other forms of spillovers that account for this and we're 418 00:35:57,880 --> 00:36:02,920 gonna try to account for that and see if our basic result is robust to the 419 00:36:02,920 --> 00:36:08,050 inclusion of those variables. A key thing is then measuring the market 420 00:36:08,050 --> 00:36:14,350 access of my city, city j, and the market access of city j is just a weighted sum 421 00:36:14,350 --> 00:36:20,470 of the expenditure in markets to which firms in city j might sell, so it's just 422 00:36:20,470 --> 00:36:27,220 the sum over, and then we have some funny things here. We have a ϕ an E and an S 423 00:36:27,220 --> 00:36:30,760 and that's going to be the things that I'm going to have to take from my 424 00:36:30,760 --> 00:36:36,550 gravity model. So expenditure is E and we know that we're going to want to sum up 425 00:36:36,550 --> 00:36:41,100 the expenditure of all the markets to which I might sell from location j and 426 00:36:41,100 --> 00:36:45,440 I've already said that I'm going to get that as this average expenditure, 427 00:36:45,440 --> 00:36:53,500 average sales, from the gravity model. But then I need to take account of supplier access 428 00:36:53,500 --> 00:36:57,300 I said I have to take account of how difficult it is to penetrate those models, 429 00:36:57,300 --> 00:37:00,800 and in fact that's already embedded in the gravity equation. 430 00:37:00,800 --> 00:37:03,900 The gravity equation doesn't just give me back GDP, 431 00:37:03,910 --> 00:37:08,770 it gives me how many, what's the average sales, so it in sense already 432 00:37:08,770 --> 00:37:13,960 incorporates this in the estimate. So I've got the E and I've got the S; the 433 00:37:13,960 --> 00:37:18,100 problem now is the ϕ's and this is actually called the 434 00:37:18,100 --> 00:37:22,960 phi-ness believe it or not, economists are just loads of fun, and the phi-ness is 435 00:37:22,960 --> 00:37:26,609 just the ease of access between producers in city j and 436 00:37:26,609 --> 00:37:31,499 consumers in region i. Now you might think it's just distance, we know 437 00:37:31,500 --> 00:37:35,160 that distance is a separator, but it's more than that. 438 00:37:35,160 --> 00:37:41,000 We may have an intracountry border as opposed to an intercountry border 439 00:37:41,000 --> 00:37:46,840 so selling between New York and Connecticut is a lot easier than between New York and Canada. 440 00:37:46,840 --> 00:37:51,020 You might not think so but consistently when we estimate gravity 441 00:37:51,029 --> 00:37:55,910 models we see that trade between two locations one in New York and one in 442 00:37:55,910 --> 00:38:01,529 Connecticut is much higher than between someone in New York and someone 443 00:38:01,529 --> 00:38:06,690 similarly located in Canada, so we have these other types of border boundaries 444 00:38:06,690 --> 00:38:10,650 which may matter and we're going to have to include that in our phi-ness measure. 445 00:38:10,650 --> 00:38:18,080 So it's distance and other forms of trade costs which may be tariffs or it 446 00:38:18,080 --> 00:38:21,360 may just be the time it takes to get through borders. 447 00:38:23,300 --> 00:38:29,380 So as I say how do we estimate these, we're going to estimate these with a gravity model as I said before. 448 00:38:29,380 --> 00:38:35,970 Now that seemed like a great idea at first until we realized that it's 449 00:38:35,970 --> 00:38:39,660 actually pretty tough and the reason why it's tough and why Hering and Poncet 450 00:38:39,660 --> 00:38:45,509 only did it for one year is because we have to take account of the fact that 451 00:38:45,509 --> 00:38:49,859 Chinese cities are each other's markets as well as foreign countries being 452 00:38:49,859 --> 00:38:57,720 markets for China. So early research looked just at how close some cities 453 00:38:57,720 --> 00:39:01,780 were to international markets and found out that wages were higher there, 454 00:39:01,780 --> 00:39:06,620 but of course the cities on the coasts are the ones that are closest to international markets, 455 00:39:06,620 --> 00:39:08,860 there's three major international ports in China, 456 00:39:08,860 --> 00:39:13,859 and so those cities were privileged in this. That doesn't get at the concept that 457 00:39:13,859 --> 00:39:16,680 we're thinking about which is that Chinese cities are growing so much 458 00:39:16,680 --> 00:39:22,619 faster than the rest of the world. If you're an interior city like Chengdu 459 00:39:22,620 --> 00:39:26,720 are you benefiting over time from increasing market integration, 460 00:39:26,720 --> 00:39:31,940 that's really what we want to get at. So we need not only sales from individual Chinese 461 00:39:31,950 --> 00:39:40,170 cities to individual countries, city or province to foreign country, 462 00:39:40,170 --> 00:39:45,270 we need province to other provinces. And the only type of way we could possibly 463 00:39:45,270 --> 00:39:50,340 do that is by looking at the Chinese customs records. So the Chinese customs records 464 00:39:50,340 --> 00:39:56,840 will also give us the location of the production so we're from and to. 465 00:39:57,580 --> 00:40:00,960 Now Hering and Poncet, called HP here 466 00:40:00,960 --> 00:40:04,260 that's not Hewlett-Packard that's actually Hering and Poncet, 467 00:40:04,260 --> 00:40:12,720 they did this for 1995 because Sandra Poncet had access to the 1997 customs records. 468 00:40:12,720 --> 00:40:16,530 I did not have access to that for a long time and earlier on I just said 469 00:40:16,530 --> 00:40:19,920 okay I don't have access to that I'll just use the old fashioned so-called 470 00:40:19,920 --> 00:40:24,119 Harris measure which is just distance weighted GDP, which works pretty well 471 00:40:24,119 --> 00:40:28,980 actually, but it wasn't enough, economists always want you know give me 472 00:40:28,980 --> 00:40:31,799 the theoretically pure approach so this is it. 473 00:40:31,800 --> 00:40:40,040 So what we need to do is to use these flows on data within China, 474 00:40:40,040 --> 00:40:45,000 that is how much does a province export to China itself which we're able 475 00:40:45,000 --> 00:40:56,150 to get by combining Chinese records, Chinese input-output tables, with data on 476 00:40:56,150 --> 00:41:01,300 trade from provinces, so any province, Hunan province, 477 00:41:01,300 --> 00:41:06,340 to every country in the world as well as to the rest of China. So the rest of China 478 00:41:06,360 --> 00:41:11,400 becomes like another country for each province. Does that makes sense? This is a lot of data 479 00:41:11,400 --> 00:41:14,640 'cause in our gravity model we're gonna look at trade from 480 00:41:14,640 --> 00:41:19,619 country to country, so foreign to foreign, Chinese province to foreign and then 481 00:41:19,619 --> 00:41:24,960 Chinese province to the rest of China. And that's a lot of data and it required 482 00:41:24,960 --> 00:41:29,940 us to put together this variety of data sources to carry this out 483 00:41:29,940 --> 00:41:33,000 so it's gonna be, I'm going to show you a table and it's gonna be nice and neat 484 00:41:33,000 --> 00:41:37,740 but actually was pretty messy. So let's think back to hypothesis number one 485 00:41:37,740 --> 00:41:43,120 which was that after 1995 China reduced barriers to interprovincial trade. 486 00:41:43,120 --> 00:41:49,740 So-called interior border costs should fall, raising the value of interior proximity, 487 00:41:49,740 --> 00:41:53,160 that is being an interior city that's close to other interior cities, 488 00:41:53,160 --> 00:41:56,740 you should become more privileged after 489 00:41:56,740 --> 00:42:00,800 these barriers start to fall. In a sense these barriers really hindered the development 490 00:42:00,810 --> 00:42:07,230 of those cities. Now what that means is that when we estimate the trade costs of 491 00:42:07,230 --> 00:42:12,900 China to itself they should fall over time, So here's our estimates and I'm 492 00:42:12,900 --> 00:42:17,580 gonna ask you to do something. These are so-called fixed effects estimates 493 00:42:17,580 --> 00:42:21,440 so don't compare them across, they're each estimated separately in 494 00:42:21,440 --> 00:42:27,540 separate gravity models. I want you to just compare them, and I'll help you do this, 495 00:42:27,540 --> 00:42:33,080 to themselves. First is the log of distance. What is going to be that 496 00:42:33,090 --> 00:42:37,770 parameter that we're going to use to discount distance? If we just took GDP 497 00:42:37,770 --> 00:42:43,440 and divided it by distance that's as if we used an exponent of 1. It turns out 498 00:42:43,440 --> 00:42:49,200 that the right number is 1.5 and that doesn't change. That is if we look here, 499 00:42:49,200 --> 00:42:54,520 here, here, here, in fact there's no significant difference between any of these and within 1.5. 500 00:42:54,520 --> 00:42:58,440 It's kind of astonishing. I think I tried to tell you how we've created 501 00:42:58,440 --> 00:43:03,090 these gravity models with these reams of data completely different right, 502 00:43:03,090 --> 00:43:10,460 different years, 1997, 2002, and 2007, and these are the estimates - 1.5. 503 00:43:10,460 --> 00:43:14,940 So now I go to bed at night and I say I know what that parameter is it's 1.5. 504 00:43:17,040 --> 00:43:21,840 Now we're gonna focus on the area inside the red box. I'm going to compare the 505 00:43:21,840 --> 00:43:29,320 intra-China border, that is what is the barriers that I face 506 00:43:29,330 --> 00:43:34,520 trading from one Chinese city to another that's in a different province versus 507 00:43:34,520 --> 00:43:38,550 trading from my city to a foreign country. And when you see here these 508 00:43:38,550 --> 00:43:44,180 numbers are different but there's still this city - this is very large compared to that right. 509 00:43:44,180 --> 00:43:51,200 So - what 70% of this? That says that there were very high barriers 510 00:43:51,210 --> 00:43:59,430 within China from province to province trade. However by 2002 we see that this 511 00:43:59,430 --> 00:44:04,350 number isn't even significantly different than 0 so it's very small 512 00:44:04,350 --> 00:44:07,340 compared to this number which is almost -6, 513 00:44:07,340 --> 00:44:11,020 that says that China still faces barriers. And again I don't want you to 514 00:44:11,020 --> 00:44:14,650 compare from here to here and say that the barriers to China trading with the 515 00:44:14,650 --> 00:44:17,349 rest of the world becomes more because that's not true, these are relative to 516 00:44:17,349 --> 00:44:21,460 something else within each of these different columns. What we really wanna 517 00:44:21,460 --> 00:44:28,690 see is that the barriers within China itself have fallen. In 2002 we see the 518 00:44:28,690 --> 00:44:36,320 same result here where the barriers between one Chinese province and another 519 00:44:36,340 --> 00:44:40,600 Chinese province are basically insignificant. Now we're gonna 520 00:44:40,600 --> 00:44:44,620 make a different comparison, which is that we're going to compare the difference between 521 00:44:44,620 --> 00:44:50,640 China and its trade with other foreign countries and any other two 522 00:44:50,640 --> 00:44:55,500 foreign countries. So I have foreign country A and foreign country B, 523 00:44:55,500 --> 00:45:00,060 what's the barrier that they face from having to cross a border versus China 524 00:45:00,060 --> 00:45:05,900 and country B, what's the barrier that it has to cross. And what we see here is that 525 00:45:05,900 --> 00:45:13,000 in 1997 China's barriers were much bigger than the barriers that were faced by any 526 00:45:13,000 --> 00:45:18,640 other two countries. So China's ability say to crack in to the German market 527 00:45:18,640 --> 00:45:26,160 was much more hindered than the ability say of France to crack into that market. 528 00:45:27,740 --> 00:45:31,640 Now when we look in 2002, again just compare the two, 529 00:45:31,640 --> 00:45:40,380 China is still hindered more than any other two random countries in accessing a foreign 530 00:45:40,380 --> 00:45:46,740 market like Germany. By 2007, which is well after its accession to the WTO, 531 00:45:46,740 --> 00:45:50,859 these numbers are no longer statistically significantly different. 532 00:45:50,859 --> 00:45:58,869 In other words by 2007 the barriers between China and a foreign trading partner are 533 00:45:58,869 --> 00:46:03,880 no different than between any other two foreign, any other two countries in the world, 534 00:46:03,880 --> 00:46:09,220 by this estimate. So this has all of the exports basically in the world in 535 00:46:09,220 --> 00:46:13,029 these estimates, okay there's some little countries we missed but it's by and large 536 00:46:13,029 --> 00:46:18,980 all of the countries within the WTO which is the vast majority of trade in the world. 537 00:46:19,780 --> 00:46:27,600 So this shows you that, remember our hypothesis which was that the so-called interior border cost 538 00:46:27,600 --> 00:46:32,700 would fall, and the answer is yes they did, at least according to our gravity estimate 539 00:46:32,700 --> 00:46:39,740 so the trade that we observe we can back out these barriers and we see that they 540 00:46:39,750 --> 00:46:44,910 are consistent with reduced both intranational province to province 541 00:46:44,910 --> 00:46:53,490 trade and China to the rest of the world trade, and by 2007 we find no internal 542 00:46:53,490 --> 00:46:57,920 barriers and international frictions that look like those of all other countries. 543 00:46:57,920 --> 00:47:01,920 So you know our instinct and getting into the project was 544 00:47:01,920 --> 00:47:07,700 1995 estimate is great but things have really changed since then and just the gravity 545 00:47:07,710 --> 00:47:13,230 models alone tell us that instinct was right. Now I'm just gonna give you a 546 00:47:13,230 --> 00:47:19,290 first look at the data which we haven't done anything but create the log of 547 00:47:19,290 --> 00:47:27,760 market access. So we have three years here in 1995, 2002, and 2007. 548 00:47:27,760 --> 00:47:34,620 And along the horizontal axis I have the log of the market access measure that I've created 549 00:47:34,620 --> 00:47:38,760 for each city, remember market access is a characteristic of the workers' city, 550 00:47:38,760 --> 00:47:45,680 and here we put the average wage in those cities. And what we see is there's a consistently 551 00:47:45,680 --> 00:47:49,460 positive relationship between these two variables, 552 00:47:49,460 --> 00:47:54,020 that the average wage is higher in cities that have better market access. 553 00:47:54,020 --> 00:47:55,680 Now we wanna ask, 554 00:47:55,680 --> 00:48:00,460 well what happens when we control for the characteristics of the workers who are there? 555 00:48:00,460 --> 00:48:05,440 Maybe it is that just the coastal cities have workers with higher education levels 556 00:48:05,440 --> 00:48:10,700 or more experience and that's really why we're seeing higher wages in those locations. 557 00:48:12,940 --> 00:48:16,780 So what we're gonna do is run individual wage regressions, 558 00:48:16,780 --> 00:48:21,380 which are often called Mincer wage regressions after a famous dead economist - 559 00:48:21,380 --> 00:48:26,300 I hope he's dead this is on tape - and we're gonna estimate 560 00:48:26,300 --> 00:48:32,720 a separate wage regression for each survey year, 1995, 2002, and 2007. 561 00:48:32,720 --> 00:48:39,500 And we're gonna include the market access of the worker's location as a separate determinant 562 00:48:39,500 --> 00:48:44,160 of the wage. And then we're gonna throw in there every available individual 563 00:48:44,160 --> 00:48:47,780 characteristics we have from the survey and we're gonna include, 564 00:48:47,860 --> 00:48:51,420 we're gonna take account the fact that you're in a particular sector. 565 00:48:51,420 --> 00:48:52,960 And here we don't know what industry you're in 566 00:48:52,960 --> 00:48:55,380 we only know what sector you're in and there's three sectors - 567 00:48:55,380 --> 00:48:59,620 agricultural, basically manufacturing and mining, and then the service sector, 568 00:48:59,620 --> 00:49:02,500 so what we call primary secondary and tertiary sectors. 569 00:49:02,500 --> 00:49:09,719 We're also going to have even greater ability to kind of control for the average wage 570 00:49:09,720 --> 00:49:14,340 by looking at the sector wage the average wage for that sector, 571 00:49:14,340 --> 00:49:19,820 so say manufacturing in a region. So we're really going to take into account 572 00:49:19,820 --> 00:49:25,080 regional differences. We're also going to control for the ownership type of your firm. 573 00:49:25,080 --> 00:49:31,049 There's strong prior evidence of pay differences in particular that 574 00:49:31,049 --> 00:49:35,399 state-owned enterprises pay more and foreign invested enterprises pay more. 575 00:49:35,399 --> 00:49:39,329 What we're really interested in is not whether they pay more but whether 576 00:49:39,329 --> 00:49:44,500 they're more sensitive to market access, so differentially sensitive. 577 00:49:44,500 --> 00:49:49,739 Remember that Hering and Poncet found that it's really only the foreign invested 578 00:49:49,740 --> 00:49:53,800 enterprises particularly and some private enterprises who were sensitive 579 00:49:53,800 --> 00:49:57,160 to this market characteristic but not the state sector. 580 00:49:58,600 --> 00:50:07,880 So I go back to our hypothesis number two, that we know that after 1995 China really reduced 581 00:50:07,880 --> 00:50:14,760 trade investment barriers and the segregation of the export activities 582 00:50:14,760 --> 00:50:21,500 and that firms, we expect, should become more sensitive to locational differences 583 00:50:21,500 --> 00:50:25,740 that drive productivity because we expect that over time wages are gonna become 584 00:50:25,740 --> 00:50:30,980 more sensitive to productivity. So we expect the elasticity of the 585 00:50:30,980 --> 00:50:35,700 local wage with respect to market access to rise. So let's see what we're talking about here. 586 00:50:35,700 --> 00:50:40,600 So here are our regression results in full detail. I'm gonna suppress 587 00:50:40,600 --> 00:50:43,620 these individual characteristics in latter tables 588 00:50:43,620 --> 00:50:47,560 but I wanted you to see. Looking first at female, 589 00:50:47,560 --> 00:50:52,640 so this is whether the worker is a man or a woman, if you're a woman you can see that over time 590 00:50:52,640 --> 00:51:00,160 it's a negative - you're getting paid 8% less, 11% less, 22% less. That's not good right 591 00:51:00,160 --> 00:51:03,240 but that's what we know from other literature is being picked up. 592 00:51:03,240 --> 00:51:08,780 When we look at education we see, this is each additional year of education, 593 00:51:08,780 --> 00:51:14,980 we're seeing that that also, each additional year of education, is paying more in terms of the wage. 594 00:51:14,980 --> 00:51:19,700 So there's educational premium pretty much is stable between these two years. 595 00:51:19,700 --> 00:51:23,859 Age and age squared, we always find this sort of hump, your salary 596 00:51:23,860 --> 00:51:28,580 increases as you age and then it actually declines a little bit. 597 00:51:28,580 --> 00:51:35,200 So what about your ownership type, we see here that state-owned enterprises pay more, 598 00:51:35,200 --> 00:51:40,000 Sino-foreign enterprises, so equity joint ventures between a Chinese firm 599 00:51:40,000 --> 00:51:47,080 and a foreign firm, and Sole-foreign firms all pay more than private firms 600 00:51:47,080 --> 00:51:51,280 or other firms or which would be co-ops and other types of firms. 601 00:51:51,280 --> 00:51:59,280 So we still do see a wage premium here but the wage premium is diminishing 602 00:51:59,280 --> 00:52:05,950 particularly for the Sole-foreign firms. But the real star of the show is 603 00:52:05,950 --> 00:52:11,050 log market access. So what happens? This is the elasticity of the wage with 604 00:52:11,050 --> 00:52:16,510 respect to market access and it says if I increase your market access quotient, 605 00:52:16,510 --> 00:52:27,880 think of it as an index about your city by 1%, your wage is going to go up by 0.085%. 606 00:52:27,880 --> 00:52:31,320 So this is - oh I'm sorry, wrong one, this one, 607 00:52:31,320 --> 00:52:35,880 it's also an 8 right. And then that elasticity is increasing over time, 608 00:52:35,890 --> 00:52:42,340 in fact it more than doubles between 1995 and 2002. Interestingly enough after 609 00:52:42,340 --> 00:52:46,420 2002 it really doesn't change. Now what we can do is we can do this 610 00:52:46,420 --> 00:52:51,540 actually for 2013 now we have the data. Whether we find that it will go up or down, 611 00:52:51,540 --> 00:52:55,600 who knows, but we have to realize that a lot of changes in the labor 612 00:52:55,600 --> 00:53:02,660 market happened in the late 1990s and that that really drew up the sensitivity 613 00:53:02,660 --> 00:53:07,000 of wages to productivity differences derived from market conditions. 614 00:53:07,000 --> 00:53:13,160 So we expected that the elasticity of the wage with respect to market access would rise, 615 00:53:13,160 --> 00:53:15,620 we didn't know by how much or for how long, 616 00:53:15,620 --> 00:53:21,740 what we find is that it doubled between 1995 and 2002 and then it becomes stable. 617 00:53:21,740 --> 00:53:25,720 But what that means is that in answer to the question which is the title of the paper, 618 00:53:25,720 --> 00:53:30,460 has market access become more valuable for Chinese workers, the answer is yes. 619 00:53:30,460 --> 00:53:36,950 Actually you know when you think about it, it puts more pressure on the 620 00:53:36,950 --> 00:53:41,980 officials who you know the desire to control migration to the most, 621 00:53:41,980 --> 00:53:48,140 the highest wage cities right, the hukou system, puts more pressure on the hukou system 622 00:53:48,140 --> 00:53:52,760 because as these locations become more valuable more workers are gonna 623 00:53:52,760 --> 00:53:58,130 wanna go there so we might think that over time as China has 624 00:53:58,130 --> 00:54:03,000 integrated that wages are equalizing but in fact market forces, 625 00:54:03,000 --> 00:54:07,740 if we think in a different way, can actually act to make some locations even more valuable. 626 00:54:07,740 --> 00:54:12,529 Now the market is also going to have something to say about that in that 627 00:54:12,529 --> 00:54:16,760 local prices there are going to go up. Now what do I mean by that? 628 00:54:16,760 --> 00:54:19,840 We're just looking at nominal wages. We know that when you live in a city 629 00:54:19,840 --> 00:54:23,340 you also want to think about local prices, particularly the price of housing. 630 00:54:23,340 --> 00:54:26,100 So this says that we're gonna see a lot of pressure 631 00:54:26,100 --> 00:54:30,960 on housing prices and other local amenities to rise in those cities 632 00:54:30,960 --> 00:54:33,380 that have better locations, and of course that is also 633 00:54:33,380 --> 00:54:37,320 consistent with what we saw throughout the last 15 years. 634 00:54:37,320 --> 00:54:43,100 So hypothesis number three is that as China liberalized wage setting we should 635 00:54:43,100 --> 00:54:48,349 see that this increase in the elasticity of the wage with respect to market 636 00:54:48,349 --> 00:54:56,140 access would hold for all types of firms. And here we do an interaction. 637 00:54:56,140 --> 00:55:00,680 So here I said this is the elasticity of the wage with respect to 638 00:55:00,680 --> 00:55:04,730 market access on average and then we look and see if that average is 639 00:55:04,730 --> 00:55:09,940 different for different types of firms. And what we see is what Hering and Poncet 640 00:55:09,940 --> 00:55:13,160 picked up in 1995 is that it was bigger, 641 00:55:13,170 --> 00:55:18,960 it was more powerful for Sino-foreign, which is equity joint ventures mainly, 642 00:55:18,960 --> 00:55:24,880 and Sole-foreign firms. So foreign firms, which are free to set their own wages, 643 00:55:24,880 --> 00:55:29,680 were more sensitive to productivity differences. However when we look across here 644 00:55:29,680 --> 00:55:33,780 we have this very tiny result here and it's negative so it's not 645 00:55:33,780 --> 00:55:42,000 much basically - no. By 2002, 2007 this elasticity is the same across all 646 00:55:42,000 --> 00:55:49,500 firm types. So that's our result number three we find that for all 647 00:55:49,500 --> 00:55:53,610 all wages for workers at all types of firms display no significant 648 00:55:53,610 --> 00:55:57,120 differences in their relationship to the city market access. 649 00:55:57,120 --> 00:56:01,720 All elasticity is more than double and they're highly statistically significant. 650 00:56:02,840 --> 00:56:06,800 What about differences by skill level? If you think back Hering and Poncet found 651 00:56:06,810 --> 00:56:11,550 that these results only held for skilled workers. And we want to revisit that 652 00:56:11,550 --> 00:56:16,640 partly because it's a limited sample of skilled workers in the 1995 CHIP data, 653 00:56:16,640 --> 00:56:21,420 but also we know there's been a lot of migration particularly 654 00:56:21,420 --> 00:56:25,380 of unskilled workers from rural areas to the city so we wanna see what happens. 655 00:56:25,380 --> 00:56:29,609 So what we're gonna do is we're gonna divide our sample into two groups, 656 00:56:29,609 --> 00:56:34,109 those with high school diploma and above and those without a high school diploma, 657 00:56:34,109 --> 00:56:38,600 and we're going to re-estimate everything. And then we're going to look across here, 658 00:56:38,600 --> 00:56:43,840 here's our elasticity estimate for this group of so-called skill workers, 659 00:56:43,840 --> 00:56:47,200 those with a high school diploma or above, and we see that 660 00:56:47,200 --> 00:56:54,620 in 1995 the elasticity was 0.14, it goes to 0.26, 0.24, 661 00:56:54,620 --> 00:56:59,400 in other words yes it almost doubles for these so-called skilled workers. 662 00:56:59,400 --> 00:57:03,680 We expected that, Hering and Poncet found that it was statistically significant. 663 00:57:03,680 --> 00:57:11,060 Here we see pretty much the same pattern. It is smaller and less significant in 1995, 664 00:57:11,060 --> 00:57:18,840 as they found, but by 2002 and 2007 there's really very little difference between 665 00:57:18,840 --> 00:57:23,560 these two estimates. In other words the effect of 666 00:57:23,560 --> 00:57:27,900 a location on the productivity of a worker holds whether you're a skilled 667 00:57:27,900 --> 00:57:30,980 worker or an unskilled worker in our data. 668 00:57:33,260 --> 00:57:35,500 So what about alternative explanations? 669 00:57:35,500 --> 00:57:39,020 Maybe there's another way to explain these systemic relationships 670 00:57:39,020 --> 00:57:46,980 than the one that I'm trying to sell you. Well first is selection, so better workers 671 00:57:46,980 --> 00:57:51,240 move to better cities and this is not adequately captured by the 672 00:57:51,240 --> 00:57:56,580 individual characteristics like education. I mean maybe you have a college degree 673 00:57:56,580 --> 00:57:59,620 but is it from Cornell right. We want to say maybe 674 00:57:59,620 --> 00:58:05,300 you're just a better more skilled more able person. We call this omitted variable, 675 00:58:05,300 --> 00:58:10,000 omitted ability bias. The other is natural advantages. 676 00:58:10,020 --> 00:58:19,820 Some places just may be richer in certain types of you know natural port. 677 00:58:21,340 --> 00:58:24,500 Ningbo is supposed to have one of the most terrific natural ports 678 00:58:24,500 --> 00:58:27,580 in the entire world, maybe that's what drives wages up. 679 00:58:27,580 --> 00:58:32,960 So as you know we've already controlled for region in our regressions 680 00:58:33,000 --> 00:58:35,760 and that's the best that we can do to control for this. 681 00:58:35,760 --> 00:58:40,170 Human capital externalities, maybe you're in a place where a lot of smart people want to go 682 00:58:40,170 --> 00:58:44,610 and they make you smarter and your wages higher so we want to try to control for 683 00:58:44,610 --> 00:58:48,330 the number of smart people in the city. And then there's spillovers through the 684 00:58:48,330 --> 00:58:52,680 density of interactions. Maybe you're in a place where there's a lot of spillovers 685 00:58:52,680 --> 00:58:54,990 'cause there's a lot of other activity and we're gonna try to 686 00:58:54,990 --> 00:59:01,560 control for those. How do we do that? Well this is kind of fun. To control for 687 00:59:01,560 --> 00:59:08,130 selection by ability we found out that in 2007 the CHIP survey asked whether or 688 00:59:08,130 --> 00:59:12,180 not you had ever taken the Gaokao. So did you take the college entrance exam, 689 00:59:12,180 --> 00:59:19,500 and if you did what was your score. So the first thing we do is here's our estimate 690 00:59:19,500 --> 00:59:24,840 without controlling for this imperfect measure of ability. It's like the SAT in the US, 691 00:59:24,840 --> 00:59:28,800 people who get a high score always say it's a perfect measure of their ability, 692 00:59:28,800 --> 00:59:32,260 people who get a low score always say they're having a bad day right, 693 00:59:32,269 --> 00:59:36,769 we know that, I would do the same. But here it is without it. Now we're just 694 00:59:36,769 --> 00:59:41,239 gonna ask did you ever try, did you ever take the test, because there's 695 00:59:41,239 --> 00:59:45,319 self-selection into the test. People who think they have a shot at it would take 696 00:59:45,319 --> 00:59:50,209 the test. And we find out that wages of people who took the test are higher than 697 00:59:50,209 --> 00:59:54,349 people who didn't take the test. So maybe they did know something about themselves. 698 00:59:54,349 --> 00:59:58,080 Maybe it's not ability maybe it's just ability to obey in the workplace. 699 00:59:58,080 --> 01:00:02,419 I don't know what it is but it associated positively with wages. But we see it has 700 01:00:02,419 --> 01:00:06,799 almost no effect on the elasticity of the wage with respect to the locational 701 01:00:06,799 --> 01:00:12,529 characteristic we're interested in. Here we actually control for the score and as 702 01:00:12,529 --> 01:00:16,639 you see here the number of observations actually drops a lot because a lot of 703 01:00:16,640 --> 01:00:20,200 people said they took it but then they couldn't remember their score. 704 01:00:21,380 --> 01:00:26,080 So what happens when we do include the score for these people? Again we see that 705 01:00:26,089 --> 01:00:31,759 it has virtually no impact on the elasticity that we estimate. So we argue 706 01:00:31,759 --> 01:00:35,839 that it's not selection, it's not just smart people going to certain cities 707 01:00:35,840 --> 01:00:38,120 that are driving these results. 708 01:00:39,680 --> 01:00:42,440 The other is about city human capital. Are you just 709 01:00:42,440 --> 01:00:45,700 picking up cities that have a lot of smart people? And that turned out to be 710 01:00:45,709 --> 01:00:49,399 kind of hard to do because we have a lot of cities and what we're able to do so far 711 01:00:49,399 --> 01:00:54,409 is to look at the number of students that are enrolled in higher education so 712 01:00:54,409 --> 01:00:58,400 enrolled in, it says higher degree, or higher degree programs per capita. 713 01:00:58,400 --> 01:01:04,120 So cities that have a lot of colleges, many people stay after they get the degree, 714 01:01:04,120 --> 01:01:07,860 and so they tend to have a higher number of people who have college degrees. 715 01:01:07,860 --> 01:01:13,420 So we're going to use this indicator, how many college students per capita, 716 01:01:13,420 --> 01:01:17,839 as a measure of a smart city. How many people actually stay behind 717 01:01:17,839 --> 01:01:21,979 in Ithaca? If I go to dinner tonight is my waiter gonna have a PhD or 718 01:01:21,979 --> 01:01:25,639 what so I'm gonna have to ask him tonight. But anyway this is our measure 719 01:01:25,640 --> 01:01:32,720 and it's not significantly correlated with the wage and it has very little effect. 720 01:01:32,720 --> 01:01:35,839 This is a little smaller, you remember is like 0.20, but not 721 01:01:35,839 --> 01:01:40,699 much so it really doesn't affect us. So it's not just that you're getting a lot 722 01:01:40,699 --> 01:01:45,320 of human capital externalities from being in a place with a lot of smart people. 723 01:01:45,320 --> 01:01:48,240 Last we look at local density, and remember I told you 724 01:01:48,240 --> 01:01:50,960 that part of the theory is that a lot of firms are gonna 725 01:01:50,960 --> 01:01:55,420 wanna locate in this area to be close to these external markets, 726 01:01:55,420 --> 01:02:00,079 so it can't just be there's a lot of firms locating there. What we do is we look at 727 01:02:00,079 --> 01:02:06,020 just urban density, so-called urbanization externalities, going back to 728 01:02:06,020 --> 01:02:11,270 the work of the famous urbanologist Jane Jacobs, and here we look at the log 729 01:02:11,270 --> 01:02:14,940 of population density. So how dense is the interaction in the city? 730 01:02:14,940 --> 01:02:22,040 And it doesn't have an effect on the wage except a negative one for 2007, 731 01:02:22,040 --> 01:02:26,440 more densities actually have lower wages here, but if we look across here it does 732 01:02:26,450 --> 01:02:31,250 influence this a little bit, it kicks it up higher, but basically doesn't really 733 01:02:31,250 --> 01:02:36,079 do anything to our basic result which is that the elasticity of the wage with 734 01:02:36,079 --> 01:02:43,700 respect to market access more than doubles between 1995 and 2002 and then basically 735 01:02:43,700 --> 01:02:49,760 stays stable after that. So I've tried to convince you there's a statistically 736 01:02:49,760 --> 01:02:55,380 significant relationship in the data. But the next question is but does it really matter? 737 01:02:55,380 --> 01:03:00,020 How big an effect is it? With a big enough sample size we can detect 738 01:03:00,020 --> 01:03:03,840 very small effects we want to know if they're economically meaningful. 739 01:03:03,840 --> 01:03:12,500 So what we did is we created a worker named Joe. My father's name was Joe so. And my 740 01:03:12,500 --> 01:03:15,560 graduate student's name was Joe spelled a little differently, 741 01:03:15,560 --> 01:03:19,819 his father was named Joe too so we decided on Joe. And we gave Joe 742 01:03:19,820 --> 01:03:23,560 certain characteristics. Obviously Joe is a man and we gave him an education, 743 01:03:23,560 --> 01:03:28,160 the average education, the average age in our sample. And then we're gonna move Joe 744 01:03:28,160 --> 01:03:33,319 around to Chinese cities and ask what would we predict as Joe's wage. 745 01:03:33,320 --> 01:03:45,820 And we start, Joe is with the pandas in Chengdu. So we see his wage, in 2007 wage, RMB, 8 RMB. 746 01:03:45,820 --> 01:03:52,220 Now these are deflated okay so don't be too worried about Joe 747 01:03:52,220 --> 01:03:56,869 he's getting enough to eat he's good. So now we move him to Anhui province. 748 01:03:56,869 --> 01:04:03,440 His wage goes up by almost 15%. Now we're gonna move him to Hubei province. 749 01:04:03,440 --> 01:04:08,580 His wage goes up another 17%. Well this is from, 750 01:04:08,580 --> 01:04:12,640 I should say, that's how much more than living in Chengdu. So he has an advantage, 751 01:04:12,650 --> 01:04:17,329 an economic advantage, in living in Hubei as compared to Chengdu. What if we 752 01:04:17,329 --> 01:04:24,049 move him to Nanjing? Well he'll earn 33% more, 1/3 higher salary by living in 753 01:04:24,049 --> 01:04:31,489 Nanjing instead of Chengdu, almost 50% more by moving to Guangzhou, and then at 754 01:04:31,489 --> 01:04:35,239 last of course Shanghai. He almost doubles his wage by moving to Shanghai. 755 01:04:35,239 --> 01:04:43,720 So by 1997 these location - I'm sorry 1997 - 2007 these locational differences remain 756 01:04:43,720 --> 01:04:49,450 economically salient, they're very large. In fact you can almost double your wage, 757 01:04:49,450 --> 01:04:54,380 the same worker, from moving from Chengdu to Shanghai which says there's a lot of 758 01:04:54,380 --> 01:04:58,700 economic pressure for mobility that still exists within China. 759 01:05:01,360 --> 01:05:06,520 Does it matter to inequality? When we first started this we were interested in inequality. 760 01:05:06,520 --> 01:05:10,640 And what we do here is we just decompose the variance of wages so 761 01:05:10,640 --> 01:05:14,749 now we're not looking at average wages anymore or the wages for an individual 762 01:05:14,749 --> 01:05:19,249 like Joe we're basically looking at the variation or the differences in wages 763 01:05:19,249 --> 01:05:24,190 across individuals. And there's some part of that variance that's just due to 764 01:05:24,190 --> 01:05:28,339 differences within the city. Now we know that some of those differences in the 765 01:05:28,339 --> 01:05:31,249 wages within a given city are due the fact that maybe I have a college degree 766 01:05:31,249 --> 01:05:35,749 and you don't so we're gonna control for that and just look at 767 01:05:35,749 --> 01:05:40,369 residuals and compare it to the average. So if we take account of all 768 01:05:40,369 --> 01:05:45,440 those differences in our observable characteristics and see what's the 769 01:05:45,440 --> 01:05:50,240 variation within the city, that we're going to call that the within city variation. 770 01:05:50,240 --> 01:05:53,960 We don't know why our wages differ within the city but we know they do. 771 01:05:53,960 --> 01:05:59,200 We also know that that kind of variation has increased in the US as well. 772 01:05:59,200 --> 01:06:03,259 We see a lot of increase in variation in wages even within the same 773 01:06:03,260 --> 01:06:08,060 industry and in fact even within the same firm. This is so-called polarization of wages, 774 01:06:08,060 --> 01:06:11,490 even as we drill down we see that workers are getting increasingly 775 01:06:11,490 --> 01:06:14,790 different salaries even when they look the same on their basis of the 776 01:06:14,790 --> 01:06:19,590 characteristics. So here's what we call the within city variance and here's the 777 01:06:19,590 --> 01:06:23,130 differences across city. We're looking at the variation in the 778 01:06:23,130 --> 01:06:25,080 average wage across cities. 779 01:06:26,800 --> 01:06:33,640 So here's our table so we have the three years - 1995, 2002, and 2007 - 780 01:06:33,640 --> 01:06:37,080 and here's a measure of the total variance in the log wage - 781 01:06:37,080 --> 01:06:43,140 0.27, 0.31 basically, and 0.51. And now we're gonna parse it out between 782 01:06:43,140 --> 01:06:48,600 between city variants and within city. And what we can see here is that within 783 01:06:48,600 --> 01:06:54,300 city variance even in 1995 is the most important component. Here's the between 784 01:06:54,300 --> 01:06:59,460 city variance and what we can see is that that increases over time. 785 01:07:00,720 --> 01:07:10,200 So here we look at the change from 1995 to 2002, 2002 to 2007, 1995 to 2007. 786 01:07:10,200 --> 01:07:18,740 Variance rises quite a bit. We can see it almost doubles. So inequality gets worse. How much of 787 01:07:18,740 --> 01:07:22,520 that is just due to this differences between average wages across cities? 788 01:07:22,520 --> 01:07:27,860 The answer is about 1/5 of it but it has gotten bigger 789 01:07:27,860 --> 01:07:35,960 and so it has contributed to inequality. However the main driver is this differences in 790 01:07:35,960 --> 01:07:42,000 wages that are paid within a city to similar workers which we also pick up in 791 01:07:42,000 --> 01:07:48,660 Western labor markets. So it's called the residual which means we basically don't 792 01:07:48,660 --> 01:07:53,430 know what's driving that and the answer is like other market economies China 793 01:07:53,430 --> 01:07:59,400 also is exhibiting that unknown variation in wages within a city. 794 01:07:59,400 --> 01:08:05,340 So the answer is, is it driving inequality, yes it is a factor. It has been a contributor 795 01:08:05,340 --> 01:08:09,810 to increased inequality but the more important driver is this unexplained 796 01:08:09,810 --> 01:08:14,400 component of within city wage differences. 797 01:08:17,260 --> 01:08:23,260 The last thing I'm going to do is compare my results from China to results that have been 798 01:08:23,260 --> 01:08:28,060 estimated for Western with Western samples. 799 01:08:28,060 --> 01:08:32,980 So here is first the estimated elasticity of the wage with respect to market access. 800 01:08:32,980 --> 01:08:39,680 You remember that our estimate was about 0.2 by 2002. If we look at the estimates that 801 01:08:39,680 --> 01:08:46,400 Head and Meyer made for European regions in the 1990s, they did it industry by industry, 802 01:08:46,400 --> 01:08:51,440 and when I look at labor intensive industries I see that their 803 01:08:51,450 --> 01:08:54,359 estimate's almost the same as what we estimated for China. 804 01:08:54,360 --> 01:09:01,200 Gordon Hanson did the same thing for US counties and he estimated an elasticity of 0.20. 805 01:09:01,200 --> 01:09:05,660 Kind of scarily the same. Not anything I expected when I started it. 806 01:09:05,670 --> 01:09:09,810 But it does tell you that there're these - I think it does tell us that there's these 807 01:09:09,810 --> 01:09:15,020 empirical regularities that markets are driving these kinds of spatial wage inequalities. 808 01:09:15,020 --> 01:09:18,940 In the theory we can back out from the wage inequality, 809 01:09:18,940 --> 01:09:25,820 this implied sigma, which has to do with the elasticity of substitution across 810 01:09:25,830 --> 01:09:29,580 different types of goods and that's something that the the economists are 811 01:09:29,580 --> 01:09:34,820 very interested in, and of course with these numbers so similar the implied sigma, 812 01:09:34,820 --> 01:09:37,580 which I haven't really told you about and it's really just 813 01:09:37,580 --> 01:09:41,880 really deep into the theory, is almost the same from the three samples. 814 01:09:41,880 --> 01:09:46,319 So I think that's very interesting, but when Hering 815 01:09:46,319 --> 01:09:49,890 and Poncet did these estimates looked very different because their 816 01:09:49,890 --> 01:09:53,820 estimates you see were about half of what we found. 817 01:09:55,600 --> 01:10:00,480 When I presented this recently there was a man there who had recently been an advisor to 818 01:10:00,480 --> 01:10:06,560 President Obama and his first - can I conclude that China is a market economy. 819 01:10:06,560 --> 01:10:11,540 Now that turns out to be a very loaded question because under the WTO there are 820 01:10:11,550 --> 01:10:15,690 certain tests for whether you are a market economy and one of them is 821 01:10:15,690 --> 01:10:22,280 whether your wages are set freely - independently or by the market - so to speak. 822 01:10:22,280 --> 01:10:26,060 So I'm going to leave that as an open question. What I can say is that, 823 01:10:26,070 --> 01:10:30,420 because they have very specific legal tests, we can say is that the Chinese 824 01:10:30,420 --> 01:10:36,300 labor market increasingly looks like, or the traces of it that we see on trade 825 01:10:36,300 --> 01:10:41,480 relationships, look very much like their Western counterparts. 826 01:10:42,740 --> 01:10:48,140 To sum up, we have new gravity estimates that are consistent 827 01:10:48,140 --> 01:10:54,120 with reduced intranational trade barriers and international trade barriers. 828 01:10:54,120 --> 01:10:59,190 By 2007 we find no evidence of internal trade barriers and China's 829 01:10:59,190 --> 01:11:03,690 international frictions look like those of any other country. We also find the 830 01:11:03,690 --> 01:11:07,230 elasticity of the wage with respect to market access more than 831 01:11:07,230 --> 01:11:12,900 doubles between 1995 and 2002 and then it stabilizes. In this sense we can 832 01:11:12,900 --> 01:11:17,610 answer the question yes market access has become more valuable to Chinese 833 01:11:17,610 --> 01:11:23,280 workers over time and we find that now it's for all workers without regard to 834 01:11:23,280 --> 01:11:27,800 what type of firm they work for or whether they are skilled or unskilled workers. 835 01:11:27,800 --> 01:11:31,680 And lastly we find that these spatial wage premiums 836 01:11:31,680 --> 01:11:37,080 do contribute to rising inequality but they contribute much less to this 837 01:11:37,080 --> 01:11:42,719 unexplained variation within cities. I'm going to leave it there and 838 01:11:42,720 --> 01:11:45,220 hopefully you have some questions or comments for me.