Provincial convergence to the law of one price in the People Republic
of China - A pairwise approach
1Master Thesis
Marcel Vos - University of Groningen
Under Supervision of: Dr. T.M Harchaoui - University of Groningen 15 June 2015
“When the wind of change blows, some build walls while others build windmills.” (Li Keqiang, Premier of the People Republic of China)
Abstract
______________________________________________________________________
The subject of this paper is testing the convergence of monthly provincial consumer prices in China, by using a pairwise unit root approach for monthly panel data ranging from 1995-2014. All binary combinations of province pairs are tested, which stands in contrast to the more traditional methods that use a reference region. The null
hypothesis of no convergence is rejected for a relatively large proportion of provincial pairs depending on the test utilized. Moreover the quantity of converging provincial pairs drastically increased over time due to the development of China in the last 10 years. Furthermore a probit analysis shows that distance, and differentials in development- and income growth rate negatively influence the likelihood of convergence.
______________________________________________________________________
2
1 Introduction
The pace and scale of China’s structural transformations are nothing short of
remarkable leaving, little room for disagreement in the economics profession. Economic growth in the manufacturing sector has been advancing at a staggering rate triggered by the displacement of inefficient state-owned enterprises with new, more efficient, private enterprises with a focus on consumers’ goods exports. This development remains inextricably meshed with a massive rural-urban migration, which has been at the core of the success in the ‘Made in China’ model. The advance in real per capita income that resulted from this process has lifted millions of Chinese out of poverty and contributed along the way to the creation of a middle class with spending patterns towards discretionary items. This advance in standard of living, combined with a
massive surplus in the current account, low real interest rates, has set in motion inflation pressures with a series of adverse effects such as higher wages, the prospect of
deterioration of China’s international competitive edge and a pass-through of inflation to developed nations. This paper examines the long-run consumer price convergence across China’s provinces using recent advances in the econometrics of panel data to a problem that has a clear spatial dimension. Owing to the presence of striking disparities between the coastal and inland regions in China, it is furthermore investigated whether adjustment towards long-run equilibrium is inversely related to distance, and the growth rates of development, income and population.
3
Testing for price converging makes it possible to assess the degree of market integration (Gluschenko, 2010). The null hypothesis of no convergence is rejected for a relatively large amount of provincial pairs, which implies a moderate degree of
subnational price convergence and thereby market integration in China. Moreover in the last ten years more provincial pairs are converging than in the ten years before. Spatial separation, as well as differentials in development-, population- and income growth rates between provincial pairs negatively influences the likelihood of converging during certain time periods.
The motivation for this research is twofold, first of all the pairwise approach has major benefits over the usual international and sub-national price convergence
methodologies which mostly build on panel unit-root tests as well, however those methodologies use a certain geographical area as a reference point to which all other regions are then compared. The pairwise approach has the advantage that all provincial binary combinations of price levels are compared to each other. Using a reference region makes the analysis sensitive to the choice of reference, in this case a province. Moreover it does not exploit the information of all other cross price levels, Peseran et al. (2009) argue that the estimation is subject to a high level of cross-sectional
4
Moreover while there is a small body of research on price convergence in China,
the country is starting a new step in their development by implementing the 13th
five-year plan in 2016. Given the quick development in China the research body should stay up to date. Since the current research body, which uses the above mentioned reference based research methodology, is conducted China has seen a drastic increase in the amount of ‘affluent’ consumers as well as a growth of the middle class, this
phenomenon relates to one of the most important developments regarding the
integration of provinces, which is the focus on increasing the domestic demand in the current five year plan of the central government.
Adding to this, the increase of domestic buying power most likely also ‘upgrades’ the products Chinese consumers procure and thereby adjust the basket that is utilized to calculate CPI. All these developments combined with the objectives of the new five year plan that influence or require market integration such as development of the western regions, an inflation target of below 4% as well as increasing the share of the service sector (KPMG, 2011; BCG, 2012) require a new look into Chinese provincial price convergence and thereby the degree of market integration.
The remainder of the paper is structured as following, in part 2 the existing literature will be reviewed. Part 3 describes the data, while part 4 discussed the
5
2 Literature review
Before looking into the research on price convergence and purchasing power parity later in this chapter, it is important to familiarize with the development path China took, with special focus on the years 1995 until 2014, which are the subject of this study.
2.1 Institutional background
China is a nation in transition from a state planned economy towards a market-based economy in a socialist setting. While this section is by no means a recollection of all stages of liberalization the goal is to give an insight in the major events in China’s unprecedented journey towards a liberal market, and to give the background that is required to study market integration. China’s modern liberalization strategy is based on gradualism, instead of great leaps forward. Meaning that the planned changes do not happen abruptly but indeed are implemented over time in order to avoid enormous shocks like the ones witnesses in Eastern Europe during their transition, and avoiding the mistaken in earlier rounds of reform. The reform in China can be split up in three stages (Fan and Wei, 2005)(We,1999)(Suliman, 1998):
First stage (1979-1984)
The first stage mostly consisted of decentralization of the government and reform in the rural areas of China as most of the Chinese occupation was in agriculture at the time. Moreover China created the economic trading zones in 1979.
Second stage (1984-1992)
The second stage of reform consisted of a shift towards urban policy. The first major change related to price policy was initiated in 1984 when the so-called dual-track
6
services and goods. While certain agricultural goods were allowed to rise in order to reach the equilibrium price.
Third stage (1992-to day)
A couple of years after the Tiananmen square riots, China embarked upon further reforms, in 1993 the 8th national people congress announced the development of a so called ‘socialist market economy’ and around the same time there was also a reform in price controlling policy with the introduction of three pricing categories, namely planned prices; guided prices and the most liberal: market prices. Goods could gradually
progress through the three stages. By the end of the 1990s, 90 percent of prices were set by the market. One of the most important steps in liberalization however is the entrance into the WTO, which happened in December 2001, after years of negotiation. And as of today China is still in the process of opening certain markets to the rest of the world, gradually of course.
7
Figure 1: China per capita GDP
Figure 1 displays the growth in GDP per capita for all of China, this measure dramatically increased from roughly 5000 Yuan to almost 47000 Yuan over the twenty year period. When the sample is separated in two equal 10 year timeframes namely 1995-2004 and 2005-2014, as is done for the estimations later in this paper, the
average annual growth in per capita GDP was 9.3 percent for the first ten years. While this is cerntainly remarkable, during the years 2005-2014 an even more impressive 12.6 percent average annual growth in per capita GDP was reache
Figure 2 shows the employment in agriculture as a percentage of the total population, which can be used as a proxy for development (see chapter 3). From 1995-1999 the agricultural employment only decreases slightly before the downward trend picked up speed from 2000-2015. This makes the average annual decrease 1.25 percent over the first ten years. The strong downward trend then proceeds until 2007 where the downward trend became less strong. Overall the average annual decrease over the remaining years for which data is published (2005-2012) is 1.71 percent, which is slightly higher as the trend in the first ten year of the sample.
0 10,000 20,000 30,000 40,000 50,000
8
So overall it can be concluded that based on the higher average annual drop in agricultural employment as a percentage of population and higher average annual growth in GDP per capita in the 2004-2015 sample compared the first sample, in the last ten years of the total 20 year sample China seem to have a faster development pace than the ten years before.
Figure 2: Employment in agriculture
2.2 Price convergence literature
At the basis of price convergence, which can be used to test the degree of market integration (Gluschenko,2010), is the law of one price (LOP). LOP can be defined as: “Goods will be traded for the same price across all markets in different regions” (Lee et al, 2011 p.71). The intuition behind this proposition is quite simple, when identical goods are not equal in price across regions, opportunists could easily use arbitrage to make a
0 5 10 15 20 25 30 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Agricultural employment as percentage of
9
profit. Purchasing Power Parity (PPP) is the multi good extension that is used in international economics: ”PPP is the proposition that once prices are converted to a common currency, national price levels should be equal (Rogoff, 1996 p647). An often-used measure of convergence is the half-life; this is “the period it takes until half of the effect of a given shock dissipates” (Lan & Sylwester, 2010 p.226).
Intra-country studies, such as this one, on price convergence and the law of one price are preceded by the study of cross-country price convergence. Frankel and Rose (1996) found evidence that PPP deviations erode with a rate of 15 percent annually. For OECD countries Egger, Gruber, and Pfaffermayr (2009) conclude that for the majority of the countries there is price convergence, especially for tradable goods. Rogers (2001) finds similar results for Eurozone countries. Cheung and Lai (2000) discover half-lives of 3.31 years for industrialized countries and a much shorter 1.36 years for developing countries. Cuestas and Ordonez (2007) find evidence for price convergence in South-American nations that are members of the Mercosur agreement. Engel and Rogers (1996) argue that the distance between cities accounts for a significant variation in prices of similar goods. However it is concluded that a border between two cities plays an even larger role.
There are two seminal works in the sub-national price convergence literature. Parsley and Wei (1996) state that tradable goods converge relatively fast to parity, while services take longer. Moreover there is a clear spatial component as cities that are farther apart converge slower. Cecchetti, Sonora and Price (2002) conclude that price index convergence in the USA is surprisingly slow. With the same pairwise methods as this research, Ikeno (2014) finds price convergence to a limited extend. Further
10
Given the large difference in infrastructure, market structure and the share of the
service sector between China and western- and developed countries, findings of studies on price convergence of other developing countries and in particular fellow BRIC
countries are perhaps more applicable to the situation of China, as those countries are also rapidly developing and have large domestic markets. In Russia which like China is in transformation from a state planned economy to a market based economy,
Gluschenko (2011) reports that 40 out of the 74 regions are integrated, 18 tend towards integration and only 16 are both not integrated and are not tending towards it. For Indian cities, Morshed et al. (2006) find an overall half-life of only 3 months. However in
another study on India, Das and Bhattacharya (2008) found longer half-lives.
Unfortunately at the moment of writing there are no (English language) articles written on the law of one price or converging prices for the other BRIC country Brazil. Other research on rapidly developing countries however include Mexico (Sonora, 2005) with regional half-lives between approximately 1 to 2 years, and Indonesia, where Wimand (2009) finds a mean half-life of approximately 1.5 years.
11
Ritola (2008) studied price convergence among 36 Chinese cities; it was found that China's market integration level is fairly high when compared to international standards. It was discovered that on the long term 52 per cent of the cities exhibit convergence while in the short term only 9 percent does. Moreover the half-live of price shocks, 8 months on average, is surprisingly small especially when services are
excluded.
Based on a multi product study in 36 major cities over a time period of 13 years, Fan and Wei (2006) found evidence that most prices of goods and services in China do converge to the law of one price. They find an average half-life of 2.35 months. It is argued that as price convergence is confirmed, the Chinese provincial markets are integrated, and the policies of the last decades are successful, thereby contradicting the results of Young (2000). The authors argue that the explanations for the differential conclusion are due to a more comprehensive data set, aggregation bias, and the more sophisticated unit-root approach. Lan and Sylwester (2010) use the same data as Fan and Wei (2006) but however utilize a 3-dimension panel (time, goods, cities) method to check the robustness of Wei and Fan findings. They find half-lives of only a few months which all range around 2 months. Lee et al. (2011) argue that the 2008 financial crisis reversed some of the steps taken toward internal price CPI convergence, as provincial governments have an incentive to protect heir own firms in response to international protectionism, this protectionism hinders the trade among Chinese provinces.
The emerging Chinese energy market is the subject of Ma et al. (2007), Coal-, electricity-, gasoline- and diesel prices are converging and the energy markets are integrated. The power of the governments in price setting in the energy market
12
effect’ for prices, which starts in the large cities Beijing, Shanghai and Guangzhou and ‘flows’ to their hinterland.
So overall based on the literature and the overview of all half-lives, which can be found in appendix 2 several conclusions can be drawn. First of all conversion to
LOP/PPP in the reviewed studies is faster for intra-country studies than for cross-country studies. Based on this we can conclude and thereby use the words of Rogoff (1996), who tried to find the answer to ‘the purchasing power parity puzzle’ that while international markets are highly integrated, they however remain not nearly as
integrated as domestic markets.
Moreover while it has to be noted that the sample of reviewed studies is small and the individual papers utilize different methods, the overview of half-lives (appendix 2) indicates a difference between developed and developing countries where
surprisingly the literature indicates that developing countries are converging at a faster rate, given their shorter half-lives. One would expect higher market integration in more developed nations such as the USA. Cecchetti et al. (2000) Ikeno (2000) and Rogoff (1996) sum the general reasons given for incomplete, slow or a lack of price
convergence for both cross-country and subnational studies: (i) Trade barriers such as tariffs
(ii) Non-tariff barriers, such as bureaucratic challenges related to trade (iii) Exchange rates that fail to adjust for price level shocks
(iv) Sticky nominal price level adjustments
(v) Differential price setting in segmented markets (vi) High transport cost that reduces arbitrage profits
13
(viii) Monopoly firms (ix) Lack of labor mobility (x) Information asymmetry
In this case the generally larger share of services and other non-tradable goods in developed nations, which relates to reason vii, might explain the findings. Several of the reviewed papers (Ritola, 2008; Egger et al., 2009) found longer half-lives for these product categories. Lan and Sylwester (2010) did not only find that when compared to international convergence studies, national convergence rates are faster. After
14
3 Data
The main data set used for the pair wise unit root tests is the monthly ‘consumer price index’ from the National bureau of statistics of China. This data set “reflects the trend and degree of changes in prices of consumer goods and services purchased by urban
and rural households during a given period”. The data is monthly and split up by
15
Table 1: Provinces with the Highest and Lowest CPI Level Year
Province with the Lowest CPI Level (same month
previous year = 100)
Province with the Highest CPI Level (same month previous
year = 100)) 1995 Fujian 105.7 Guizhou 131.8 1996 Hainan 102.3 Beijing 115.0 1997 Guangxi 96.6 Sichuan 108.8 1998 Guangxi 96.1 Shanghai 106.0 1999 Heilongjiang 94.9 Shanghai 106.5 2000 Heilongjiang 94.5 Beijing 107.2 2001 Yunnan 96.6 Beijing 106.0 2002 Beijing 96.6 Qinghai 103.8 2003 Hainan 98.0 Anhui 105.5 2004 Beijing 98.7 Yunnan 108.4 2005 Qinghai 98.9 Guangxi 105.7 2006 Fujian 99.2 Hunan 104.6 2007 Beijing 100.5 Qinghai 110.4 2008 Liaoning 98.8 Qinghai 113.6 2009 Guangdong 95.8 Qinghai 104.6 2010 Beijing/Guangdong 100.0 Qinghai 108.1 2011 Qinghai 101.5 Qinghai 109.2 2012 Ningxia 100.6 Xinjiang 105.4 2013 Hunan 101.0 Qinghai 105.5 2014 Heilongjiang 100.3 Hainan/Qinghai 103.6
16
clearly discernable geographical pattern between provinces displaying the highest/lowest inflation level.
Figure 3 reports the provincial CPI based on provinces that display the highest and those displaying the lowest levels of the CPI. There is a large discrepancy between the lowest and highest levels that remains persistent until 2011-2014 where the
beginning of convergence process emerges. Other than this interesting episode, others such as the 1995-2000 and the 2000-2011 are worth emphasizing. For example, the first period reports by far the largest discrepancy, which can be considered as evidence in favor of the absence of regional convergence. The latter period indicates a
reasonably rapid pick up in the index for the provinces traditionally with the lowest inflation and a leveling off of its counterpart from the provinces with the high inflation until 2011 when it observed a significant convergence towards the lowest inflation level.
Figure 3: Descriptive statistics: Lowest and highest CPI
17 Probit estimation data and variables
For the independent variables of the probit analysis several more data sets are used. In order to create a distance matrix, consisting of the distance between all pairs of provincial capitals, the spatial separation in kilometers has been calculated using the coordinates of every provincial capital (Tageo, 2015). In order to come to a distance matrix it is assumed that the earth is a perfect sphere.
The ‘year-end provincial population’ data is downloaded from the National bureau of statistics of China. There is no data available at the moment of writing for the year 2014 and for the period 1995-2000. In order to match the time periods of the CPI sample, data for the years 1995-2000 is extrapolated by first calculating the average distribution of the total Chinese population over the provinces over the period 2000-2004. This average distribution is then applied to the total population number of the whole country for the years 1995-2000, which is in fact published for these years by the statistics bureau. There is no national population data for 2014, and therefore the total population the data is extrapolated based on the average growth of the total population over the previous five years. Again the provincial population distribution for the five years before 2014 is calculated and applied to the extrapolated total population number for 2014. After all calculations are done the provinces Tibet and Chongqing are
excluded which creates a dataset of 29 provinces for a period of 20 years, which equals 580 observations and no missing data.
18
not an option as much of that GDP ‘contains’ goods that are produces for export out of the province and country. The data is then adjusted for inflation by using the dataset ‘consumer price index (preceding year 100)’ As a last step the data is divided by the ‘year end province population (times 10000)’ In order to come to per capita values. They are then transformed (multiplied by 100 million) to represents the household
expenditure per capita in Yuan’s (RMB).
Another important factor to include in the probit analysis is development; agricultural employment can be used as a proxy for development as in general a decreasing share of employment in agriculture indicates an increase in development. This follows the development theory that agricultural productivity is the key to
economical development. It builds on the principle that Countries experiencing
increases in agricultural productivity are able to release labor from agriculture into other sectors of the economy” (Gollins, Parente & Rogerson, 2002 p.163). Data shows that the new sectors, agricultural labor flows towards have a higher output per worker. So a shift from labor in agriculture towards other sectors increases overall productivity (Gollins, Parente & Rogerson, 2002). The dataset ‘number of employed persons in agriculture, forestry, animal husbandry and fishery’ is used a measure of agricultural employment. The dataset ‘year-end provincial population’ is used to calculate the agricultural employment as a share of the total provincial population.
For the probit analysis the dependent variable CONVERGENCEij is a dummy with a
value of 1 if the CPI of a province pair i and j is converging (when the null hypothesis for the unit root test is rejected) and 0 when there is no convergence (when the null
hypothesis is not rejected). This is based on the unit root results of the KSS unit root test using a 5 percent significance level.
19
DISTANCEij is the distance in kilometers between the capitals of a provincial pair.
POPULATIONi is calculated using the following formula:
((Year end provincial population at the end/ Year end provincial population at the
start)^(1/ (end year- start year))-1)*100 (1)
Then POPULATIONij is the absolute difference in population growth between province
pairs i and j:
POPULATIONij = |POPi- POPj| (2)
In order to come to a measurement of INCOMEi the same method is utilized:
((Household expenditure per capita in Yuan at the end/ Household expenditure per capita in Yuan at the start)^(1/ (end year- start year))-1)*100 (3) Then INCOMEij is the absolute difference in income growth between the province pairs:
INCOMEij= |INCOMEi-INCOMEj| (4)
Then again the same principle is applied in order to come to DEVELOPMENTij:
20
Then DEVELOPMENTij is the absolute difference in income growth between the
province pairs:
DEVELOPMENT ij= | DEVELOPMENTi- DEVELOPMENT j|
An overview of the summary statistics for the probit analysis datasets and correlation matrices, which indicates that there are no multicollinearity issues, can be found in appendix 3.
Chinese statistics are often criticized for their lack of accuracy. China being a country in development which is controlled by an authoritarian state receives doubts related to the degree of political will and institutional capacity to publish reliable
statistics, it is however argued that while the Chinese national bureau of statistics made landmark improvement, yet the level of (output) statistics is not on par with Europe and the USA (Koch-Weser, 2014).
Moreover Nakamura, Steinson and Liu’s (2014) find that Chinese inflation measures are ‘smoothed out’, meaning that during the high inflation in the 2000’s the authors find that their corrected estimate was significantly higher than the official
21
22
4
M
ETHODSApart from the initial work of Pesaran (et al.) (2007; 2009), this part of the research heavily builds on the adaptions of the pair wise approach of Holmes et al. (2011) and especially the pairwise method Ikeno (2014) utilized for Japanese price convergence. The pair wise unit root approach analyses the convergence of time series by testing for the null hypothesis of a unit root in the bivariate provincial CPI differentials (pij). If the
alternative hypothesis holds pij follows a stationary process, the price level in province i
and j then moves around a long term level which implies a relative purchasing power parity and this can be interpreted as price convergence (Ikeno, 2014). As mentioned the main benefit of the pair wise approach is that there is no requirement for a benchmark province. Pi,t denotes the Consumer price index in province i at time t.
The relative price indices in provinces i and j at period t expressed in logarithms then are
!!" ≡ ℓ! !!"
!!" ! = 1,2, … , !; ! = 1,2, … , !. (6)
Formula 7 measures the fraction of provincial pairs (N*N-1)/2 or with our data (29*28)/2=406 for which the null hypothesis is rejected based on the used significance level. !!" = ! ! !!! !!",! ! !!!!! !!! !!! , (7)
Zij,T = 1 if the null hypothesis of the unit root of pij can be rejected for the signifance level
23
In order to calculate ZNT and in line with Ikeno’s (2014) method the following
univarate unit root test are utilized: The augmented Dicky Fuller test (ADF test), the generalized least-square detrending test (ADF-GLS) and the non-linear unit root test (KSS test), the significance levels are set to 5- and 10 percent. The Aikaike Information criterion (AIC) determines the number of lags, and the maximum lag is set to 5.
To analyze the causes of convergence or a lack thereof on provincial pairs the following probit model is utilized:
CONVERGENCEij =α + β1DISTANCEij + β2POPULATIONij+ β3 INCOME + β4 DEVELOPEMENT
(8)
24
5 Pairwise unit root tests results
The pair wise root tests of which the results are displayed in table 2 wich evaluate the proportion of Chinese provincial pairs that exhibit converging prices.
Table 2: Pairwise unit root test, Znt (in percent) for all samples at the 5% and 10% significance level
Sample period Significance Level ADF GLSDF KSS
1995-2001 (Pre-WTO) 0.05/5% 31.03 35.22 24.88 0.1/10% 46.06 51.48 33.25 2002-2014 (During WTO) 0.05/5% 78.81 79.55 53.45 0.1/10% 88.17 88.18 65.27 1995-2004 (First Half) 0.05/5% 54.67 43.1 39.16 0.1/10% 70.93 55.42 44.09 2005-2014 (Second Half) 0.05/5% 66.26 75.86 41.38 0.1/10% 80,79 83.99 55.91 1995-2014 (Complete) 0.05/5% 96.31 52.96 68.72 0.1/10% 97.78 60.34 76.11
In addition to the whole sample ranging from 1995-2014, several sub sets are tested. First in order to test the influence of the increased openness of the Chinese economy, we selected China’s membership of the World Trade Organization (WTO) as a turning point. In December 2001 China joined the WTO. Therefore the whole sample is
separated in two time periods namely 1995-2001 (Pre-WTO) and 2002-2014 (During WTO). However as this sample is not of two equal time periods, the sample has also simply been separated in two equal parts namely 1995-2004 as well as 2005-2014.
25
even close to 100 percent for the augmented Dicky-Fuller test (ADF). This indicates that over a time frame of 20 years prices in most, or at least more than half, of the provinces in China are converging.
26
6 Probit model
As stated in the introduction it is of particular interest to investigate the potential spatial component of pair wise intra-national convergence. To test for this and other factors a probit model is estimated. A negative relationship between distance and convergence is expected as spatial segregation of two locations increases the transaction costs
associated with trade and thereby negatively affects the opportunity of arbitrage, which is so important for price convergence.
Based on the Balassa Samuelson/Penn effect, which is already identified as a potential driver of divergence in chapter 2, it is expected that an increased income as proxied by increased spending result in higher prices (Ikeno, 2014). Therefore the differential of income growth for province pairs is selected as another potential driver of divergence. Timmer and Inklaar (2012) explain the effect further “Rising income is
related to technical change that leads to higher productivity in the tradable goods sector. This leads to higher overall wages, driving up prices in the non-tradable services sector that has much lower productivity growth”.
Population growth might have an influence on price convergence as it may put pressure on the housing market and thereby drive residential and commercial rent upwards, following from this increase rent prices of commercial products and services increase as well (Ikeno, 2014). Rising differences in population growth may therefore diverge the price levels of two geographical areas.
27
service sector, as well as other non-tradable good sectors. Which again relates to the Balassa-Samuelson effect and increase the price level in the province.
Table 3: Result probit estimation
Probit 1995-2004
Variable Coefficient Standard Error
DISTANCEij -0.00009 0.00010 POPULATIONij 0.07217 0.15368 INCOMEij -0.09404* 0.02946 DEVELOPMENTij 0.00354 0.03818 CONSTANT 0.04737 0.17610 Probit 2005-2014 DISTANCEij -0.00010 0.00009 POPULATIONij -0.12646 0.08502 INCOMEij -0.07656 0.07707 DEVELOPMENTij 0.01605 0.05201 CONSTANT 0.09544 0.17460 Probit 1995-2014 DISTANCEij -0.00018** 0.00010 POPULATIONij -0.17676 0.13037 INCOMEij 0.07838 0.07365 DEVELOPMENTij -0.05842* 0.02847 CONSTANT 0.92454 0.17734
Note: * = Significant at 5% level, ** = significant at 10% level
Table 3 shows the results of the probit test, three of the samples have been used, namely first of all the complete sample, which again ranges from 1995-2014 as well as the sub-samples 1995-2004 and 2005-2014. Before analyzing the results, based on the arguments above it is expected that all signs are negative as spatial differences and differences in income-, population- and development growth rate intuitively should decrease the propensity of price convergence and in fact do the opposite, which is increase the change of divergence between provincial pairs.
28
development is in fact significant at the five percent level. Income has an unexpected positive sign but is insignificant
For the 1995-2004 sample the results are different, income is in fact the only significant effect in the first ten year of the sample period, and it has the expected
negative sign. Distance does have the expected sign but it is not significant, income and development neither have the expected signs nor are they significant.
The last sample, which describes the drivers of convergence for the year 2005-2014 shows that all the variables except development have the expected negative signs, but none of the variables are in fact significant at the five- nor ten percent level.
In order to analyze the actual magnitude of the effects of the variables on the propensity of provincial price convergence the average marginal effects are estimated, these results can be found in table 4 and display the marginal effect the variables have on the likelihood of success, in this case convergence.
Table 4: Marginal effect probit estimation
Variable Coefficient Standard Error
Marginal effect 1995-2004 DISTANCEij -0.000034 0.000036 POPULATIONij 0.026960 0.057372 INCOMEij -0.035132* 0.010585 DEVELOPMENTij 0.001321 0.014263 Marginal effect 2005-2014 DISTANCEij -0.000040 0.000036 POPULATIONij -0.048744 0.032492 INCOMEij -0.029510 0.029589 DEVELOPMENTij 0.006188 0.020041 Marginal effect 1995-2014 DISTANCEij -0.000062** 0.000034 POPULATIONij -0.061201 0.044877 INCOMEij 0.027136 0.025410 DEVELOPMENTij -0.020227* 0.009711
29
For the years 1995-2014 there are two significant drivers of divergence, indicated by the asterisks. First of all an increase of the development rate differential of a
provincial pair of one percent point negatively affects the chance of convergence with 2 percent. The other significant variable for these years is distance; every marginal 1000-kilometer of distance between two provincial capitals negatively influences the
probability of convergence between those provinces with 6.2 percent. Given that China is a very large country and the average distance between two provincial capitals in China is 1,326 kilometers it can be concluded that next to development, distance is an important driver for divergence.
When looking at the two other time frames there are different results, a one percent point increase in the difference between the income growth rates of two
provinces had a negative effect of 3.5 percent on the propensity of convergence during the years 1995-2004. The other variables in the 1995-2004 sample are not significant. Furthermore as concluded earlier, for the years 2005-2014 none of the estimated variables actually had a significant influence on price convergence.
Table 5 then contains the Wald test, where all tested variables are included. If the Wald null hypothesis holds, it means that all of the tested variables fail to affect the probability of convergence. (Adkins & Hill, 2011). The Wald test for the full sample and 1995-2004 sample can be rejected at the 5 percent significance. The 2005-2014 sample is not significant, which corresponds with the lack of significant estimators for those sample years in the probit test results.
Table 5: Wald tests for probit estimation
Sample Wald statistics (Chi2) Significance
1995-2014 10.81 0.0288
1995-2004 12.39 0.0147
30
7 Conclusion
This paper analyzed market integration of Chinese provinces. Generally it can be concluded that based on pairwise unit root tests on a sample of 29 provinces over a period of 240 months there is a high degree of price convergence based on the ADF test while the KSS and GLS-DF are less in favor of convergence. When looking at the full sample there is convergence ranging from about 50 percent of the provincial pairs in the GLS-DF test to near 100 percent of the pairs in the ADF test. The results are not completely in line with most of the other research on convergence in China, which found short half-lives and high convergence rates (Ritola, 2008; Lan and Sylwester, 2010; Fan & Wei, 2006).
Howeover in the last ten years price convergence and thereby market integration considerably increased compared to the ten years before. This gives an indication that among others the openness, modernization policies, economical development and growth as well as the WTO membership of China increased the integration of the Chinese internal market.
With regard to the causes of convergence the claim that there is an inverse relation with distance can be confirmed for the 1995-2014 sample, moreover differences in development as proxied by agricultural employment also increase the likelihood of divergence. This means that two provinces that follow the same development path are more likely to convergence. This combined with the increase in convergence in the last ten year provides further reason to argue that in fact development has a strong
influence on convergence on a subnational level in China.
In the subsample which ranges from the years 1995 until 2004, the only
31
that the effect of income growth differentials dissipated when Chinese development increased in the last ten year.
With regard to future research it is the aim of the author to continue to work together with Dr. T.M. Harchaoui and attempt to publish the article. Adding
bootstrapping methodologies, with the assistance of Dr. H. Ikeno of Surugadai University, will make improvements in the robustness of the pairwise unit root tests.
Other opportunities for future research are to empirically test the relationship between increased openness of the Chinese market to the world and price
convergence. Moreover China could be split up in several geographical areas to test if the convergence rates are higher in for example the eastern coastal provinces. Another interesting approach is to use the same methodology on both developed- and
32
References
Adkins, C. & Hill, R. 2013. Using stata for principles of economics fourth edition, Wiley. Boston Consulting Group. 2012. The age of the affluent. Accessed 08-05-2015.
Available at:
http://www.bcg.com.cn/export/sites/default/en/files/publications/reports_pdf/BCG_The_ Age_of_the_Affluent_Nov_2012_ENG.pdf
Cecchetti, S. Sonora, M. & Price, R. 2002. Index Convergence Among United States Cities. International Economic Review, 43(4):1081-1099
Cheung, Y. & Lai, K. S. 2000. On cross-country differences in the persistence of real exchange rates. Journal Of International Economics, 50(2), 375-397.
Cuestas, J. C., & Ordonez, J. 2007. Testing for price convergence among Mercosur countries. Working Paper, 2007/1, Nottingham Trent University.
Das, S. & Bhattacharya, K. 2008. Price convergence across regions in India. Empirical
Economics, 3,(2):299-313.
Dreger, C., & Kosfeld H. 2010. Do Regional Price Levels Converge? Jahrbücher für
Nationalökonomie und Statistik / Journal of Economics and Statistics
33
Egger, P. Gruber, S. & Pfaffermayr, M. 2009. The law of one price: conditional convergence evidence from disaggregated data. Applied Economics, 41(26), 3345-3357
Fan, Y.,& Liao. H, Y. 2007. Can market oriented economic reforms contribute to energy efficiency improvement? Evidence from China. Energy Policy, 35 2287–2295.
Fan, C., & Wei, X. 2006. The law of one price: evidence from the transitional economy of china. Review of economics & statistics, 88(4):682-697.
Frankel, J., & Rose, A. 1996. A panel project on purchasing power parity: Mean reversion within and between countries. Journal Of International Economics, ;40(1/2):209-224.
Chow, G. C. 1993. Capital formation and economic growth in China. Quarterly Journal
Of Economics, 108(3), 809.
Gluschenko, K. 2011. Price convergence and market integration in Russia. Regional
Science & Urban Economics. 41(2):160-172.
Horvarth, J., & Vidovic, S. 2004. Price variability and the speed of adjustment to the law of one price: Evidence from Slovakia. BOFIT Discussion Paper, Bank of Finland, 3: Holmes, M. Otero, J. & Panagiotidis, T. 2011. Investigating regional house price convergence in the United States: Evidence from a pair-wise approach. Economic
34
Inklaar, R. & Timmer, M. The relative price of services. Research Memorandum
Groningen Growth and Development Centre. Available at:
http://www.ggdc.net/publications/memorandum/gd124.pdf accessed 18-06-15 Lee, S,. Woo,. & K. Yeung, R. 2011. Impact of the Financial Tsunami on the
Convergence of the Consumer Price Index in China. Chinese Economy 44(3):71-83. De Masi, P. 1996 Relative price convergence in Russia. IMF Staff Paper
Available from: Business Source Premier, Ipswhich, MA. Accessed March 14, 2015. Parsley, D,. & Wei, P. 1996. Convergence to the law of one price without trade barriers or currency fluctuations. Quarterly Journal Of Economics, 111(4):1211.
Pesaran M. 2007. A pair-wise approach to testing for output and growth convergence.
Journal Of Econometrics. 138(1):312-355.
Pesaran, M. Smith, H. Yamagata, T., & Hvozdyk, L. (2009). Pairwise Tests of Purchasing Power Parity. Econometric Reviews, 28(6), 495-521.
Poncet, S. 2003. Measuring Chinese domestic and international integration. China
Economic Review 14(1)
Young, A. 2000. The razor’s edge: distortions and incremental reform in the People’s Republic of China. The Quarterly Journal of Economics CXV, 1091–1135.
Koch-Weser, I. 2013. The Reliability of China’s economic data: An analysis of
national output. U.S.- China economic and security review commision staff research
35
Lan, Y. & Sylwester, K. 2010. Does the law of one price hold in China? Testing price convergence using disaggregated data. China Economic Review 21(2), 224-236. Lee S, Woo K, Yeung R. 2011. Impact of the Financial Tsunami on the Convergence of the Consumer Price Index in China. Chinese Economy 2011;44(3):71-83.
Nakamura, E. Steinsson, J,. & Miao, L. 2014. Are chinese growth and inflation too
smooth? Evidence from engel curves. NBER Working Paper NO. 19893 Available at
http://www.nber.org/papers/w19893
Ratfai, A. (2006). How Fast Is Convergence to the Law of One Price? Very (When Price Differentials Are Properly Measured). Economics Bulletin, 6(10), 1-12.
Ritola, M. 2008. Price convergence and geographic dimension of market integration: Evidence from China. BOFIT Discussion Papers 13. available at
http://www.suomenpankki.fi/bofit/tutkimus/tutkimusjulkaisut/dp/Documents/dp1308.pdf Rogers, J. 2001. Price Level Convergence, Relative Prices, and Inflation in Europe.
Working Papers -International Finance Discussion Papers Number 699. available
at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=266505 accessed 13-03-2015 Rogoff, K. 1996. The Purchasing Power Parity Puzzle. Journal Of Economic
Literature,34(2), 647-668.
Sonora, R. J. 2005. City CPI Convergence in Mexico. Review Of Development
36
Suliman, M. 1998. China's Transition to a Socialist Market Economy. Westport, Conn: Greenwood Publishing Group.
Tageo (2015) Chinese city and town population Availble at:
http://www.tageo.com/index-e-ch-cities-CN.htm accessed: 21-05-2015
The Econonomist. 2014. An alternative view. The Economist. Accessed 12-05-15. Available at: http://www.economist.com/news/finance-and-economics/21599051-chinas-official-figures-both-understate-and-overstate-inflation-alternative-view
Wimanda, R. E. (2009). Price variability and price convergence: Evidence from Indonesia. Journal Of Asian Economics, 20(4), 427-442.
Xu, Z. & Fan, J. 2012. China's Regional Trade and Domestic Market Integrations. Review Of International Economics, 20(5), 1052-1069.
Yin-Wong, C. & Lai, K. 2000. On cross-country differences in the persistence of real exchange rates. Journal Of International Economics, 50(2), 375-397.
We, W. 1999. Special Economic Zones and the Opening of the Chinese Economy: Some Lessons for Economic. World Development, 27(7), 1267.
Gollin D, Parente S, &Rogerson R.2002. The Role of Agriculture in Development. American Economic 92(2):160-164.
Zhang, F., & Morley, B. 2014. The convergence of regional house prices in China.
37
Appendix 1: Data sources
National Bureau of Statistics China, 2015. Per capita GDP 1995-2014 available at: http://data.stats.gov.cn/english/easyquery.htm?cn=E0103. Accessed: June 2015
National Bureau of Statistics China . Employment in Agriculture 1995-2012 Available at:http://data.stats.gov.cn/english/easyquery.htm?cn=C01.
Accessed June 2015
National Bureau of Statistics China . Consumer Price Index (same month last year =
100). available at: http://data.stats.gov.cn/english/easyquery.htm?cn=A01.
Accessed: April 2015
National Bureau of Statistics China . Resident population. available at:
http://data.stats.gov.cn/english/easyquery.htm?cn=E0103. Accessed: June 2015
National Bureau of Statistics China . Household Consumption Expenditure. available at:http://data.stats.gov.cn/english/easyquery.htm?cn=E0103. Accessed: June 2015 National Bureau of Statistics China . Consumer Price Index (preceding year=100) available at: http://data.stats.gov.cn/english/easyquery.htm?cn=E0103. Accessed June 2015
National Bureau of Statistics China . Number of employed persons in agriculture,
forestry, animal husbandry and fishery (10000 persons). Available at:
38
National Bureau of Statistics China. Total Population (year-end)(10000 persons). available at: http://data.stats.gov.cn/english/easyquery.htm?cn=C01 accessed: July 2015
Tables and figures
Figure 1 based on: National Bureau of Statistics China, 2015. Per capita GDP
1995-2014
Figure 2 based on: National Bureau of Statistics China, 2015. Employment in
Agriculture 1995-2012 / National Bureau of Statistics China. Total Population (year-end)(10000 persons).
Figure 3 based on: National Bureau of Statistics China, 2015. Consumer Price Index
(same month last year = 100)
Table 1: based on: National Bureau of Statistics China, 2015. Consumer Price Index
39
Appendix 2: Half-lives of selected studies
40
Appendix 3: Probit summary statistics and correlation matrix
Table 7: Summary statistics probit estimation
(1) (2) (3) (4) (5)
VARIABLES N mean sd min max
FSPOPULA TIONIJ 406 0.488 0.427 1.09e-05 2.112 SSPOPULA TIONIJ 406 0.922 0.889 0.000608 3.719 CSPOPULA TIONIJ 406 0.553 0.506 1.09e-05 2.311 FSINCOMEI J 406 2.743 2.567 0.0189 14.60 SSINCOMEI J 406 1.144 0.884 0.00578 4.628 CSINCOMEI J 406 1.269 0.948 0.00622 5.077 KMIJ 406 1,326 689.8 112.5 3,507 FSAGRICU LTUREIJ 406 2.099 1.721 0.000784 9.816 SSAGRICU LTUREIJ 406 1.651 1.445 0.000110 8.538 CSAGRICU LTUREIJ 40 3.255 2.405 0.00727 11.82
Note: CS stands for Complete sample (1995-2015), SS stands for Second Sample (2005-2014) and FS is the First Sample (1995-2004)
Table 8 Correlation matrix for all samples probit estimation
1995-2004
41 POPULATIONIJ 1 INCOMEIJ 0.1644 1 KMIJ -0.0238 0.1291 1 AGRICULTURIJ -0.1328 0.1536 0.1354 1 2005-2014
POPULATIONIJ INCOMEIJ KMIJ AGRICULTURIJ
POPULATIONIJ 1
INCOMEIJ 0.2792 1
KMIJ -0.0848 -0.0604 1
AGRICULTURIJ 0.5009 0.2744 0.0183 1
POPULATIONIJ INCOMEIJ KMIJ AGRICULTURIJ
POPULATIONIJ 1
INCOMEIJ 0.0631 1
KMIJ -0.0402 0.2953 1
