City Size Distribution and Economic Growth: The Case of China

Size: px
Start display at page:

Download "City Size Distribution and Economic Growth: The Case of China"

Transcription

1 City Size Distribution and Economic Growth: The Case of China Ting JIANG, Ryo OKUI and Danyang XIE Department of Economics, School of Business and Management The Hong Kong University of Science and Technology December 27, 2008 Abstract This paper empirically explores the relationship between city size distribution and economic growth, based on a panel analysis using China provincial data from 1984 to We first estimate Zipf s coefficient, a measure of city size distribution, for each province in each sample year using Gabaix and Ibragimov s (2006) method. Second, we propose and verify a nonlinear relationship between Zipf s coefficient and GDP growth rate capturing the idea that government intervention on labor migration distorts city size distribution and entails productivity loss, so that a deviation of Zipf s coefficient from its equilibrium value has a negative impact on economic growth. Third, we construct a VAR model with which the co-evolution of city size distribution and economic growth is illustrated. Simulation results based on the estimated VAR model uncover the effect of an exogenous shock to city size distribution on economic growth and provide policy implications. Keywords: City size distribution; Economic growth; Zipf s law; Migration JEL Classifications: O18; R11; R12; R23 We have benefited from Kazuhiro Yamamoto and seminar participants at the Renmin University of China and the Hong Kong University of Science and Technology for helpful comments and suggestions. This research is fully supported by a grant from the Research Grants Council of the Hong Kong SAR (Project No. HKUST6470/06H). The usual disclaimer applies. 1

2 1 Introduction Urbanization is a ubiquitous phenomenon accompanying economic development because in cities most of industrial production and innovation take place. 1 This is especially true for developing countries when they experience transformation from agricultural to industrial economies. China is a vivid example. Since the economic reform and open-door policy was implemented in 1978, China has witnessed explosive economic growth for thirty years, with annual real GDP growth rate exceeding 8% in most of the years. At the same time, consistently rapid urban expansion has totally changed the geographical composition of the population. In 1978 when the reform began, only 17.92% of the total population lived in cities. In 2005, urban population represented 42.99% of the total population. This number is close to the world average (48.7%), although still much lower than that for developed countries (74.1%). 2 An important consequence of urbanization is a change in city size distribution. Here, the size of a city is measured by the number of people residing in that city. City size distribution is described by the number of cities and individual city sizes. In Figure 1, we compare the difference between city size distributions of China in 1984 and 2005 by dividing all the cities into six size categories. Two basic observations can be made: First, there is a large increase in the total number of cities (from 295 in 1984 to 656 in 2005, more than doubled), due to creation of new cities; Second, there is a large increase in the size of individual cities, which results in a boom of medium-sized cities and emergence of mega-cities. These results are striking, even when we take into account the fact that the total population of China increases by about 25% during this period. Despite that the coexistence of accelerated urbanization and economic growth seems to be self-explanatory, the evolution of city size distribution in the process of economic development is far from straightforward. A natural question is whether a faster growth leads to a more equal distribution of cities as we see more and more medium-sized cities are formed, or to a more unequal city size distribution as people might concentrate in large cities and smaller cities might decay and die? Yet it is more apposite to ask in the opposite way: Is more equal city size distribution conducive or harmful to economic growth? This paper aims to investigate the interaction between 1 The causal relationship between urbanization per se and economic growth is not the theme of this paper. See Gallup et al. (1999) which implies, and Bertinelli and Black (2004) which explicitly models such causality that urbanization promotes economic growth. 2 Source: World Urbanization Prospects: The 2005 Revision, available at the United Nations web-site, 2

3 city size distribution and economic growth in the context of China, and provide insights into the above questions via empirical analysis. We measure city size distribution by a single coefficient obtained from an auxiliary regression following the existing literature. For a group of cities, we first order them by size and record their ranks, and then regress the ranks on corresponding sizes using ordinary least squares (OLS) after taking natural logs on both sides. The estimated coefficient associated with the log of size (i.e., α 1 in the following equation) is a statistic that represents the size distribution of this group of cities: ln (Rank) = α 0 α 1 ln (Size) + Error. (1) A larger α 1 means city sizes are more evenly distributed, 3 while smaller α 1 means that there is a wider dispersion in city sizes. Remarkably, when researchers apply this simple exercise to various sets of real world data, they find in most cases α 1 is very close to unity (see Gabaix and Ioannides (2004) for a comprehensive survey). It has now become a well-accepted empirical regularity bearing the name Zipf s law. We call α 1 Zipf s coefficient henceforth throughout the text. Like many other studies, this paper acknowledges Zipf s law as being a good approximation to city size distribution in the real world. Moreover, many theoretical works predict that in a standard decentralized economy the city size distribution will evolve to its steady state with Zipf s coefficient approaching unity. We thus conjecture that city size distribution with Zipf s coefficient equal to one is actually the optimal distribution that would be observed in the absence of a friction or an obstacle to the functioning of the economy. A deviation of Zipf s coefficient from one could be viewed as a direct consequence of some distortion of individual decisionmaking (e.g., their migration and residence choices), which may be due to exogenous shocks or government intervention to the economic environment. We may reasonably assume that those distorted decisions are detrimental to economic growth. This hypothesis is the starting point of our attempt to build a relationship between city size distribution and economic growth. We argue that internal migration restriction in China is an important reason why Zipf s coefficient fails to be one in Chinese data, and Zipf s coefficient has a nonlinear effect on economic growth when it serves as a proxy of the enforcement of the migration restriction policy. This paper examines the effect of variation of Zipf s coefficient on aggregate economic performance, the speed of economic growth in particular. As far as we are aware, this is new in the literature. We estimate dynamic panel data models for the relationship between Zipf s coefficient and economic growth using Chinese data. 3 Imagine the extreme case when all the cities in an economy were of the same size, then if they were plotted in a rank-size diagram, we would have a vertical line and hence an infinitely large α 1. 3

4 Province is our definition of economy where a system of cities exists and evolves. By using provincial data, we are able to construct a rich panel dataset. The application of panel data analysis enables us to study the dynamics of city size distribution and economic growth and to avoid the bias caused by unobserved time-invariant province specific factors. We note that cross-sectional data do not allow us to address these issues. We estimate the models by GMM estimators for panel data. The results show that a deviation of Zipf s coefficient from its optimal value has a negative impact on economic growth. Besides, we also estimate a VAR (vector autoregressive) model which describes the co-evolution of city size distribution and economic growth. We then examine how the economy behaves in the short and long run when it faces exogenous shocks to initial Zipf s coefficient or growth rate. The rest of the paper is organized as follows: Section 2 summarizes the existing studies on the link between Zipf s law and economic development; Section 3 describes the dataset and discusses the choice of control variables and the theories behind them; Section 4 discusses how to construct a panel of Zipf s coefficients; Section 5 demonstrates the nonlinear effect of Zipf s coefficient on economic growth; Section 6 examines an economic theory in which economic growth affects Zipf s coefficient; In Section 7, we estimate a VAR model and carry out some simple simulations; Section 8 concludes the paper. 2 Zipf s Law and Economic Development The discovery of Zipf s law has inspired a considerable amount of subsequent research. Numerous researchers try to model an economy that generates a city size distribution which satisfies Zipf s law at its steady state. These studies not only justify our adoption of Zipf s coefficient as a proper measure of city size distribution, but also motivate our design of empirical strategies. Gabaix (1999) introduces exogenous idiosyncratic amenity shocks to explain the growth and decline of cities. Both wage and amenity enter workers utility function multiplicatively. A positive amenity shock leads to an inflow of workers but their utility increase is partly offset by wage decrease. The resulting population shock to a city is proportional to its current size, therefore in the steady state there must be a hierarchy of cities with different sizes. Rossi-Hansberg and Wright (2007) study a system of mono-centric cities each of which specializes in only one industry. Industryspecific exogenous technology shock affects city population through its effect on the average product of labor in the city, as residents seek to get highest earnings net of commuting costs associated with city size growth. Their theory also produces a city size distribution that is well approximated by Zipf s law. Duranton (2006) develops a 4

5 Romer-type endogenous growth model of product proliferation. Urban population is proportional to the number of varieties of industries (products) located in a city, and the growth of city size depends solely on the creation of new industries through R&D. The existence of local externality in R&D activities leads to a similar mechanism as in Gabaix (1999) in which large cities are more likely to expand, which in turn gives rise to Zipf s law. Another paper by Duranton (2007) examines the same issue using a Grossman- Helpman-type endogenous growth model of quality ladders. The channel through which economic development occurs in this model is not the enrichment of product varieties, but the improvement of product qualities. City sizes fluctuate as the locations of industries switch from one city to another. This is so far the only model in which the steady-state economic growth and the steady-state city size distribution are simultaneously solved. Using this model, we exploit an explicit link between economic growth and city size distribution. We empirically examine the validity of this theory in Section 6. Empirical studies related to our paper try to answer the question what factors drive the variation of Zipf s coefficient (i.e., treating Zipf s coefficient as the dependent variable in a regression). 4 Most of them are cross-sectional studies with country level data. Rosen and Resnick (1980) is an early comprehensive study on this topic. They estimate Zipf s coefficients for 44 countries in 1970 (or around that year), and find that variation in Zipf s coefficients is found to be positively associated with population level and GNP per capita, and negatively associated with land area and railway density. Alperovich (1993) reproduces Rosen and Resnick s results by employing a richer set of explanatory variables and also reaches some new findings. In particular, he finds that high degree of government involvement in the economy and high proportion of GDP produced by industrial sectors promote urban concentration (lower Zipf s coefficient). In a recent paper, Soo (2005) deals with a dataset of 75 countries with the latest record in He also uses OLS in the first stage to estimate Zipf s coefficient. In the second stage by OLS with panel corrected standard errors, he finds that political variables have more explanatory power of the variation than economic variables. All the four variables in Rosen and Resnick (1980) plus the size of non-agricultural sectors, the size of international trade, and the degree of scale economy either are insignificant or enter with opposite sign to what theoretical models would predict. Interestingly, he finds that the size of government expenditure is positively related to Zipf s coefficient, which contradicts Alperovich s (1993) 4 The validity of Zipf s law is also a serious concern in empirical works. Dobkins and Ioannides (2001), Ioannides and Overman (2003), and Black and Henderson (2003) are a few examples of recent attempt in this direction. 5

6 finding. Alternative approaches have been taken in closer examination of Zipf s coefficient. First, different techniques other than simple OLS could be applied. Alperovich (1992) carries out a pooled OLS analysis on panel data of Israel cities for various years between 1922 and He simultaneously estimates Zipf s coefficient and investigates their variation. Moomaw and Alwosabi (2004) is a panel analysis based on fixed effects models. The dataset consists of 30 countries from Asia and America for 7 time periods between 1960 and They use urban primacy 5 rather than Zipf s coefficient as the measure of city size distribution and find economic and political variables both to be important. Second, different empirical questions have been raised and tested. Alperovich (1992) finds out that Zipf s coefficient first decreases and then increases with per capita GNP when the country goes through different phases of development. Henderson (2003) investigates how city size distribution would affect economic growth. He demonstrates a nonlinear relationship between urban primacy and productivity growth across countries. This result implies that there might be an optimal degree of urban concentration, and over- or under-concentration may be very costly. Third, the study of developing countries is of independent interest, since they are experiencing more or less dramatic transition and restructuring of the socioeconomic system in which human activities are being organized and their city size distributions tend to exhibit larger fluctuation. The case of China is considered by Anderson and Ge (2005). They estimate Zipf s coefficients for the nation from 1961 to 1999 and check from a pure statistical perspective that there is a structural change of city size distribution due to economic reform and one child policy. 3 Data We make use of the data on 23 provinces in China from year 1984 to Four municipality cities (Beijing, Tianjin, Shanghai, and Chongqing) and four provinces (Hainan, Tibet, Qinghai, and Ningxia) which have few cities are excluded. The choice of time span is constrained by data availability. On the other hand, restricting the data in post-reform period (after 1978) may avoid the problem of structural change pointed out by Anderson and Ge (2005). There are two key variables. One is annual real GDP per capita growth rate which is our measure of the speed of economic growth. The other is Zipf s coefficient, which has to be estimated from city population data for each province in each year. Besides, we have a candidate set 5 The concept of urban primacy will be reviewed in Section 4. 6

7 of control variables which are discussed in Section 3.2. The population data come from China Population Statistics Yearbook and China Urban Statistics for various years. The data for GDP and control variables come from China Compendium of Statistics, and China Statistical Yearbook for various years. 3.1 Issues on population data In China, there are two kinds of cities, one is prefecture-level city (some of them are further defined as sub-provincial-level city), controlled by the province, the other is county-level city, most of which are controlled by the prefecture-level cities while some are controlled directly by the province. We refer the notion of city to both prefecture-level cities and county-level cities. Regarding city population, there are four categories. They are population in the whole region, population in town, nonagricultural population in the whole region, and non-agricultural population in town. Since the data for the last category is the most complete, we choose it as our measure of city size. The National Bureau of Statistics of China also uses this data to construct a size ranking of all cities in China every year. 6 We eliminate two tiny cities from the data. 7 We make these adjustments to remove sudden and sizable rise and fall in the evolution of Zipf s coefficient. We believe that city size distribution results from the aggregation of numerous uncoordinated individual actions, so that it should evolve only smoothly. The existence of these tiny cities can only be explained by administrative reasons rather than representing the effects of economic forces, therefore they are eliminated from the dataset. 8 6 Shen (2006) points out that urban population data for China may not be reliable and consistent. We are fully aware of this possibility. There are two concerns. One is the employment of rural labor in TVEs (township and village enterprises). The other is temporary population of cities mostly composed of rural migrants. The former is also related to another concern that economically meaningful definition of city agglomeration may be different from administratively defined cities. The latter casts more doubt on the credibility of existing data, since temporary population unquestionably contribute to city production and also place burden on city amenity and infrastructure. For some cities, the volume of economic activities by temporary population takes up a non-negligible share of the total number. Potential drawbacks notwithstanding, no systematic adjustment has been made in the present study simply because there is no proper and accepted way of doing so, especially with respect to city level data. 7 Wuda Lianchi is deleted from Heilongjiang from 1985 to It formed in 1985, remained at a low population level of less than 10 thousand before its population explosively increased to more than 110 thousand in Second, Wanting is deleted from Yunnan for all the sample years. It formed in 1985 and deceased in 1999, with its population never exceeding 6 thousand, far less than other cities in the same province. 8 Gabaix and Ioannides (2004) also mention the choice of cutoff size to remove tiny cities (see footnote 4 of their paper). This criterion is not applicable here because the sample size for a 7

8 3.2 Choice of control variables In previous empirical studies seeking for the determinants of variation in city size distributions, besides the variables indicating GDP level or growth, control variables often include population level, land area, size of industrial (or non-agricultural) sectors, size of external (or international) trade, road (or railway) density, and political variables such as size of government, bureaucratic efficiency, political freedom. In addition, growth accounting exercises typically use (physical and human) capital investment rates, inflation rate, black market exchange rate premium, and geographical dummies as independent variables. In the current context, some of the variables are impossible to obtain (such as measures of provincial bureaucratic efficiency or political freedom), some are invariant over time so that they are already captured by fixed effects in panel analysis (such as land area or the dummy indicating whether a province is along the coastal line), and some are unlikely to affect or be affected by city size distribution (such as capital investment rates, inflation rate, or black market exchange rate premium). The remaining candidates are population level, size of industrial sectors, size of external trade, road density, and size of government. It is helpful to first address very briefly two distinct features of cities before we continue to discuss the hypothesized effects of these variables on the direction of change in Zipf s coefficient. The benefit of urbanization comes from agglomeration economies. A static view is that in a city where labor is affluent and trade among firms is costless, production can be organized cheaply. The production network also enables each firm to specialize in the industry where it has comparative advantage, and economy-wide specialization leads to higher efficiency. A dynamic view is that when many firms and industries are gathered around, increased stock of knowledge will greatly facilitate R&D and make technological innovations much easier to take place. The cost of urbanization is due to congestion and hence decreased individual utility from amenities. Looking at the production side, employees will ask firms to compensate for the decrease of living standard, which results in higher production cost and less output. Cities of different sizes will typically face different scales of trade-off between these benefit and cost. When the population level is high, congestion cost may outweigh agglomeration benefit in large cities, so that people are motivated to move to smaller cities. When the size of industrial (non-agricultural) sectors is small and that of agricultural sectors is large, industrial cores have to be scattered to serve many agricultural peripheries, so that many cities are formed. The size of industrial sectors can also be viewed as an indicator of the degree of scale economy. The more prevalent modprovince is fairly small, so we have to keep as many data points as possible. 8

9 ern industries are, the more profitable to form large cities to extract agglomeration externalities. When an economy is involved in a greater extent of external trade (enjoying higher degree of openness), it is less pressing that the demand of its people for different varieties of goods has to be satisfied by internal trade, and that suppliers have to locate themselves near local consumers to sell more of their goods. Thus there is less demand-driven force to form large cities. The effect of road density on city formation is straightforward. More roads mean less transportation cost between the cores and the peripheries and more firms can be located together to form larger cities. More roads also mean less commuting cost, and make it much easier for people to work in mega-cities and live in suburb areas. The effect of the size of government on city distribution is ambiguous. There are two competing hypotheses. One emphasizes the distributive role of governments. The stronger the government is, the more able it is to reduce regional inequality and provide infrastructure and other public service to small cities to foster their growth. This point of view predicts a province with large government size to have equal city size distribution. The other mechanism relies on the fact that larger government size is usually associated with more political rents, and rent-seeking activities are more possible in large cities (usually provincial capital cities). This perspective would predict the opposite effect. Because of the fairly long time span of our dataset, we should avoid using any level variables that would exhibit certain trend over time. Another limitation is that there is no appropriate measure of the size of external trade for a province, which is the counterpart of international trade for a country. We do have the data on total imports and exports for each province, but the data on cross-province trade volume, probably more relevant in our case, are unavailable. Finally, the control variables (all of which are yearly data on the provincial level) we choose are: population natural growth rate (P opu); change in the size of industrial sectors (Ind) as a share of the provincial total GDP; 9 the size of government (Gov), measured by provincial government expenditure as a share of provincial GDP; growth rate of road mileage (Road); 10 and the size of international trade (T rade), measured by total imports and exports as a share of total GDP. Summary statistics for the five control variables are provided in Table 1. Their predicted effects on the direction of change of Zipf s coefficient are summarized in the following table. To reiterate, an increase in Zipf s 9 We also construct the relative size of industrial sectors, measured by first calculating the size of industrial sectors for each province (and the nation) and then taking the difference between the number for each province and that for the nation. But it does not affect the qualitative results and is not reported. 10 We also construct the growth rate of railway mileage. Since the data is incomplete and for many years there is no variation, it does a much worse fit in all regressions and is not reported. 9

10 coefficient means more equal size distribution, or more small and medium-sized cities are formed. On the other hand, a decrease in Zipf s coefficient means more unequal size distribution, or people are more concentrated in large cities. Control variable Predicted effect on Zipf s coefficient Population natural growth rate (P opu) + Change in size of industrial sectors (Ind) Size of government (Gov)? Growth rate of road mileage (Road) Size of international trade (T rade) + 4 Estimation of Zipf s coefficient Although the entire city size distribution may be drawn for any economy of interest, a representative yet operational measure that characterizes the property of the distribution should be made available before direct empirical testing. Several candidate measures have been proposed in the literature. One is spatial Gini index, constructed following exactly the same idea which has been extensively applied to measure income inequality. A perfectly equal distribution of city sizes implies a zero spatial Gini index and more unequal distribution corresponds to higher index. A similar analogy can be made to construct a spatial Hirschman-Herfindahl index, using the sum of the squares of the population share of all cities in an economy. The magnitude of the index is also positively associated with the degree of inequality of the city sizes distribution. A third measure is called urban primacy. It has a couple of variants. The most commonly used is the ratio of the size of the largest city to that of the whole economy. Others include the ratio of the size of the largest city to that of the second largest city, the ratio of the size of the largest city to the next three largest cities, and the ratio of the size of the two largest cities to the next two largest cities, etc.. Again, an economy with more unequal city size distribution is likely to have a larger urban primacy. However, the measure that attracts economists and geographers interest most is Zipf s coefficient. As was first discovered by Auerbach (1913) and later propagated by Zipf (1949), the size distribution of a group of cities is such that the number of cities whose sizes exceed a certain level S is proportional to the inverse of S. Concisely, we have the following probabilistic version of Zipf s law: P rob (Size > S) = α 0 S α 1 with α 1 1, (2) 10

11 and refer to α 1 as Zipf s coefficient. 11 If we take natural logs on both sides it becomes: ln (Rank) = α 0 α 1 ln (Size). (3) Statistically we should observe α 1 1 if Zipf s law is satisfied. Equation (1) is the empirical counterpart of this rank-size rule. However, Gabaix and Ibragimov (2006) show that the OLS estimator based on equation (1) or (3) is heavily biased and should be modified in the following way to correct the bias: 12 ( ln Rank 1 ) = α 0 α 1 ln (Size) + Error. (4) 2 Zipf s law implies that the null hypothesis α 1 = 1 should not be rejected. The interpretation on the magnitude of α 1 is that α 1 gets larger as the city size distribution becomes more even. We illustrate the idea in Figure 2, where we plot the samples of all cities in China in 1984 and 2005 respectively, and the fitted lines are depicted according to the regression equation (4). We can see that the slope of the fitted line for 2005 is steeper than that for 1984, suggesting the city size distribution of China in 2005 has become more equal as opposed to that in We calculate the four aforementioned measures for the entire period and the results are displayed in Figure 3. The evolution of the four measures over time generates similar patterns. For example, before year 2000, all the first three measures show a downward trend while Zipf s coefficient is going upward. The spatial Gini index curve has a U-shape, hitting the bottom in 1999, which matches the inverted- U-shape observed in the Zipf s coefficient curve. Actually, Zipf s coefficient has a strong pairwise correlation with the other three measures. 13 Although we cannot prefer one measure over another on pure theoretical ground as we do not know the underlying data-generating process, we shall expect that the results does not rely heavily on which measure to be used. Since Zipf s coefficient is generally favored in both theoretical and empirical literature, we choose it as the measure of city size distribution in carrying out our empirical study. Henderson (2005) asserts that spatial Gini index might be more reliable than Zipf s coefficient if we want to make any inference in a time-series framework, because 11 It is sometimes labeled Pareto exponent, because equation (2) is the defining property of Pareto distribution. Zipf s law is a special case of Pareto law with shape parameter α 1 = 1. (See Gabaix (2008) for an introduction on Pareto law.) 12 The formula for the standard error of α 1 is S.E. (α 1 ) = 2/nα The correlations between Zipf s coefficient and spatial Gini index, spatial HHI, and urban primacy are.904,.937, and.910, respectively. 11

12 the rank-size rule may not approximate the real-world distributions in all cases. His argument does not necessarily challenge the validity of our approach because we use a different estimation procedure for Zipf s coefficient from former empirical studies. 14 A panel of Zipf s coefficients for all 23 provinces during the years is estimated using the new rank-size rule (equation (4)). Summary statistics for the panel are shown in the first row of Table 1. Note that the mean value is larger than one, implying that provinces in China on average have a more even distribution of city sizes than in the ideal case where Zipf s law holds perfectly. Figure 4 further provides an overview of the evolution of Zipf s coefficient for each province from 1984 to A general conclusion is that provinces exhibit quite distinct time paths and no uniform pattern of evolution can be observed. For instance, cities in Fujian and Sichuan are consistently very evenly distributed (high Zipf s coefficient), while Hebei and Inner Mongolia have a wide dispersion of city sizes (low Zipf s coefficient) which also persists over time. Guangdong and Heilongjiang have once reached the highest and lowest values of Zipf s coefficient among the panel, respectively, although the deviation (from one) seems to be decreasing in recent years for both cases. There are also a number of provinces (e.g., Guizhou and Shanxi) whose city size distributions are nearly perfectly described by Zipf s law. For all the Zipf s coefficients we estimate, the null hypothesis α 1 = 1 is never rejected due to large standard errors. We now turn to formal examination of the interrelationship between city size distribution and economic growth, using the estimated panel as input for secondstage regressions. 5 Migration restriction, city size distribution and economic growth China has implemented strong migration restriction through household registration (hukou) system for many years. The system was set up in 1958, prescribing each individual a place of legal residence. Differential treatments are given to urban and rural residents, in favor of the former, regarding various civil rights pertaining to employment, housing, education, social security, and health care, etc., although the original intention of the system was only to find a way to ration citizens when supplies in necessities of life were scarce. Au and Henderson (2006a, 2006b) argue that migration restriction results in substantial loss in GDP. This argument can be justified in three aspects. First, it would impede large number of implicit rural unemployment 14 Henderson and Wang (2007) is an empirical work using spatial Gini index as the measure of city size distribution. 12

13 to be absorbed by urban industries, retard the formation and prosperity of labor market and market for human capital, and distort the allocative role of market economy for economic factors to achieve efficiency. Second, it reduces aggregate demand, since numerous illegal or temporary laborers are not expected to have growing consumption profiles before legitimate identity is obtained, basic needs for survival are met, and stable working condition is secured. The low life quality of a substantial number of de facto but not de jure urban population results in a vicious circle of low consumption and human capital investment, low productivity, and low earnings, which eventually jeopardizes the health of national economy. Third, migration restriction violates the idea of social justice based on equal opportunity and brings on urban-rural stratification, which would exacerbate social tensions and possibly may cause massive social conflicts that generate adverse economic effects. If people were allowed to move freely across cities, ceteris paribus, the economy would operate on the efficiency frontier. At the same time, the size distribution of cities will be shaped according to Zipf s law. This idea is drawn from the theoretical literature addressing the steady-state city size distribution in an endogenously growing economy, especially Rossi-Hansberg and Wright (2007) and Duranton (2006). In Rossi-Hansberg and Wright s (2007) formulation, land is owned by property developers who extract rents from their land. To achieve this goal, they offer a rich set of subsidies to attract firms and workers to migrate into the city. Developers maximize total rents net of subsidy costs, and by doing so they can fully internalize productivity externalities originated from industrial agglomeration. Therefore the equilibrium allocation is Pareto optimal, and government intervention which blocks free mobility across cities will inevitably reduce the efficiency and growth of the economy. Duranton (2006) does not make explicitly clear whether and how government policy can play any efficiency-enhancing role. Since the growth engine of the economy is R&D activities of competitive research firms aiming to invent new varieties of products based on existing general stock of knowledge, we might think that the government could boost the economy by coordinating independent research efforts in order to avoid duplication of experiments. However, it is hard to believe that the migration restriction mimics this sort of coordination policy. 15 We cannot measure the enforcement of migration restriction policy in different provinces, but we do know that the policy will have a direct effect on city size distribution. A large deviation of Zipf s coefficient from one could be viewed as a large 15 The only positive effect of migration restriction on such an economy we can barely think of is that by prohibiting high-skilled labors from changing their working places, firms have right incentive to invest in human capital accumulation. Not only is this possibility difficult to exploit in a general case, but in China, high-skilled labors actually find themselves easier to move. 13

14 distortion of government policy on the economic activities of individual citizens, and will entail great efficiency loss on the economy. Therefore we should observe a nonlinear relationship between GDP growth rate and Zipf s coefficient, in which Zipf s coefficient serves as a proxy for the intensity of the migration restriction policy s impact on economic performance. We model the nonlinear relationship between city size distribution and economic growth using the quadratic function of α: g it = β 1 g i,t 1 + β 2 α i,t 1 + β 3 α 2 i,t 1 + X itγ + D tδ + η i + u it, (5) where g is real GDP per capita growth rate, α is Zipf s coefficient, X it is a vector of control variables, D t is a vector of year dummies. η i denotes time-invariant provincespecific factors. Error term u it is independently and identically distributed. i is province index and t is time index. The marginal effect of α on g is: g it α i,t 1 = β 2 + 2β 3 α i,t 1. It varies with different values of α, and the sign of the effect and the direction of change depend on the signs of β 2 and β 3. Since Zipf s coefficient is obtained using end-of-year population data, we expect it to have potential influence on the next year s economic growth, so both α and α 2 enter the equation with lagged rather than contemporaneous terms. Fixed-effects estimation results of equation (5) are reported in columns (1) to (3) of Table 2. We can see that output growth is moderately persistent, which is captured by a significantly positive β 1. The parameter associated with α 1 is positive and the parameter associated with α 1 2 is negative, both of which are significant no matter whether control variables and year dummies are added or not. This result implies that the impact of an increase in lagged Zipf s coefficient on GDP growth is positive when it is small (too unequal size distribution), and the impact is negative when Zipf s coefficient is already large enough (too equal size distribution). One concern with fixed-effects estimator is that after within-group transformation, we have E [( g i,t 1 1 T 1 ) ( T 1 g i,j j=1 u i,t 1 T 1 )] T u i,j 0, so that the regressor and the error term after the transformation are correlated. 16 It is now well-known that the fixed-effects estimator has a downward bias of order 16 u it is correlated with g i,k 1 for k t + 1. j=2 14

15 O (1/T ), i.e., it will retain consistency only when T tends to infinity. In particular, Judson and Owen (1999) show that the bias is not negligible even when T = 20. To overcome this problem, Holtz-Eakin et al. (1988) and Arellano and Bond (1991) proposed a first-differenced GMM estimator. We first take the first difference of equation (5): g it g i,t 1 = ( ) β 1 β 2 β 3 γ δ g i,t 1 g i,t 2 α i,t 1 α i,t 2 α 2 i,t 1 α 2 i,t 2 X it X i,t 1 D t D t 1 + (u it u i,t 1 ). Given that u is serially uncorrelated, g i,t k (k 2) can serve as valid instruments for (g i,t 1 g i,t 2 ). We assume that α it (α 2 it) and X it are assumed to be endogenous, i.e., they are uncorrelated with future realizations of u but correlated with current and lagged values of u: E(α it u is ) = E(α 2 itu is ) = 0, E(X it u is ) = 0 for s > t, E(α it u is ) 0, E(α 2 itu is ) 0, E(X it u is ) 0 for s t. This assumption allows the simultaneity of g and α (α 2, X), as well as the potential feedback of α (α 2, X) on g. The difference-gmm estimator is the GMM estimator based on the following moment conditions: E [Φ i,t s (u it u i,t 1 )] = 0 for s 2; t = 3,... T ; i where Φ = { g, α, α 2, X }. (6) For the GMM estimation, we use the weighting matrix that is optimal in homoskedastic cases, but we report the cluster robust standard errors (Arellano (1987)). The results for difference-gmm estimation are shown in columns (7) and (8) of Table 2. Because T is large in our sample, we have to use a truncated set of instruments to ensure the number of instruments does not exceed the sample size. In column (7), we report the result using a maximum of two lagged variables as instruments. In column (8), we also report the result using only one lag of all pertinent variables as instruments, for the purpose of robustness check. The parameters associated with α 1 and α 2 1 remain significant and the estimate of growth rate persistence is adjusted slightly upward, as expected. The Arellano-Bond tests at the last row indicates that there is no evidence of serial correlation in the error term (thus, there is no evidence against instrument validity). We also consider a restricted specification derived from our hypothesis g it = β 1 g i,t 1 + β 2 (1 α i,t 1 ) 2 + X itγ + D tδ + η i + u it, (7) 15

16 where β 2 is expected to be negative. This specification is based on the idea that it is the deviation of Zipf s coefficient from unity that is associated with distortion to the economy. Fixed-effects estimation results of equation (7) are reported in columns (4) to (6) of Table 2, and difference-gmm estimation is similarly implemented and reported in columns (9) and (10). Note that the instrument set, Φ, is now { g, (1 α) 2, X }. The coefficient β 2 appears significantly negative in all the specifications. In particular, our preferred estimates, given by GMM, are.115 and.142. However, the effects of control variables vary upon the inclusion of year dummies and different choices of instrument sets. We now state our main result formally: Other things being equal, any deviation from optimal city size distribution will result in lower economic growth. Specifically, when Zipf s coefficient deviates from one by x, real GDP per capita growth rate will decrease by about 11.5x 2 % to 14.2x 2 %. 6 City size distribution in an innovation-driven economy The previous section identifies the influence that city size distribution exerts on economic growth. However, economic development may also affect the evolution of city size distribution. As mentioned in the Introduction, Duranton (2007) provides a plausible argument. We test the theory of Duranton (2007) using Chinese data. In Duranton (2007), city size is determined by the number of industries located in that city. An industry means a line of same goods with different qualities. In equilibrium only the good with highest quality is produced and each industry is located in only one city. Research is undertaken to improve the quality of current goods, but it would also accomplish unexpected cross-industry innovations with some probability. When research in city A leads to an innovation of some other industry located in city B, the production of that industry will be transferred to city A. Therefore the size of city A increases by one unit while the size of city B decreases by the same unit. A steady state requires that the expected number of cities of each size remains constant. Here, output growth has no direct effect on city size distribution, but it serves as a legitimate proxy of the likelihood of technological innovation. The latter is the driving force for the city size distribution to converge to its steady state. Consider an increase in the probability that a cross-industry technological innovation takes place. It will increase output growth rate for sure, and will also speed up the convergence of city size distribution to the steady state Strictly speaking, Duranton s model does not produce a steady-state city size distribution with 16

17 If we initially have a Zipf s coefficient larger than one, then it should drop more drastically when output growth rate is higher. If we instead have an initial Zipf s coefficient smaller than one, then it should have a larger increase along with higher output growth rate. We start from the following first-order autoregressive structure: α it = b 0 + b 1 α i,t 1. (8) The speed of convergence b 1 depends on the probability of cross-industry technological innovation, for which we use output growth rate as a proxy. The larger b 1 is, the slower it converges to its steady state. We assume a linear structure upon this relationship: b 1 = c 0 + c 1 g i,t, where c 1 < 0. Moreover, Zipf s law predicts the limit of α it to be one, that is Rewriting (8), we get lim α it = b 0 = 1. t 1 b 1 α it = 1 b 1 + b 1 α i,t 1 = 1 c 0 + c 0 α i,t 1 c 1 (1 α i,t 1 ) g i,t. Therefore, an econometric specification we propose is: α it = β 1 α i,t 1 + β 2 (1 α i,t 1 ) g i,t + X itγ + D tδ + η i + u it, where all the terms are defined in the same way as in equation (5) and (7). We are supposed to observe β 2 > 0 in light of Duranton s model. However, the fixed-effects estimates in columns (1) to (3) of Table 3 reveal that the theory is at odds with our data. The coefficient β 2 is found to be insignificantly negative in all the specifications. 18 Of course, this result may be caused by the bias of the fixed effects estimator. As in the previous subsection, we employ GMM estimators to overcome this problem. However, Arellano and Bover (1995) and Blundell and Bond (1998) argue that when the individual series are highly persistent, instruments used in the standard first-differenced estimator are likely to be weak and may underlying Zipf s coefficient equal to one, rather, it produces a complete distribution profile of city sizes, which could be well approximated by Zipf s law. 18 It is possible that GDP growth affects the convergence of city size distribution gradually, but replacing GDP growth rate in the interaction term by its lagged value does not change the result. 17

18 induce serious bias in finite samples. This argument is relevant in the current context as we notice that the estimated β 1 is around 0.8. To cure this problem, in addition to the instruments described in (6), they propose a new set of moment conditions in levels where lagged difference of endogenous (or predetermined) explanatory variables serve as corresponding instruments: E [(α i,t 1 α i,t 2 ) (η i + u it )] = 0 for t = 3,... T ; i, E [((1 α i,t 2 ) g i,t 1 (1 α i,t 3 ) g i,t 2 ) (η i + u it )] = 0 for t = 4,... T ; i. Following the existing empirical studies, we assume all control variables to be strictly exogenous, so they do not serve as GMM-type instruments. 19 The results of this system-gmm estimation are shown in columns (6) and (7) of Table 3, where the maximum numbers of lagged variables used as instruments for differenced equations are again fixed at two and one, respectively. We find that β 2 is still negative. Of course it would be hasty to say that the direction of causality goes solely from city size distribution to economic growth, not the other way round, based on the above result. What may be concluded is that even if output growth does influence city size distribution, the effect is not likely to take place through the channel described in Duranton (2007). This result casts some doubt on the applicability of Duranton s story to China and is a potential topic for future research. 7 A vector autoregressive framework In this section, we illustrate how well our model fits the data and then examine how city size distribution and economic growth evolve in face of exogenous shocks. We first set up a VAR structure in the form of simultaneous equations system: { git = β 11 g i,t 1 + β 12 (1 α i,t 1 ) 2 + X itγ 1 + D tδ 1 + η g,i + u g,it α it = β 21 α i,t 1 + β 22 g i,t 1 + X itγ 2 + D tδ 2 + η α,i + u α,it (9) The first equation is identical to equation (7), which we have discussed in Section 5. As Section 6 shows, the theory we borrow from Duranton (2007) predicting the effect of GDP growth on city size distribution is not supported by the data, so we simply assume a reduced-form linear relationship of g on α, as is presented in the second equation of the above VAR system. 19 If X it is indeed endogenous with city size distribution, OLS and fixed-effects panel estimation will no longer be valid since they both lead to inconsistent estimates. However, in our GMM case results are not affected qualitatively if X it are treated as endogenous and thus not reported. 18

19 Fixed-effects estimation results of the second equation in equations (9) are shown in columns (4) and (5) of Table 3. The corresponding results of System-GMM estimation are reported in columns (8) and (9) of the same table. Note that in a VAR system such as equations (9), control variables should be treated in the same way across equations. Because the control variables are clearly endogenous in the first equation, we assume they are also endogenous in the second one. 20 The estimated effect of GDP growth rate is insignificant in all specifications. The size of government is only significant when assuming no time-specific effect, indicating that the rent-seeking effect of having large government more than offsets the distribution effect. The growth of industrial sectors is significant except when we use a smaller set of instruments in the system-gmm estimation. However, the sign of the effect is opposite to what we have argued in Section 3.2. These results show potential limitation of the existing cross-sectional studies and panel studies using pooled OLS estimates. The results obtained in those studies explaining the variation of city size distribution may simply be due to their failure of taking account of time-specific factors and explicitly modeling the dynamic nature of city size distribution. Next we use our estimated model to interpolate GDP growth and city size distribution during the sample period Since significance of estimates is not a concern here, the parameter values of the above system are taken from column (9) of Table 2 and column (8) of Table 3. The fitness of the VAR system can be illustrated by comparing the real data and the fitted values of g and α. In Figure 5 we plot two examples. Province Guangdong has the largest variation of Zipf s coefficient during the sample period, and Shanxi has the largest variation of real GDP per capita growth rate. For the fitted lines, g 1984 and α 1984 from real data are set as initial values, then g and α for subsequent years are computed in succession according to equations (9). Because of the endogeneity of control variables, we cannot simply insert the real data year by year. Instead, we substitute the sample mean into the equations and keep them constant over the years. We can see from Figure 5 that the model fits the path of GDP growth fairly well, and generates less variation than the real data of Zipf s coefficient, but still matches the trend. The model can also be used for simulation exercises to predict the pattern of co-evolution of economic growth and city size distribution when an exogenous shock is introduced. The long run values of real GDP per capita growth rate and Zipf s coefficient predicted by this model are g = 9.61% and α = The finding α > 1 implies that migration restriction will inhibit the growth of mega-cities in the long run. This finding is in line with Henderson s (2005) observation for developing 20 Again, in column (8), the first two lags of all possible instruments are used for differenced equations, and in column (9) instruments are restricted to include only the first lag. 19

20 countries. Simulation results are shown in Figure 6. Suppose initially the provincial GDP growth and city size distribution are at their respective steady states. The magnitude of shocks are assumed to be one standard deviation of Zipf s coefficient and real GDP per capita growth rate in the whole sample. If the economy is hit by a positive growth shock, the Zipf s coefficient will jump up in the first three years and then decrease to the steady state level. During the transition period, we will observe more equal city size distribution. An interesting observation is that if the economy experiences some sort of negative distribution shock (driving down Zipf s coefficient) that leads to a more unequal city size distribution (we call it polarization shock ), the economy will grow faster in the following years than if there were no such shock. On the contrary, a positive distribution shock (driving up Zipf s coefficient) which has an impact on the equalization of city sizes, will result in a slowdown of GDP growth, and the decrease is disproportionately high as compared to the increase associated with a negative shock of the same magnitude. Imagine that the source of such distribution shock could be government policies. Our simplistic model would generate strong and unambiguous policy implications. One example is the policy that promotes increased enrollment of higher education, which has been intensely debated during recent years in China. Since most universities in China are located in large cities, this policy induces many students to flow into large cities. Most of them will get employed in the cities after graduation. They will be entitled citizenship (hukou) and become permanent residents of the cities. Now that more people are concentrated in large cities than before, the city size distribution will be more unequal, which, according to our model, is conducive to future economic growth. An opposite example is available in the history of Cultural Revolution, during which China has witnessed a swelling tide of educated urban youth going and working in the countryside and mountain areas. This migration choice was, to a large extent, also forced by government policy or Chairman Mao s personal will. The analysis is a bit more complicated than in the previous example, since the outflow of young population happened not only in large cities, but also in smaller cities. Moreover, this forced migration policy aimed not only at college students, but high school students as well. If large cities and small cities have roughly the same proportion of educated youth to total population, reducing the sizes of all cities proportionally will not change the size distribution. However, part of the educated youth in small cities who would have moved to large cities to pursue higher education were actually driven to countryside. Therefore large cities lost more population than smaller cities, and as a consequence, city size distribution became more equal. Again, our model implies 20

21 that the policy during the Cultural Revolution was harmful to economic growth. 8 Conclusion In urban economics literature, a vast quantity of research has emerged in answering how and why city size distributions differ across regions and evolve over time, but few study has been conducted to investigate the effect of city size distribution on economic growth. This paper is an attempt in this direction. Rather than directly modeling the mechanism through which city size distribution would possibly influence the level or pace of economic growth, this paper takes a different strategy. Our results are based on two crucial assumptions. First, in acknowledging that Zipf s law is a widely accepted empirical regularity, especially in developed countries where there is no contrived restriction distorting free internal labor migration, we assume that the optimal city size distribution in the absence of any exogenous intervention is such that Zipf s coefficient of the distribution equals to one. The optimal city size distribution can be viewed as a necessary demographic condition to ensure an optimal growth path. Second, internal migration restriction across cities in China introduces efficiency losses in aggregate production, as well as distorts the convergence of city size distribution. We then argue that there is an inverted-u-shape relationship between Zipf s coefficient and economic growth rate, peaked at the point where Zipf s coefficient equals to one. The argument is not rejected in the context of China. A VAR model characterizing the co-evolution of city size distribution and economic growth is further constructed and estimated. The simulation results based on the model reveal that, given migration restriction as a binding institutional constraint in the long run, any exogenous distribution shock that brings the cities toward a more uneven distribution will be accompanied by faster economic growth. This result has a clear policy implication. References [1] Alperovich, G. (1992). Economic Development and Population Concentration. Economic Development and Cultural Change 41 (1): [2] Alperovich, G. (1993). An Explanatory Model of City-Size Distribution: Evidence From Cross-Country Data. Urban Studies 30 (9): [3] Anderson, G. and Y. Ge (2005). The Size Distribution of Chinese Cities. Regional Science and Urban Economics 35 (6):

22 [4] Arellano, M. (1987). Computing Robust Standard Errors for Within-groups Estimators. Oxford Bulletin of Economics and Statistics 49 (4): [5] Arellano, M. and S. Bond (1991). Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58 (2): [6] Arellano, M. and O. Bover (1995). Another Look at the Instrumental Variable Estimation of Error-Components Models. Journal of Econometrics 68 (1): [7] Au, C. C. and J. V. Henderson (2006a). Are Chinese Cities Too Small? Review of Economic Studies 73 (3): [8] Au, C. C. and J. V. Henderson (2006b). How Migration Restrictions Limit Agglomeration and Productivity in China. Journal of Development Economics 80 (2): [9] Auerbach, F. (1913). Das Gesetz der Bevölkerungskonzentration. Petermanns Geographische Mitteilungen 59: [10] Black, D. and J. V. Henderson (2003). Urban evolution in the USA. Journal of Economic Geography 3 (4): [11] Blundell, R. and S. Bond (1998). Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics 87 (1): [12] Bertinelli, L. and D. Black (2004). Urbanization and Growth. Journal of Urban Economics 56 (1): [13] Dobkins, L. H. and Y. M. Ioannides (2001). Spatial Interactions among U.S. Cities: Regional Science and Urban Economics 31 (6): [14] Duranton, G. (2006). Some Foundations for Zipf s Law: Product Proliferation and Local Spillovers. Regional Science and Urban Economics 36 (4): [15] Duranton, G. (2007). Urban Evolutions: The Fast, the Slow, and the Still. American Economic Review 97 (1): [16] Gabaix, X. (1999). Zipf s Law for Cities: An Explanation. Quarterly Journal of Economics 114 (3):

23 [17] Gabaix, X. (2008). Power Laws, entry in The New Palgrave Dictionary of Economics, 2nd Edition, Steven N. Durlauf and Lawrence E. Blume (eds.), MacMillan. [18] Gabaix, X. and R. Ibragimov (2006). Rank -1/2: A Simple Way to Improve the OLS Estimation of Tail Exponents. Working paper. [19] Gabaix, X. and Y. M. Ioannides (2004). The Evolution of City Size Distributions. Handbook of Regional and Urban Economics Volume 4: [20] Gallup, J. L., Sachs, J. D. and A. D. Mellinger (1999). Geography and Economic Development. International Regional Science Review 22 (2): [21] Henderson, J. V. (2003). The Urbanization Process and Economic Growth: The So-What Question. Journal of Economic Growth 8 (1): [22] Henderson, J. V. (2005). Urbanization and Growth. Handbook of Economic Growth Volume 1: [23] Henderson, J. V. and H. G.Wang (2007). Urbanization and City Growth: The Role of Institutions. Regional Science and Urban Economics 37 (3): [24] Holtz-Eakin, D., W. Newey and H. S. Rosen (1988). Estimating Vector Autoregressions with Panel Data. Econometrica 56 (6): [25] Ioannides, Y. M. and H. G. Overman (2003). Zipf s Law for Cities: An Empirical Examination. Regional Science and Urban Economics 33 (1): [26] Judson, R. A. and A. L. Owen (1999). Estimating Dynamic Panel Data Models: A Guide for Macroeconomists. Economics Letters 65 (1): [27] Moomaw, R. L. and M. A. Alwosabi (2004). An Empirical Analysis of Competing Explanations of Urban Primacy Evidence From Asia and the Americas. Annals of Regional Science 38 (1): [28] Rosen, K. T. and M. Resnick (1980). The Size Distribution of Cities: An Examination of the Pareto Law and Primacy. Journal of Urban Economics 8 (2): [29] Rossi-Hansberg, E. S. and M. A. Wright (2007). Urban Structure and Growth. Review of Economic Studies 74 (2):

24 [30] Shen, J. (2006). Estimating Urbanization Levels in Chinese Provinces in International Statistical Review 74 (1): [31] Soo, K. T. (2005). Zipf s Law for Cities: A Cross-Country Investigation. Regional Science and Urban Economics 35 (3): [32] Zipf, G. L. (1949). Human Behavior and the Principle of Least Effort: An Introduction to Human Ecology. MA, Addison-Wesley. 24

25 Figure 1: The composition of cities in China in 1984 and 2005 by size. Note: City size is measured by non-agricultural population in town. 25

26 Figure 2: Zipf s law in China in 1984 and

27 Figure 3: Comparison of different measures of city size distribution. Notes: (a) Spatial Gini index is computed by first drawing a Lorenz curve of the distribution of city sizes and then taking the size ratio of the area enclosed by the Lorenz curve and the line of perfect equality to the lower-right triangle. (b) Spatial HHI (spatial Hirschman-Herfindahl index) = n i=1 (s i/ S 100) 2, where s i is the population of city i, S is the total population, n is the number of cities. (c) Urban primacy is computed as the ratio of the population of the largest city to the total population. (d) Zipf s coefficient is equal to the absolute value of the slope of the fitted line as in Figure 2. 27

28 Figure 4: Evolution of Zipf s coefficient from 1984 to 2005 by province. 28

29 Figure 5: Fitness of the model: two examples. 29

30 Figure 6: Simulation of city size distribution and economic growth. Notes: The figure above illustrates the evolution of Zipf s coefficient, deviating from the steady state, after a positive growth shock and a negative growth shock, respectively. The figure below illustrates the evolution of real GDP per capita growth rate, deviating from the steady state, after a positive (equalization) shock and a negative (polarization) shock on city size distribution, respectively. 30

More on Roy Model of Self-Selection

More on Roy Model of Self-Selection V. J. Hotz Rev. May 26, 2007 More on Roy Model of Self-Selection Results drawn on Heckman and Sedlacek JPE, 1985 and Heckman and Honoré, Econometrica, 1986. Two-sector model in which: Agents are income

More information

A Numerical Simulation Analysis of (Hukou) Labour Mobility Restrictions in China

A Numerical Simulation Analysis of (Hukou) Labour Mobility Restrictions in China A Numerical Simulation Analysis of (Hukou) Labour Mobility Restrictions in China John Whalley Department of Economics, The University of Western Ontario and Shunming Zhang Department of Finance, School

More information

Route of Urbanisation in China from an International Perspective

Route of Urbanisation in China from an International Perspective 18 The Route of Urbanisation in China from an International Perspective Xiaolu Wang 1 Introduction This chapter examines China s urban development strategy from an international perspective. There are

More information

Analysis for Regional Differences and Influence Factor of Rural Income in China

Analysis for Regional Differences and Influence Factor of Rural Income in China Modern Economy, 2012, 3, 578-583 http://dx.doi.org/10.4236/me.2012.35076 Published Online September 2012 (http://www.scirp.org/journal/me) Analysis for Regional Differences and Influence Factor of Rural

More information

Secondary Towns and Poverty Reduction: Refocusing the Urbanization Agenda

Secondary Towns and Poverty Reduction: Refocusing the Urbanization Agenda Secondary Towns and Poverty Reduction: Refocusing the Urbanization Agenda Luc Christiaensen and Ravi Kanbur World Bank Cornell Conference Washington, DC 18 19May, 2016 losure Authorized Public Disclosure

More information

The challenge of globalization for Finland and its regions: The new economic geography perspective

The challenge of globalization for Finland and its regions: The new economic geography perspective The challenge of globalization for Finland and its regions: The new economic geography perspective Prepared within the framework of study Finland in the Global Economy, Prime Minister s Office, Helsinki

More information

Applied Microeconometrics (L5): Panel Data-Basics

Applied Microeconometrics (L5): Panel Data-Basics Applied Microeconometrics (L5): Panel Data-Basics Nicholas Giannakopoulos University of Patras Department of Economics ngias@upatras.gr November 10, 2015 Nicholas Giannakopoulos (UPatras) MSc Applied Economics

More information

A Meta-Analysis of the Urban Wage Premium

A Meta-Analysis of the Urban Wage Premium A Meta-Analysis of the Urban Wage Premium Ayoung Kim Dept. of Agricultural Economics, Purdue University kim1426@purdue.edu November 21, 2014 SHaPE seminar 2014 November 21, 2014 1 / 16 Urban Wage Premium

More information

Subject: Note on spatial issues in Urban South Africa From: Alain Bertaud Date: Oct 7, A. Spatial issues

Subject: Note on spatial issues in Urban South Africa From: Alain Bertaud Date: Oct 7, A. Spatial issues Page 1 of 6 Subject: Note on spatial issues in Urban South Africa From: Alain Bertaud Date: Oct 7, 2009 A. Spatial issues 1. Spatial issues and the South African economy Spatial concentration of economic

More information

National Spatial Development Perspective (NSDP) Policy Coordination and Advisory Service

National Spatial Development Perspective (NSDP) Policy Coordination and Advisory Service National Spatial Development Perspective (NSDP) Policy Coordination and Advisory Service 1 BACKGROUND The advances made in the First Decade by far supersede the weaknesses. Yet, if all indicators were

More information

Econometrics of Panel Data

Econometrics of Panel Data Econometrics of Panel Data Jakub Mućk Meeting # 6 Jakub Mućk Econometrics of Panel Data Meeting # 6 1 / 36 Outline 1 The First-Difference (FD) estimator 2 Dynamic panel data models 3 The Anderson and Hsiao

More information

Urban Economics City Size

Urban Economics City Size Urban Economics City Size Utility and City Size Question: Why do cities differ in size and scope? While NYC has a population of more 18 million, the smallest urban area in the U.S. has only 13,000. A well

More information

Do bigger cities contribute to economic growth in surrounding areas? Evidence from county-level data in China

Do bigger cities contribute to economic growth in surrounding areas? Evidence from county-level data in China Do bigger cities contribute to economic growth in surrounding areas? Evidence from county-level data in China Yan Liu School of Economics, Fudan University Xingfeng Wang China Academy of Urban Planning

More information

Does agglomeration explain regional income inequalities?

Does agglomeration explain regional income inequalities? Does agglomeration explain regional income inequalities? Karen Helene Midelfart Norwegian School of Economics and Business Administration and CEPR August 31, 2004 First draft Abstract This paper seeks

More information

Identifying the Monetary Policy Shock Christiano et al. (1999)

Identifying the Monetary Policy Shock Christiano et al. (1999) Identifying the Monetary Policy Shock Christiano et al. (1999) The question we are asking is: What are the consequences of a monetary policy shock a shock which is purely related to monetary conditions

More information

CROSS-COUNTRY DIFFERENCES IN PRODUCTIVITY: THE ROLE OF ALLOCATION AND SELECTION

CROSS-COUNTRY DIFFERENCES IN PRODUCTIVITY: THE ROLE OF ALLOCATION AND SELECTION ONLINE APPENDIX CROSS-COUNTRY DIFFERENCES IN PRODUCTIVITY: THE ROLE OF ALLOCATION AND SELECTION By ERIC BARTELSMAN, JOHN HALTIWANGER AND STEFANO SCARPETTA This appendix presents a detailed sensitivity

More information

Wooldridge, Introductory Econometrics, 4th ed. Chapter 15: Instrumental variables and two stage least squares

Wooldridge, Introductory Econometrics, 4th ed. Chapter 15: Instrumental variables and two stage least squares Wooldridge, Introductory Econometrics, 4th ed. Chapter 15: Instrumental variables and two stage least squares Many economic models involve endogeneity: that is, a theoretical relationship does not fit

More information

FEDERAL RESERVE BANK of ATLANTA

FEDERAL RESERVE BANK of ATLANTA FEDERAL RESERVE BANK of ATLANTA On the Solution of the Growth Model with Investment-Specific Technological Change Jesús Fernández-Villaverde and Juan Francisco Rubio-Ramírez Working Paper 2004-39 December

More information

Global Value Chain Participation and Current Account Imbalances

Global Value Chain Participation and Current Account Imbalances Global Value Chain Participation and Current Account Imbalances Johannes Brumm University of Zurich Georgios Georgiadis European Central Bank Johannes Gräb European Central Bank Fabian Trottner Princeton

More information

Journal of Asian Business Strategy ON THE RELATIONSHIP BETWEEN FOREIGN TRADE AND REGIONAL DISPARITY IN CHINA IN THE POST-REFORM ERA

Journal of Asian Business Strategy ON THE RELATIONSHIP BETWEEN FOREIGN TRADE AND REGIONAL DISPARITY IN CHINA IN THE POST-REFORM ERA 2016 Asian Economic and Social Society. All rights reserved ISSN (P): 2309-8295, ISSN (E): 2225-4226 Volume 6, Issue 3, 2016, pp. 50-62 Journal of Asian Business Strategy http://aessweb.com/journal-detail.php?id=5006

More information

Repeated observations on the same cross-section of individual units. Important advantages relative to pure cross-section data

Repeated observations on the same cross-section of individual units. Important advantages relative to pure cross-section data Panel data Repeated observations on the same cross-section of individual units. Important advantages relative to pure cross-section data - possible to control for some unobserved heterogeneity - possible

More information

A Study on Differences of China s Regional Economic Development Level Based on Cluster Analysis

A Study on Differences of China s Regional Economic Development Level Based on Cluster Analysis MATEC Web of Conferences 22, 0 5 022 ( 2015) DOI: 10.1051/ matec conf / 201 5 220 5 022 C Owned by the authors, published by EDP Sciences, 2015 A Study on Differences of China s Regional Economic Development

More information

1 Bewley Economies with Aggregate Uncertainty

1 Bewley Economies with Aggregate Uncertainty 1 Bewley Economies with Aggregate Uncertainty Sofarwehaveassumedawayaggregatefluctuations (i.e., business cycles) in our description of the incomplete-markets economies with uninsurable idiosyncratic risk

More information

International Development

International Development International Development Discipline/Multi-discipline or trans-disciplinary field Tahmina Rashid Associate Professor, International Studies What is Development? a. Development as a state or condition-static

More information

A Summary of Economic Methodology

A Summary of Economic Methodology A Summary of Economic Methodology I. The Methodology of Theoretical Economics All economic analysis begins with theory, based in part on intuitive insights that naturally spring from certain stylized facts,

More information

A Study of Hotel Performance Under Urbanization in China

A Study of Hotel Performance Under Urbanization in China A Study of Hotel Performance Under Urbanization in China Jian Ming LUO, Faculty of International Tourism and Management, City University of Macau, Macao. E-mail: kennyluo@cityu.mo Abstract This paper studied

More information

LOCATIONAL PREFERENCES OF FDI FIRMS IN TURKEY

LOCATIONAL PREFERENCES OF FDI FIRMS IN TURKEY LOCATIONAL PREFERENCES OF FDI FIRMS IN TURKEY Prof. Dr. Lale BERKÖZ Assist. Prof. Dr.S. SenceTÜRK I.T.U. Faculty of Architecture Istanbul/TURKEY E-mail: lberkoz@itu.edu.tr INTRODUCTION Foreign direct investment

More information

System GMM estimation of Empirical Growth Models

System GMM estimation of Empirical Growth Models System GMM estimation of Empirical Growth Models ELISABETH DORNETSHUMER June 29, 2007 1 Introduction This study based on the paper "GMM Estimation of Empirical Growth Models" by Stephan Bond, Anke Hoeffler

More information

DATABASE AND METHODOLOGY

DATABASE AND METHODOLOGY CHAPTER 3 DATABASE AND METHODOLOGY In the present chapter, sources of database used and methodology applied for the empirical analysis has been presented The whole chapter has been divided into three sections

More information

1 The Basic RBC Model

1 The Basic RBC Model IHS 2016, Macroeconomics III Michael Reiter Ch. 1: Notes on RBC Model 1 1 The Basic RBC Model 1.1 Description of Model Variables y z k L c I w r output level of technology (exogenous) capital at end of

More information

Linear dynamic panel data models

Linear dynamic panel data models Linear dynamic panel data models Laura Magazzini University of Verona L. Magazzini (UniVR) Dynamic PD 1 / 67 Linear dynamic panel data models Dynamic panel data models Notation & Assumptions One of the

More information

Do Economic Reforms Accelerate Urban Growth? The Case of China

Do Economic Reforms Accelerate Urban Growth? The Case of China Do Economic Reforms Accelerate Urban Growth? The Case of China Gordon Anderson* and Ying Ge** Abstract This paper examines the determinants of city growth in China. We provide evidence that economic reforms

More information

RBC Model with Indivisible Labor. Advanced Macroeconomic Theory

RBC Model with Indivisible Labor. Advanced Macroeconomic Theory RBC Model with Indivisible Labor Advanced Macroeconomic Theory 1 Last Class What are business cycles? Using HP- lter to decompose data into trend and cyclical components Business cycle facts Standard RBC

More information

PubPol 201. Module 3: International Trade Policy. Class 4 Outline. Class 4 Outline. Class 4 China Shock

PubPol 201. Module 3: International Trade Policy. Class 4 Outline. Class 4 Outline. Class 4 China Shock PubPol 201 Module 3: International Trade Policy Class 4 China s growth The The ADH analysis Other sources Class 4 Outline Lecture 4: China 2 China s growth The The ADH analysis Other sources Class 4 Outline

More information

Linking Proximity to Secondary Cities vs Mega Cities, Agricultural Performance, & Nonfarm Employment of Rural Households in China

Linking Proximity to Secondary Cities vs Mega Cities, Agricultural Performance, & Nonfarm Employment of Rural Households in China Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Linking Proximity to Secondary Cities vs Mega Cities, Agricultural Performance, & Nonfarm Employment of Rural Households

More information

Sixty years later, is Kuznets still right? Evidence from Sub-Saharan Africa

Sixty years later, is Kuznets still right? Evidence from Sub-Saharan Africa Quest Journals Journal of Research in Humanities and Social Science Volume 3 ~ Issue 6 (2015) pp:37-41 ISSN(Online) : 2321-9467 www.questjournals.org Research Paper Sixty years later, is Kuznets still

More information

The TransPacific agreement A good thing for VietNam?

The TransPacific agreement A good thing for VietNam? The TransPacific agreement A good thing for VietNam? Jean Louis Brillet, France For presentation at the LINK 2014 Conference New York, 22nd 24th October, 2014 Advertisement!!! The model uses EViews The

More information

Recent Advances in the Field of Trade Theory and Policy Analysis Using Micro-Level Data

Recent Advances in the Field of Trade Theory and Policy Analysis Using Micro-Level Data Recent Advances in the Field of Trade Theory and Policy Analysis Using Micro-Level Data July 2012 Bangkok, Thailand Cosimo Beverelli (World Trade Organization) 1 Content a) Endogeneity b) Instrumental

More information

IDE Research Bulletin

IDE Research Bulletin http://www.ide.go.jp IDE Research Bulletin Research Summary based on papers prepared for publication in academic journals with the aim of contributing to the academia Empirical studies on industrial clusters

More information

The Ramsey Model. (Lecture Note, Advanced Macroeconomics, Thomas Steger, SS 2013)

The Ramsey Model. (Lecture Note, Advanced Macroeconomics, Thomas Steger, SS 2013) The Ramsey Model (Lecture Note, Advanced Macroeconomics, Thomas Steger, SS 213) 1 Introduction The Ramsey model (or neoclassical growth model) is one of the prototype models in dynamic macroeconomics.

More information

Team Production and the Allocation of Creativity across Global and Local Sectors

Team Production and the Allocation of Creativity across Global and Local Sectors RIETI Discussion Paper Series 15-E-111 Team Production and the Allocation of Creativity across Global and Local Sectors NAGAMACHI Kohei Kagawa University The Research Institute of Economy, Trade and Industry

More information

The Output Effect of Trade Openness in China: Evidence from Provincial Data

The Output Effect of Trade Openness in China: Evidence from Provincial Data The Output Effect of Trade Openness in China: Evidence from Provincial Data Jang C. Jin (jcjin@cuhk.edu.hk) Chinese University of Hong Kong, Hong Kong Abstract Unlike other studies that concentrate on

More information

Advanced Macroeconomics

Advanced Macroeconomics Advanced Macroeconomics The Ramsey Model Marcin Kolasa Warsaw School of Economics Marcin Kolasa (WSE) Ad. Macro - Ramsey model 1 / 30 Introduction Authors: Frank Ramsey (1928), David Cass (1965) and Tjalling

More information

A System Dynamics Glossary

A System Dynamics Glossary A System Dynamics Glossary Compiled by David N. Ford, Ph.D., P.E. Last Update: July 3, 2018 accumulation (integration): the collection of some quantity over time. Examples of accumulation include water

More information

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities Chinese Journal of Urban and Environmental Studies Vol. 2, No. 1 (2014) 1450004 (9 pages) World Scientific Publishing Company DOI: 10.1142/S2345748114500043 Analysis of the Tourism Locations of Chinese

More information

Commuting, Migration, and Rural Development

Commuting, Migration, and Rural Development MPRA Munich Personal RePEc Archive Commuting, Migration, and Rural Development Ayele Gelan Socio-economic Research Program, The Macaulay Institute, Aberdeen, UK 003 Online at http://mpra.ub.uni-muenchen.de/903/

More information

1 Motivation for Instrumental Variable (IV) Regression

1 Motivation for Instrumental Variable (IV) Regression ECON 370: IV & 2SLS 1 Instrumental Variables Estimation and Two Stage Least Squares Econometric Methods, ECON 370 Let s get back to the thiking in terms of cross sectional (or pooled cross sectional) data

More information

Sampling Scheme for 2003 General Social Survey of China

Sampling Scheme for 2003 General Social Survey of China Sampling Scheme for 2003 General Social Survey of China 1. Sampling Unit This survey uses a five-stage stratified sampling scheme with unequal probabilities. The sampling units at each stage are as follows:

More information

Economic Growth: Lecture 8, Overlapping Generations

Economic Growth: Lecture 8, Overlapping Generations 14.452 Economic Growth: Lecture 8, Overlapping Generations Daron Acemoglu MIT November 20, 2018 Daron Acemoglu (MIT) Economic Growth Lecture 8 November 20, 2018 1 / 46 Growth with Overlapping Generations

More information

Linear Models in Econometrics

Linear Models in Econometrics Linear Models in Econometrics Nicky Grant At the most fundamental level econometrics is the development of statistical techniques suited primarily to answering economic questions and testing economic theories.

More information

Empirical approaches in public economics

Empirical approaches in public economics Empirical approaches in public economics ECON4624 Empirical Public Economics Fall 2016 Gaute Torsvik Outline for today The canonical problem Basic concepts of causal inference Randomized experiments Non-experimental

More information

Wooldridge, Introductory Econometrics, 3d ed. Chapter 16: Simultaneous equations models. An obvious reason for the endogeneity of explanatory

Wooldridge, Introductory Econometrics, 3d ed. Chapter 16: Simultaneous equations models. An obvious reason for the endogeneity of explanatory Wooldridge, Introductory Econometrics, 3d ed. Chapter 16: Simultaneous equations models An obvious reason for the endogeneity of explanatory variables in a regression model is simultaneity: that is, one

More information

DEVELOPMENT OF URBAN INFRASTRUCTURES AND POPULATION CHANGE IN CHINA

DEVELOPMENT OF URBAN INFRASTRUCTURES AND POPULATION CHANGE IN CHINA DEVELOPMENT OF URBAN INFRASTRUCTURES AND POPULATION CHANGE IN CHINA A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the

More information

Macroeconomics II. Dynamic AD-AS model

Macroeconomics II. Dynamic AD-AS model Macroeconomics II Dynamic AD-AS model Vahagn Jerbashian Ch. 14 from Mankiw (2010) Spring 2018 Where we are heading to We will incorporate dynamics into the standard AD-AS model This will offer another

More information

Dynamic AD-AS model vs. AD-AS model Notes. Dynamic AD-AS model in a few words Notes. Notation to incorporate time-dimension Notes

Dynamic AD-AS model vs. AD-AS model Notes. Dynamic AD-AS model in a few words Notes. Notation to incorporate time-dimension Notes Macroeconomics II Dynamic AD-AS model Vahagn Jerbashian Ch. 14 from Mankiw (2010) Spring 2018 Where we are heading to We will incorporate dynamics into the standard AD-AS model This will offer another

More information

Income Distribution Dynamics with Endogenous Fertility. By Michael Kremer and Daniel Chen

Income Distribution Dynamics with Endogenous Fertility. By Michael Kremer and Daniel Chen Income Distribution Dynamics with Endogenous Fertility By Michael Kremer and Daniel Chen I. Introduction II. III. IV. Theory Empirical Evidence A More General Utility Function V. Conclusions Introduction

More information

A Course in Applied Econometrics Lecture 4: Linear Panel Data Models, II. Jeff Wooldridge IRP Lectures, UW Madison, August 2008

A Course in Applied Econometrics Lecture 4: Linear Panel Data Models, II. Jeff Wooldridge IRP Lectures, UW Madison, August 2008 A Course in Applied Econometrics Lecture 4: Linear Panel Data Models, II Jeff Wooldridge IRP Lectures, UW Madison, August 2008 5. Estimating Production Functions Using Proxy Variables 6. Pseudo Panels

More information

Relationships between phases of business cycles in two large open economies

Relationships between phases of business cycles in two large open economies Journal of Regional Development Studies2010 131 Relationships between phases of business cycles in two large open economies Ken-ichi ISHIYAMA 1. Introduction We have observed large increases in trade and

More information

Wooldridge, Introductory Econometrics, 3d ed. Chapter 9: More on specification and data problems

Wooldridge, Introductory Econometrics, 3d ed. Chapter 9: More on specification and data problems Wooldridge, Introductory Econometrics, 3d ed. Chapter 9: More on specification and data problems Functional form misspecification We may have a model that is correctly specified, in terms of including

More information

Field Course Descriptions

Field Course Descriptions Field Course Descriptions Ph.D. Field Requirements 12 credit hours with 6 credit hours in each of two fields selected from the following fields. Each class can count towards only one field. Course descriptions

More information

Opportunities and challenges of HCMC in the process of development

Opportunities and challenges of HCMC in the process of development Opportunities and challenges of HCMC in the process of development Lê Văn Thành HIDS HCMC, Sept. 16-17, 2009 Contents The city starting point Achievement and difficulties Development perspective and goals

More information

Interest Rate Liberalization and Capital Misallocation 1

Interest Rate Liberalization and Capital Misallocation 1 Interest Rate Liberalization and Capital Misallocation 1 Zheng Liu 1 Pengfei Wang 2 Zhiwei Xu 3 1 Federal Reserve Bank of San Francisco 2 Hong Kong University of Science and Technology 3 Shanghai Jiao

More information

Online Appendix The Growth of Low Skill Service Jobs and the Polarization of the U.S. Labor Market. By David H. Autor and David Dorn

Online Appendix The Growth of Low Skill Service Jobs and the Polarization of the U.S. Labor Market. By David H. Autor and David Dorn Online Appendix The Growth of Low Skill Service Jobs and the Polarization of the U.S. Labor Market By David H. Autor and David Dorn 1 2 THE AMERICAN ECONOMIC REVIEW MONTH YEAR I. Online Appendix Tables

More information

Advanced Economic Growth: Lecture 3, Review of Endogenous Growth: Schumpeterian Models

Advanced Economic Growth: Lecture 3, Review of Endogenous Growth: Schumpeterian Models Advanced Economic Growth: Lecture 3, Review of Endogenous Growth: Schumpeterian Models Daron Acemoglu MIT September 12, 2007 Daron Acemoglu (MIT) Advanced Growth Lecture 3 September 12, 2007 1 / 40 Introduction

More information

Economic Growth: Lecture 9, Neoclassical Endogenous Growth

Economic Growth: Lecture 9, Neoclassical Endogenous Growth 14.452 Economic Growth: Lecture 9, Neoclassical Endogenous Growth Daron Acemoglu MIT November 28, 2017. Daron Acemoglu (MIT) Economic Growth Lecture 9 November 28, 2017. 1 / 41 First-Generation Models

More information

Implementation Performance Evaluation on Land Use Planning: A Case of Chengdu, China

Implementation Performance Evaluation on Land Use Planning: A Case of Chengdu, China Cross-Cultural Communication Vol. 8, No. 4, 2012, pp. 34-38 DOI:10.3968/j.ccc.1923670020120804.1020 ISSN 1712-8358[Print] ISSN 1923-6700[Online] www.cscanada.net www.cscanada.org Implementation Performance

More information

November 29, World Urban Forum 6. Prosperity of Cities: Balancing Ecology, Economy and Equity. Concept Note

November 29, World Urban Forum 6. Prosperity of Cities: Balancing Ecology, Economy and Equity. Concept Note November 29, 2010 World Urban Forum 6 Prosperity of Cities: Balancing Ecology, Economy and Equity Concept Note 1 CONTENT Thematic Continuity Conceptualizing the Theme The 6 Domains of Prosperity The WUF

More information

Multi scale and multi sensor analysis of urban cluster development and agricultural land loss in China and India

Multi scale and multi sensor analysis of urban cluster development and agricultural land loss in China and India Multi scale and multi sensor analysis of urban cluster development and agricultural land loss in China and India Karen C. Seto, PI, Yale Michail Fragkias, Co I, Arizona State Annemarie Schneider, Co I,

More information

Dealing With Endogeneity

Dealing With Endogeneity Dealing With Endogeneity Junhui Qian December 22, 2014 Outline Introduction Instrumental Variable Instrumental Variable Estimation Two-Stage Least Square Estimation Panel Data Endogeneity in Econometrics

More information

Cities in Bad Shape: Urban Geometry in India

Cities in Bad Shape: Urban Geometry in India Cities in Bad Shape: Urban Geometry in India Mariaflavia Harari MIT IGC Cities Research Group Conference 21 May 2015 Introduction Why Study City Shape A wide range of factors determine intra-urban commuting

More information

Lecture 4 The Centralized Economy: Extensions

Lecture 4 The Centralized Economy: Extensions Lecture 4 The Centralized Economy: Extensions Leopold von Thadden University of Mainz and ECB (on leave) Advanced Macroeconomics, Winter Term 2013 1 / 36 I Motivation This Lecture considers some applications

More information

Markov Perfect Equilibria in the Ramsey Model

Markov Perfect Equilibria in the Ramsey Model Markov Perfect Equilibria in the Ramsey Model Paul Pichler and Gerhard Sorger This Version: February 2006 Abstract We study the Ramsey (1928) model under the assumption that households act strategically.

More information

Beyond the Target Customer: Social Effects of CRM Campaigns

Beyond the Target Customer: Social Effects of CRM Campaigns Beyond the Target Customer: Social Effects of CRM Campaigns Eva Ascarza, Peter Ebbes, Oded Netzer, Matthew Danielson Link to article: http://journals.ama.org/doi/abs/10.1509/jmr.15.0442 WEB APPENDICES

More information

The Geography of Development: Evaluating Migration Restrictions and Coastal Flooding

The Geography of Development: Evaluating Migration Restrictions and Coastal Flooding The Geography of Development: Evaluating Migration Restrictions and Coastal Flooding Klaus Desmet SMU Dávid Krisztián Nagy Princeton University Esteban Rossi-Hansberg Princeton University World Bank, February

More information

Wage and price setting. Slides for 26. August 2003 lecture

Wage and price setting. Slides for 26. August 2003 lecture 1 B&W s derivation of the Phillips curve Wage and price setting. Slides for 26. August 2003 lecture Ragnar Nymoen University of Oslo, Department of Economics Ch 12.3: The Battle of the mark-ups as a framework

More information

Solow Growth Model. Michael Bar. February 28, Introduction Some facts about modern growth Questions... 4

Solow Growth Model. Michael Bar. February 28, Introduction Some facts about modern growth Questions... 4 Solow Growth Model Michael Bar February 28, 208 Contents Introduction 2. Some facts about modern growth........................ 3.2 Questions..................................... 4 2 The Solow Model 5

More information

The Contribution Rate of Thrice Industrial Agglomeration to Industrial Growth in Ningxia The Calculate Based on Cobb-Douglas Function.

The Contribution Rate of Thrice Industrial Agglomeration to Industrial Growth in Ningxia The Calculate Based on Cobb-Douglas Function. International Conference on Economics, Social Science, Arts, Education and Management Engineering (ESSAEME 2015) The Contribution Rate of Thrice Industrial Agglomeration to Industrial Growth in Ningxia

More information

Next, we discuss econometric methods that can be used to estimate panel data models.

Next, we discuss econometric methods that can be used to estimate panel data models. 1 Motivation Next, we discuss econometric methods that can be used to estimate panel data models. Panel data is a repeated observation of the same cross section Panel data is highly desirable when it is

More information

Stagnation Traps. Gianluca Benigno and Luca Fornaro

Stagnation Traps. Gianluca Benigno and Luca Fornaro Stagnation Traps Gianluca Benigno and Luca Fornaro May 2015 Research question and motivation Can insu cient aggregate demand lead to economic stagnation? This question goes back, at least, to the Great

More information

Population Growth and Economic Development: Test for Causality

Population Growth and Economic Development: Test for Causality The Lahore Journal of Economics 11 : 2 (Winter 2006) pp. 71-77 Population Growth and Economic Development: Test for Causality Khalid Mushtaq * Abstract This paper examines the existence of a long-run relationship

More information

Dynamic Panel Data Models

Dynamic Panel Data Models June 23, 2010 Contents Motivation 1 Motivation 2 Basic set-up Problem Solution 3 4 5 Literature Motivation Many economic issues are dynamic by nature and use the panel data structure to understand adjustment.

More information

Toulouse School of Economics, M2 Macroeconomics 1 Professor Franck Portier. Exam Solution

Toulouse School of Economics, M2 Macroeconomics 1 Professor Franck Portier. Exam Solution Toulouse School of Economics, 2013-2014 M2 Macroeconomics 1 Professor Franck Portier Exam Solution This is a 3 hours exam. Class slides and any handwritten material are allowed. You must write legibly.

More information

Spatial Aspects of Trade Liberalization in Colombia: A General Equilibrium Approach. Eduardo Haddad Jaime Bonet Geoffrey Hewings Fernando Perobelli

Spatial Aspects of Trade Liberalization in Colombia: A General Equilibrium Approach. Eduardo Haddad Jaime Bonet Geoffrey Hewings Fernando Perobelli Spatial Aspects of Trade Liberalization in Colombia: A General Equilibrium Approach Eduardo Haddad Jaime Bonet Geoffrey Hewings Fernando Perobelli Outline Motivation The CEER model Simulation results Final

More information

On Spatial Dynamics. Klaus Desmet Universidad Carlos III. and. Esteban Rossi-Hansberg Princeton University. April 2009

On Spatial Dynamics. Klaus Desmet Universidad Carlos III. and. Esteban Rossi-Hansberg Princeton University. April 2009 On Spatial Dynamics Klaus Desmet Universidad Carlos and Esteban Rossi-Hansberg Princeton University April 2009 Desmet and Rossi-Hansberg () On Spatial Dynamics April 2009 1 / 15 ntroduction Economists

More information

Applied Econometrics (MSc.) Lecture 3 Instrumental Variables

Applied Econometrics (MSc.) Lecture 3 Instrumental Variables Applied Econometrics (MSc.) Lecture 3 Instrumental Variables Estimation - Theory Department of Economics University of Gothenburg December 4, 2014 1/28 Why IV estimation? So far, in OLS, we assumed independence.

More information

Under-Employment and the Trickle-Down of Unemployment - Online Appendix Not for Publication

Under-Employment and the Trickle-Down of Unemployment - Online Appendix Not for Publication Under-Employment and the Trickle-Down of Unemployment - Online Appendix Not for Publication Regis Barnichon Yanos Zylberberg July 21, 2016 This online Appendix contains a more comprehensive description

More information

Econometrics in a nutshell: Variation and Identification Linear Regression Model in STATA. Research Methods. Carlos Noton.

Econometrics in a nutshell: Variation and Identification Linear Regression Model in STATA. Research Methods. Carlos Noton. 1/17 Research Methods Carlos Noton Term 2-2012 Outline 2/17 1 Econometrics in a nutshell: Variation and Identification 2 Main Assumptions 3/17 Dependent variable or outcome Y is the result of two forces:

More information

Economics 270c Graduate Development Economics. Professor Ted Miguel Department of Economics University of California, Berkeley

Economics 270c Graduate Development Economics. Professor Ted Miguel Department of Economics University of California, Berkeley Economics 270c Graduate Development Economics Professor Ted Miguel Department of Economics University of California, Berkeley Economics 270c Graduate Development Economics Lecture 2 January 27, 2009 Lecture

More information

Chapter 4. Explanation of the Model. Satoru Kumagai Inter-disciplinary Studies, IDE-JETRO, Japan

Chapter 4. Explanation of the Model. Satoru Kumagai Inter-disciplinary Studies, IDE-JETRO, Japan Chapter 4 Explanation of the Model Satoru Kumagai Inter-disciplinary Studies, IDE-JETRO, Japan Toshitaka Gokan Inter-disciplinary Studies, IDE-JETRO, Japan Ikumo Isono Bangkok Research Center, IDE-JETRO,

More information

A STUDY OF HUMAN DEVELOPMENT APPROACH TO THE DEVELOPMENT OF NORTH EASTERN REGION OF INDIA

A STUDY OF HUMAN DEVELOPMENT APPROACH TO THE DEVELOPMENT OF NORTH EASTERN REGION OF INDIA ABSTRACT A STUDY OF HUMAN DEVELOPMENT APPROACH TO THE DEVELOPMENT OF NORTH EASTERN REGION OF INDIA Human development by emphasizing on capability approach differs crucially from the traditional approaches

More information

Part VII. Accounting for the Endogeneity of Schooling. Endogeneity of schooling Mean growth rate of earnings Mean growth rate Selection bias Summary

Part VII. Accounting for the Endogeneity of Schooling. Endogeneity of schooling Mean growth rate of earnings Mean growth rate Selection bias Summary Part VII Accounting for the Endogeneity of Schooling 327 / 785 Much of the CPS-Census literature on the returns to schooling ignores the choice of schooling and its consequences for estimating the rate

More information

Creating a Provincial Long-Term Growth Model for China using the Kohonen Algorithm

Creating a Provincial Long-Term Growth Model for China using the Kohonen Algorithm Creating a Provincial Long-Term Growth Model for China using the Kohonen Algorithm Elpida Makriyannis, Department of Computer Science, University College London, Gower Street, London, WC1E 6BT, U.K. Philip

More information

Neoclassical Business Cycle Model

Neoclassical Business Cycle Model Neoclassical Business Cycle Model Prof. Eric Sims University of Notre Dame Fall 2015 1 / 36 Production Economy Last time: studied equilibrium in an endowment economy Now: study equilibrium in an economy

More information

Impact Evaluation of Rural Road Projects. Dominique van de Walle World Bank

Impact Evaluation of Rural Road Projects. Dominique van de Walle World Bank Impact Evaluation of Rural Road Projects Dominique van de Walle World Bank Introduction General consensus that roads are good for development & living standards A sizeable share of development aid and

More information

Decentralisation and its efficiency implications in suburban public transport

Decentralisation and its efficiency implications in suburban public transport Decentralisation and its efficiency implications in suburban public transport Daniel Hörcher 1, Woubit Seifu 2, Bruno De Borger 2, and Daniel J. Graham 1 1 Imperial College London. South Kensington Campus,

More information

New Notes on the Solow Growth Model

New Notes on the Solow Growth Model New Notes on the Solow Growth Model Roberto Chang September 2009 1 The Model The firstingredientofadynamicmodelisthedescriptionofthetimehorizon. In the original Solow model, time is continuous and the

More information

Modeling firms locational choice

Modeling firms locational choice Modeling firms locational choice Giulio Bottazzi DIMETIC School Pécs, 05 July 2010 Agglomeration derive from some form of externality. Drivers of agglomeration can be of two types: pecuniary and non-pecuniary.

More information

Economics 2450A: Public Economics Section 8: Optimal Minimum Wage and Introduction to Capital Taxation

Economics 2450A: Public Economics Section 8: Optimal Minimum Wage and Introduction to Capital Taxation Economics 2450A: Public Economics Section 8: Optimal Minimum Wage and Introduction to Capital Taxation Matteo Paradisi November 1, 2016 In this Section we develop a theoretical analysis of optimal minimum

More information

problem. max Both k (0) and h (0) are given at time 0. (a) Write down the Hamilton-Jacobi-Bellman (HJB) Equation in the dynamic programming

problem. max Both k (0) and h (0) are given at time 0. (a) Write down the Hamilton-Jacobi-Bellman (HJB) Equation in the dynamic programming 1. Endogenous Growth with Human Capital Consider the following endogenous growth model with both physical capital (k (t)) and human capital (h (t)) in continuous time. The representative household solves

More information

Simultaneous Equation Models Learning Objectives Introduction Introduction (2) Introduction (3) Solving the Model structural equations

Simultaneous Equation Models Learning Objectives Introduction Introduction (2) Introduction (3) Solving the Model structural equations Simultaneous Equation Models. Introduction: basic definitions 2. Consequences of ignoring simultaneity 3. The identification problem 4. Estimation of simultaneous equation models 5. Example: IS LM model

More information