1
Bachelor Thesis
The Compatibility of Balassa-Samuelson Hypothesis on Chinese
Economy
Author:
Jianfeng Zhou 10495010
Supervisor:
Chan, Stephanie
A thesis submitted in fulfillment of the requirements for the degree of
Bachelor of Economics and Finance in Economics and Business
15 July 2016
Abstract
The Balassa–Samuelson effect is employed to test the feasibility of the hypothesis on Chinese economy. The pure Balassa–Samuelson model is proposed to accommodate seventeen different countries and twelve pro vinces in China. The model includes the Balassa–Samuelson effect, the production approach of GDP measurement and the basic production function. Data involved in the research covers the period from 1980 to 2014and the results proved a long term relationship between the Balassa–Samuelson hypothesis and Chinese economy. However, the recession in 2008 has turned down the consistency of hypothesis and the regional difference around China weaken the predictability of the Balassa–Samuelson model on the entire Chinese economy.
2 Statement of Originality
This document is written by Student [Jianfeng Zhou 10495010] who declares to take full responsibility for the contents of the document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 1. Introduction
Economists have long sought to explain exchange rate fluctuations and they have done so with varying degrees of success. Purchasing power parity (PPP), the law of ‘one price’, along with a perfect world economy, predict that relative prices will be equalized across the globe. However according to Taylor, more recent work using time series analysis of short spans of data has ‘led many to conclude that purchasing power parity failed to hold’ (Taylor, 2002, p. 139).
Another most popular hypotheses against the absolute form of PPP on explaining the long-term real exchange rate movements is called the Balassa-Samuelson hypothesis. Independently advanced by Balassa and Samuelson, the hypothesis suggest that the purchasing power parity only holds for the traded sector of the economy and it is often impossible for it to hold for the non-traded sector (Balassa, 1964, p. 584−596). As a result, the greater the productivity difference in the traded sector, the greater the gap of real exchange between countries.
Since the differences in productivity increase are expected to be larger in high- growth countries, the Balassa-Samuelson prediction should be most visible among rapidly growing countries. Given this, it is intriguing to note that over the period of 1980 to 1994, the real price of Chinese RMB has depreciated to only around one fourth of its initial value (Fig. 1) despite the expansion of Chinese real GDP of nearly four times (Fig. 2). Under the standard prediction of the Balassa–Samuelson hypothesis, Chinese real exchange rate should have appreciated tremendously. However, since the real exchange rate has actually depreciated, it means that over the period from 1980 to 1994, Chinese exchange rate has evolved in a manner which is incompatible with the Balassa–Samuelson hypothesis.
The inconsistency of Balassa-Samuelson hypothesis on Chinese economy drew the attention of economists around the globe and numerous research conducted to assess the compatibility has been processed. Inspiring as they may be, these research are mainly processed at the end of the 20th century and they mainly focused on the period when Chinese currency contradict with the prediction of Balassa–Samuelson
4
hypothesis. As the trend of exchange depreciation converted in 1995 (Fig. 1), few literatures have been found discussing the relationship between the Balassa-Samuelson hypothesis and Chinese economy.
Fig. 1. Real effective exchange rate index of real currency value (2010 = 100) Sources: The World Bank (WB)
Fig. 2. Chinese Real GDP (1980 = 100) Sources: China Statistical Yearbook (2015)
0 50 100 150 200 250 300 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 0 500 1000 1500 2000 2500 3000 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
5
However two decades later, China has grown into the world's second largest economy. Chinese real GDP has been nearly five times larger than that in 1995 whereas the real value of Chinese currency has only doubled. It seems that Balassa-Samuelson hypothesis is still not an appropriate estimator for China. But from both positive (descriptive) and normative (policy prescription) perspective, an accurate prediction on Chinese economy is needed. Common investors can't overlook the world's second largest market full of chance and opportunities, politicians under the inevitable trend of globalization require further understanding of world economy for policy setting, central banks need to forecast foreign exchange to hedge domestic currency and economists need more data to come up with a solution for the world economic recession broke out in 2008.
Given this, it is now high time to rethink about the Balassa-Samuelson hypothesis and my principal purpose here is to assess the extent to which Chinese economy have been compatible with the Balassa–Samuelson hypothesis over the period. During the entire research, annual measures are employed and the pure model of the Balassa-Samuelson hypothesis is implemented on the correlation between the productivity increase in tradable sector and the changes in real exchange rate. Enlighten by empirical tests conducted by previous researchers, a continuous time series as well as nation differences are also taken into account. Internal factors such as inflation are also fully considered during the research and province behaviors within China are analyzed for further accuracy.
The remaining sections of the paper are organized as follows. Section 2 review the basic concept of Balassa-Samuelson hypothesis. Section 3 summarize previous research with similar focus. Section 4 provide instruction for the pure Balassa-Samuelson model. Section 5 describe the methodology conducting in the research. Section 6 gather the data and present the way the variables used in our empirical analysis are determined. Section 7 summarize our empirical evidence and section 8 concludes the paper.
6
2. The Basic Concept of Balassa-Samuelson hypothesis
The Balassa-Samuelson hypothesis refers to the phenomenon that countries with higher growth in productivity faces a higher price increase. It uses part of the preconditions of purchasing power parity (PPP) and in turn explains why the pure PPP fail to hold in current global economy. Although Balassa-Samuelson hypothesis is only a simplified model of real economy with many previous assumptions, it is still currently the most popular and widely accepted theory on exchange rate determination.
The three preconditions involved in the hypothesis are listed as follows.
(1) The typical economy is comprised of ‘tradable’ (T) and ‘non-tradable’ sector (N). The tradable sector (T) consists of industries like agriculture, manufacturing and mining where outputs can be cheaply and easily transported across international markets. Against this, the non-tradable sector (N) is comprised of immovable property, perishable goods and services like hospitality and tourism and these are much less easily traded across international barriers (Balassa, 1964, p. 584−596).
As a result, the purchasing power parity (PPP) only holds in the tradable sector (T) and due to the international barriers in transportation, it is not feasible for the non-tradable sector (N).
(2) Perfect labor mobility under the condition of small and open economy, which leads to the same wage level (W) between tradable (T) and non-tradable (N) sector and can be written as
(1) representsthe wage level in the tradable sector, is the symbol of wage level in the non-tradable sector and the equation can be explained by the fact that any wage change in one side can lead to job transfer between the two sectors. For the sector with higher wage level, an excessive labor supply is expected and the wage payment in this sector will decrease under market mechanism. The same reason also holds for the sector with lower wage level and the adjustment continues until wage level between the two sectors goes back to equilibrium.
7
However, the equation mentioned above doesn't hold for large and planned economy. Under large economy, perfect labor mobility can't be guaranteed since migration comes at a cost for each labor and the wage difference may not compensate the entire cost of job transfer. Whereas under planned economy, jobs are allocated rather than self chosen and labor mobility is impossible in the first place. As a result, in order to maintain the precondition of perfect labor mobility, small and open economy should also be involved as part of the assumption.
(3) All labor production (Q) are quantified and under the payment model, the price of production (P) is positively correlated with the wage level (W), negatively correlated with labor productivity and can be written as
(2) Therefore the wage (W) received by labor is the total value of production, a product of labor production (Q) and price of production (P).
With the assumptions mentioned above, the following equations are derived. According to assumption (3), for poor countries,
(3) For wealthy countries,
(4) Among the equations above, represents the price of non-tradable goods, represents the price of tradable goods, represents the labor productivity in the non-tradable sector and represents the labor productivity in the tradable sector. Star (*) marks the economy of wealthy countries with higher productivity in tradable sector.
Since assumption (2) holds for the Balassa-Samuelson hypothesis,
(5) The productivity of tradable goods is higher in wealthy countries but for non-tradable goods, the productivity are the same for both poor and wealthy countries. Which means
8
(6) Finally, as it is mentioned in assumption (1) the purchasing power parity (PPP) only holds in the tradable sector and
(7)
S represents exchange rate in equation (7) (Taylor, 2002, p. 139) and for every
economy, the price ratio ( ) for non-tradable and tradable goods are
(8)
and
(9)
Since wealthy countries have higher productivity in the tradable sector, the price ratio in the wealthy country is higher, leading to .
Repeat equation (8) and rewrite equation (9) and it comes to
(10)
and
(11)
Since for the purchasing power parity (PPP), as , (Pilbeam, 2013, p.133-134). The price of non-tradable goods in wealthy countries are higher than that in poor countries, leading to a higher overall price level across wealthy nations.
3. Review of related Literature
After a carefully examine of the basic idea for the Balassa-Samuelson hypothesis, it is time to come up with a way to carry on the research. As it is widely accepted, the starting point of new research usually lies on similar researche conducted before. This will help researchers to avoid detours or deficiency, which may take a long time to find out if the field of research is completely blank. Except for the conclusion, the idea and method leading to the conclusion are also of great help as a reference and my current research between Balassa-Samuelson hypothesis and Chinese economy is
9 greatly inspired by both of them.
The article that inspire me most is an overview of the Balassa-Samuelson hypothesis in Asia. It explored the relationship of Balassa-Samuelson hypothesis and the economy of APEC countries including China over a twenty year period from 1970s to 1990s. Across the research, growth rate, machinery export, export characteristics, stage of economic development and productivity growth for each country are selected and in order to fully examine the compatibility of the Balassa-Samuelson hypothesis on fast growing econo my, both the real exchange rate and inflation are referred to for accuracy. With all the data mentioned above, the author set up a variety of scattered diagram to reveal the relationship between the index and the coefficient correlation between the index were also considered for further accuracy. The results of these empirical research proved the feasibility of the Balassa-Samuelson hypothesis on fast growing economy. However, even if the economy is growing fast, Balassa-Samuelson may not be applicable if the economy has just come out of the primary goods exporter or planned economy phase (Ito, Isard, and Symansky, 1999, p. 109 - 132).
The inspiration of the essay above comes from four specific points. To begin with, the literature revealed the inaccuracy of the prediction by the Balassa-Samuelson hypothesis on planned economy. Since China has only adopted open policy in 1978, it still remains a question whether China has come out of the planned economy phase and to what extent it has come to. As a result, the predictability of Balassa-Samuelson hypothesis for specific years, especially for the first few years of economic reform in China, can't represent the overall compatibility of the hypothesis on Chinese economy and the results of the empirical research may vary across the year due to the same issue. In face of condition like this, a long term relationship between the Balassa-Samuelson hypothesis and Chinese economy should be examined within the research and by doing so, the instability as well as inaccuracy of the prediction affected by factors such as politics and global economy can be minimized.
10
the real exchange rate and price index should be considered. There are many other essays studying the Balassa-Samuelson hypothesis as well and they either focused on real exchange rate or inflation, a measurement of price index. The essay above however, took both directions into account during the empirical research and it has sufficient reasons for doing so. As it is concluded in section 2, , the price of non-tradable goods in wealthy countries become higher than that in poor countries and the overall price index in wealthy countries increase. In order to stabilize the price level of domestic goods, instead of increasing the face value of domestic product which is definitely not preferred by the residents, the value of domestic currency increases and it is shown as a decrease in real exchange rate (Balassa, 1964, p. 584−596).
However, the motion mentioned above only apples to countries with perfect floating exchange rate and on the other occasions, such as fixed exchange rate, it is the increase in face value of domestic goods that covers the entire productivity growth and the real exchange rate is not affected. So it comes to a conclusion that in order to fully test the capability of Balassa-Samuelson hypothesis on real economy, both the decrease in real exchange rate and increase in price index should be considered since global currencies are neither perfectly floating nor fixed and the loss of efficiency on one side can be a tradeoff of an enhance on the other side.
The third inspiration given by the essay comes from the fact that during the empirical research on the Balassa-Samuelson hypothesis, scattered plots as well as correlation tests are performed and the data used for comparison were structured in the form of percentage change. To be more specific, although much effort has been conducted by economists on model construction , up till now, no model has been commonly proven and accepted as the standard model for analyzing Balassa-Samuelson hypothesis on world economy. In face of this, previous researchers all structured models specially designed for their own studies to come up with the statistic results, such as the three-goods model and the black market model which will be discussed later. But the models constructed before were specially
11
designed and it is impossible to identify the best among them since they are never of the same standard. As a result, it is reasonable and acceptable to design a new model for my own study and two other factors should also be considered.
On the one hand, as it is highlighted in section 2 and the literature above, Balassa-Samuelson hypothesis focus on the relationship between productivity increase and real exchange rate increase (or increase in price index). Since all three variables tested in the hypothesis are measured in changes of ratios, the empirical test as well as the new model of my research should focus on the same perspective. On the other hand, the scattered diagram, together with correlation test used on APEC countries should be implemented during my empirical test since both methods have the ability to narrow the objective down to two dimension, which is applicable with the structure of the Balassa-Samuelson hypothesis and it is easier for observation and comprehension.
The last specific point highlighted in the essay informs me the importance of comparison. As it is mentioned before, a variety of countries in APEC were involved in the empirical test and the final conclusion of the essay came from the comparison between the statistic results. The reason for multinational comparison is obvious. As it is stated in the previous paragraph, previous researchers all structured models specially designed for their studies to come up with statistical results. However, since the model is specially designed for the research, it is completely new from every perspective and the feasibility of the model still remains to be seen. To be more specific, if only a single country is tested using the model, the statistical result is nothing but a number. It is hard to explain the number since there are no other bench marks to refer to and the performance of the model can't be fully illustrated by one data in the first place. Therefore, only when a variety of nations are implemented using the model can the efficiency of the model be tested. At the same time, the relationship between the results can reveal economic content within and between different economy and they are of great value to the result of the empirical research.
12
Jeanneney and Hua on the one hand, focused on the inflation differences between the Chinese provinces by using a self constructed "three-goods" model. The author divided goods into three sectors rather than two, namely tradable, semi-tradable and non-tradable and compared the estimated result with real historic data. In their conclusion, the econometric results cannot refute the hypothesis that the Balassa–Samuelson effect explains durable differences in inflation among provinces (Jeanneney and Hua, 2002, p. 134). Guo, on the other hand, focused on the effect of each province on the overall exchange rate of China. To be more specific, the author took black market exchange rate into account and performed both unit root and cointegration test. At the end of the literature, the results of the tests showed a strong long-run relationship between the real exchange rate and the relative productivity differential for the Chinese economy and the black market exchange rate appears to be more consistent with the predictions of the Balassa–Samuelson model than the official exchange rate (Guo, 2010, p.334).
The intuition within the two research above is that pure Balassa-Samuelson hypothesis is base on the background assumption of small and open economy introduced in section 2, and as the third largest country in the world, China is definitely not the case. So the analysis of Chinese economy using the Balassa-Samuelson should be divided into two steps. The first step is to test Chinese economy as a whole and if the prediction is inconsistent with other nations using the same model, an analysis within China should be processed to explain the difference together with other internal or external factors.
4. The pure Balassa-Samuelson Model
As it is discussed within the previous section, since it is impossible to identify the best model among previous research and there are no standard model for empirical test on the Balassa-Samuelson hypothesis, I set up a new model of my own to continue the research and the new model specially designed for my study is called the pure Balassa-Samuelson model. This is due to the fact that the model include only the
13
productivity increase in the tradable sector as the independent variable just like the pure Balassa-Samuelson hypothesis and no other external or internal factors are included within the model.
For the logicality of the pure model, changing ratios concerned in the previous chapter are taken in account and two assumptions should be highlighted before the construction of the model. The first assumption is the production function between total production and unit productivity and the second assumption is the production approach for GDP measurement.
As it is introduced in section 2 and 3, the Balassa-Samuelson hypothesis is a model explaining the relationship between productivity increase in tradable sector and the increase in real exchange rate (or the increase in price index). However, when the model is implemented on empirical tests, a barrier occurs for the comparison between the data since unit productivity is only a microeconomic concept for labor and both the real exchange rate and the price index are macroeconomic concepts on national bases. The two concepts can't be compared because they are of different economic scope and to solve the problem, the production function (Mankiw, 2009, p.52) between total production and unit productivity is required and it is written as
(12) represents unit productivity, represents total production, represents total labor force and the equation shows that total production ( ) is the sum of productivity ( ) of all labor force ( ) and the equation holds for both tradable and non-tradable sector of the economy.
However a relationship between micro and macroeconomic concept is still insufficient for the pure Balassa-Samuelson model. Even if the micro concept of productivity has been transformed into production, the measurement of production itself is still a problem for macroeconomics. To be more specific, a bench mark is needed to evaluate the production level between different industries and the only available standard right now is the measurement of value. In situation like this, the production approaches for GDP measurement is introduced as the second assumption
14
and the production level can now be transformed into an expression of production value. As it is shown in equation (3) of section 2,
(13) Rewrite the formula and we have
(14) According to the production approach of GDP measurement, GDP measures the total value of currently produced goods and services (Mankiw, 2010, p. 21). Since the total production is divided into tradable and non-tradable goods within assumption (1) of section 2, the total value of GDP is the total value of production for both tradable ( ) and non-tradable ( ) sector, written as
(15)
With two assumptions mentioned above, the following equations can be derived. According to the pure Balassa-Samuelson hypothesis described in section 2, only the productivity increase in the tradable sector is considered within the model and
(16) is the production value of the tradable sector and as is stated in the previous section, the Balassa-Samuelson hypothesis focus on the relationship between different changing ratios. So equation (16) is derived into equation (17) below.
(17) is the increase in GDP for tradable industry, is the increase in price for tradable goods and is the increase in production for tradable sector. The equation comes to a finding that the increase in GDP for tradable sector is consisted of two parts, the price increase in the tradable product and the enhance in production within the tradable department. As purchasing power parity (PPP) only holds for the traded sector of the economy according assumption (1) for section 2, the price of tradable product remains constant and the increase in price equals to zero. which can be written as
15 Plug (18) into (17), and it comes to
(19)
The GDP increase in tradable sector now completely rely on the production increase of the sector and further analysis of is now required. As it is assumed in assumption (1) of this section, the production function between total production and unit productivity is written as
(20) Rewrite equation (20) and it comes to
(21) Since the formula holds for both tradable and non-tradable sector of the economy, (22) is the labor productivity in tradable sector and is the total labor within the tradable industry. The derivation of formula (22) is written as
(23) In equation (23), is the increase in unit productivity for labor in the tradable sector and is the increase in total labor in the tradable department. According to the formula, the increase in production of the tradable sector is consisted of unit productivity and labor increase within the industry and since formula (19) and (23) are equal on one side of the equation,
(24) Rewrite equation (24) and it comes to
(25) The productivity increase in the tradable sector can now be calculated using the difference between the GDP increase for tradable industries and labor increase in the tradable departments. In order to implement the empirical test using the pure Balassa-Samuelson model, numerous data and calculations are required.
16
In order to apply the pure Balassa-Samuelson model on Chinese economy, according to the hypothesis, tradable and non-tradable sectors should be diversified in the first place. But in reality, most of the goods are tradable to some extent, except for parts of the real estate and service industry. So in the empirical analysis of the national economy, it is difficult to divide two departments in details. Experts have carried out heated debate on the division and it is most commonly agreed that the service industry occupies a major portion of the non-tradable division whereas agriculture and industry are listed in the tradable segment (Officer, 1976, p.1-60).
At present, there are around 250 nations across the globe, each of which has unique area, population, politic and economic situation. In order to examine the capability of the pure Balassa-Samuelson model on economy prediction, large amount of data is required for the research. However, many problem emerges under the first process of data collection. For example, the year of liberty varies from country to country and it is impossible to implement horizontal comparison for some period of time due to data shortage before liberation. This is also the case under social instability, during which government's capability of economic supervision is disabled and relative data are not collected for the specific year. To make matters worse, different countries often apply different statistic regulations and the data require transformation into the same standard after collection for precision. In order to overcome all difficulties mentioned above and for simplicity reasons, all the data are extract from the same database of the World Bank (WB) for the same standard. However, due the unavoidable data constraint, only sixteen countries possess sufficient data for further research and they are Austria, Colombia, Czech Republic, Denmark, Dominican Republic, Finland, France, Germany, Italy, Netherlands, Norway, Philippines, Romania, Sweden, Switzerland and United Kingdo m.
Collection of data from Chinese economy is also an important sector for the essay. Along with macroeconomic data for the entire Chinese economy from China Statistical Yearbook (CSYB), data of different regions are also demanded. According to diversification, China is a country comprised of twenty-two provinces, five
17
autonomous regions, four municipalities, and two special administrative regions. Special administrative regions are not within the scope of the study due to the different political and administrative systems they operate in comparison to mainland China. Among the rest of the sectors, twelve provinces are chosen for the empirical test (Chongqing, Fujian, Guangxi, Hainan, Hebei, Henan, Hubei, Hunan, Jiangsu, Liaoning, Qinghai, Shandong). These provinces can also be differentiated into economic regions of east and west, where coastal provinces in the east often have higher level of economic development than inland provinces in the west. The detailed data that can't be extracted from the China Sta tistical Yearbook (CSYB) are searched in the Statistical Yearbook of each province.
The data collected from different nations and provinces are required to be sufficient between year 1995 and 2014, a time when Chinese exchange rate overcame depreciation. This is also the period that hasn't been paid much attention to by most economist so far and a period relatively further away from the initial year of economic reform (1978). Only by examining pure Balassa-Samuelson model on Chinese economy through this period and comparing the result with other nations, can the hypothesis be proved to correctly reflect the current status of Chinese economy. However, a period of twenty years is still too short for the stability measurement of the hypothesis. As it is stated in the literature review, China has only adopted open policy in 1978 and it still remains a question whether China has come out of the planned economy phase. Possibility exist that the test between the period of 1995 and 2014 is still an inaccurate measurement of the model and a long term stability test of Balassa-Samuelson hypothesis on Chinese economy using the pure model should be performed to compensate the risk. Therefore, the entire period of Chinese economy since the open policy from 1980 to 2014 should be taken into test for the sustainability of Balassa-Samuelson hypothesis on Chinese economy.
As it is stated in section 4, the productivity increase in the tradable sector ( ) is the difference between the GDP increase for tradable industry ( ) and labor increase in the tradable department ( ) written as
18
(26) and the data collection should focus on two important factors mentioned above to come up with the independent variable of productivity change in the tradable sector.
is measured using the "contribution of primary and secondary industry to real GDP growth". It is the increase of real GDP over the year for the first and secondary industry on the entire real GDP bases. The reason for choosing real GDP instead of nominal GDP is due to the fact that price index is one of the dependent variable of the Balassa-Samuelson model and the nominal GDP level is affected by the price index. By using nominal GDP within the empirical test, interaction exist between the independent and dependent variable, leading to inaccurate result of the test.
To come up with the , two date sources are required and they are "the growth rate of real GDP" and "the total share of the contribution of the primary and secondary industry to the increase of the real GDP". The former data is the total percentage increase of real GDP over the year and the later one measures the proportion for the primary and secondary industry within the total percentage increase of real GDP. The "contribution of primary and secondary industry to GDP growth" measuring is the product of two data mentioned above and the story is the same for , which is measured by the "contribution of primary and secondary labor increase to total labor growth", a product of "the growth rate of total labor" and "the total share of the contribution of the primary and secondary labor increase to the increase of total labor". The definition of the three data are similar with the three data for measuring and the only difference is the change from the increase of real GDP to the increase of labor.
However, since "the total share of the contribution " is not often available due to different statistical standards among different regions as stated before, a substitute way to calculate the "contribution of primary and secondary industry to total increase " is implemented by using the "Composition of GDP (labor) by the three strata of industry" and the "growth rate of real GDP (labor)". The method is more complicated
19
than the previous one since the result needs to be measured using the different between total productivity (or labor) each year but the idea and result remain the same.
Both methods above take total economy as the denominator when producing increase ratio. To be more specific, the real GDP and labor increase of the two tradable industry are divided by the total real GDP and the total labor. This is rather complicated comparing to other data such as the "value add index by industry", a direct data of percentage change for each industry. However, we have to be aware of the fact that the denominator of "value add index by industry" is based on standalone industries. On the one hand, since tradable sector is a combination of both primary and secondary industry, a denominator of standalone industry is never applicable. On the other hand, the dependent variable of the hypothesis for both changes in real exchange rate and price index are all based on the scale of total economy. The difference in the denominator will affect the accuracy of the research result, therefore the productivity increase should be base on the scale of total economy and take total real GDP as the denominator.
Two more important data that should be referred to are the change in real exchange rate and the increase in price index. For the change in real exchange rate, "real effective exchange rate index" extract from World Bank (WB) Database is applied. For the World Bank, "the real effective exchange rate is the nominal effective exchange rate (a measure of the value of a currency against a weighted average of several foreign currencies) divided by a price deflator or index of costs". By using the "real effective exchange rate", the exchange rate is treated equally across all nations since they are of the same standard and the effect of price index, which will also be tested later on, is removed for precision. However, the "real effective exchange rate index" is not the same as "real effective exchange rate". To be more specific, the index is a measurement of the real currency value according to the real effective exchange rate and the decrease in real exchange rate is presented as an increase of real currency value. Therefore, in the remaining part of the essay, "currency appreciation" is
20 implemented as the decrease in real exchange rate.
For the increase in price index, the Consumer Price Index (CPI) is chosen among many other price index such as the Producer Price Index (PPI), GDP deflator as well as normal inflation rate. The reason why the later are not favored for estimate are as follows. To begin with, real appreciation predicted by the Balassa-Samuelson hypothesis is the final stage of economic movement affected by productivity growth, whereas PPI is located at the initial position of the industrial chain. In addition, the economic scope that GDP deflator concerned is much wider than CPI. Since the real appreciation first occur in the service industry according to theory, CPI can be less affected by the fluctuation of other departments compared with GDP deflator and offer more accurate predictions. Finally, as it is widely accepted, inflation rate is a comprehensive and secondary value conducted by many other index including CPI, PPI, GDP deflator and many more. As a result, precision of the result can't be guaranteed comparing with CPI. The data for CPI is also extracted from the World Bank Database for equality across nations and the data derived from CPI for the increase in price index is called " inflation conducted by CPI" in the remaining parts of the essay.
6. Research Methodology
As it is talked about in section 3, the research method of my literature is greatly inspired by the overview of the Balassa-Samuelson hypothesis in Asia. To be more specific, The author of the book took the average increase of both dependent and independent variable within the researched years and observed the distribution of the points by using the X-Y diagram. The independent variables within the research, including the average growth rate, average machinery export and average productivity growth for each country were put on the horizontal line and dependent variables of both average change in real exchange rate and in inflation were put in the horizontal line. After the observation of scattered diagram, coefficient correlation between the dependent and independent variables are calculated with and without China to
21
measure the feasibility of Balassa-Samuelson hypothesis on planned economic and the correlation test is also performed on the self constructed model of the author for the feasibility of the model.
Since the only dependent variable in the pure Balassa-Samuelson model of my essay is the productivity increase in the tradable sector, the average productivity increase of each country should be calculated. And as it is stated in the literature review, to fully test the capability of Balassa-Samuelson hypothesis on real economy, both the decrease in real exchange rate and increase in price index should be considered. Therefore, average currency appreciation and the average inflation conducted by CPI are also calculated and a scattered plot can be implemented by either two of the variables above for statistical analysis. However, although diagrams with only average measurements are possible to reveal some basic relationship between the three variables, in order to test the feasibility of the pure Balassa-Samuelson model and for further accuracy, empirical test of coefficient correlations for each country between the three variables should be considered. In addition, correlation results should also be included for the construction of scattered diagram together with three average measurements above for further statistical analysis and if the result is inconsistent with the expectation, internal factors of region difference between Chinese provinces should be examined using the same method.
The long term affect of Balassa-Samuelson hypothesis on Chinese economy is also considered using the compatibility of the model. As it is mentioned in the previous chapter, due to the open policy and economic reform in China, the static result of short time period may be inappropriate and should be compared with the long term result for consistency. However, it is insufficient to only calculate the average and correlation difference between two different time periods since the data is too scarce and the variation across time can't be fully revealed. Therefore, coefficient correlation as well as the mean value are calculated under a twenty year period from the start of 1980 until 2014 on a continuous time series namely 1980-2000, 1981-2001, 1982-2002 and until 1995-2014. The results derived from these time periods are also
22
performed on the X-Y diagram for the consistency of the research and only after all empirical tests mentioned above, can the true relationship between the Balassa-Samuelson hypothesis and Chinese economy be revealed.
7. Empirical results
As it is introduced in the previous sectors, the main prepoce of this article is to testing the feasibility of Balassa-Samuelson hypothesis on the current status of Chinese economy. To justify the result, seventeen different nations are tested as a reference for the pure Balassa-Samuelson model and the result are shown in figures below.
Fig. 3. Average productivity change in tradable sector versus correlation between productivity change in tradable sector and change in real currency value for eighteen nations (1995-2014) Austria Switzerland Colombia Czech Republic Germany Denmark Dominican Republic Finland France United Kingdom Italy Netherland Norway Philippines Romania Sweden China -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 C or re la ti on b et w ee n pr od uc ti vi ty cha nge i n t ra d abl e se ct or a nd c ha nge i n r ea l cur re nc y va lue
23
With the help of three figures listed above and below, three important factors are revealed for the nations analyzed within the research. To begin with, the pure Balassa-Samuelson model display different statistic relationship for different nations according to the level of economic development of the country. As it is widely accepted, nations around the globe can be divided into two types by using economic measurement and they are the developed countries and developing countries. According to Fig. 3 and 4, developing countries of Colombia, Dominican Republic and Philippines always have a positive correlation between productivity change in
Fig. 4. Average productivity change in tradable sector versus correlation between productivity change in tradable sector and inflation conducted by CPI for eighteen nations (1995-2014) Austria Switzerland Colombia Czech Republic Germany Denmark Dominican Republic Finland France United Kingdom Italy Netherland Norway Philippines Romania Sweden China -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 C or re la ti on b et w ee n pr od uc ti vi ty cha nge i n t ra d abl e s ec tor a nd inf la ti on cond uc te d by C P I
24
tradable sector and change in real currency value whereas the coefficients between productivity change in tradable sector and inflation conducted by CPI is negative. The trend however, is completely reversed for developed countries like Austria, Czech Republic, Denmark, Finland, France, Germany, Italy, Netherlands, Norway, Switzerland and United Kingdom as they dominate the positive correlation between productivity change in tradable sector and inflation conducted by CPI and the correlation between productivity change in tradable sector and change in real currency value is negative.
In addition, according to the distribution in Fig. 5, most nations analyzed within the research are displayed around a negative trend line passing through the second and fourth quadrant of the graph and a negative relationship is revealed between the two types of correlations. The statistical result is consistent with the prediction of the Balassa-Samuelson hypothesis since the fully effect of the hypothesis is either displayer in the change in real exchange rate or the change in price index according to the previous literature and the two dependent variables can be seen as the alternative of each other. As a result, a positive correlation on one side will lead to a negative coefficient on the other and this is why the second and fourth quadrant are dominated the most.
However, this still can't fully explain the negative trend line they converge to and the phenomenon that the stronger the positive coefficient on one correlation the higher the negative coefficient on the other for each nation. According to the substitute effect between the real exchange rate and price index mentioned above, a ratio of tradeoff exist between the two dependent variables and the negative trend line can be explained as the appearance of sacrifice ratio between the two tradeoffs. Meanwhile, a higher positive correlation on one index always lead to a higher increase. Since the total effect productivity increase in one country is fixed, a significant decrease in the other index is needed for the feasibility of the Balassa-Samuelson hypothesis, leading to a higher negative coefficient on the other index.
25
increase in tradable sector for different countries are about the same and they are mostly distributed around the value of 0.5, which means a average growth of GDP by about 0.5 percent each year for different countries despite their difference in economic development. The result of average productivity growth reveals an external factor of globalization and under the trend of economic globalization, new production methods to increase productivity are spend throughout the world immediately without barriers, leading to the same productivity increase for both developed and developing countries.
Fig. 5. Correlation between productivity change in tradable sector and change in real currency value versus correlation between productivity change in tradable sector and inflation conducted by CPI with trend line for eighteen nations (1995-2014)
With all the finding listed above, despite the weak numerical value of the Austria China Colombia Czech Republic Denmark Dominican Republic Finland France Germany Italy Netherlands Norway Philippines Romania Sweden Switzerland United Kingdom -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 C or re la ti on be tw ee n pr od uc ti vi ty cha nge i n t ra d abl e se ct or a nd inf la ti on cond uc te d by C P I
Correlation between productivity change in tradable sector and change in real currency value
26
correlation, the feasibility of the pure Balassa-Samuelson model is proved. The weak correlation value can be explained by the simplicity of the model and it is the country distribution that contribute to the compatibility of the pure model.
However, as we compare the data of different countries with Chinese economy, it is always spotted at the right up corner of Fig. 3, 4 and 5, a point far away from the distribution of both developing and developed countries as well as the negative trend line. In face of this, as it is mentioned in the literature review and accepted in the research methodology, both the internal factors of Chinese province and the stability of the Balassa-Samuelson model on Chinese economy should be tested. The analysis of province factor is performed in the Fig. 6 below and the scattered plot for provinces is compared with the distribution of different countries in Fig. 4. Since the real exchange of china is of different scale with the productivity change of each province, only the correlation between productivity change in tradable sector and inflation conducted by CPI of each province is considered.
The result of Fig. 6 have both similar and different characteristics comparing to Fig. 4. For the similar perspective, the distribution of correlation between productivity change in tradable sector and inflation conducted by CPI for each province correspond with the distribution of different countries and they both contain parts above and below the line of zero. For the diverged perspective, the productivity increase in all the estimated provinces are higher than the level of both developing and developed countries in similar statistics and the economic difference between provinces can't be revealed by using correlation results.
The reason for the productivity difference is beyond the scope of this research since Balassa-Samuelson hypothesis only focus on the effect of productivity change and the productivity gap can be broadly explained by the open policy of China. However, the prediction failure of the coefficients for using province test on Chinese economy need to the resolved. As it is previously mentioned in section 5, t hese provinces can be differentiated into economic regions of east and west, w ith coastal provinces in the east often have higher level of development than inland provinces in
27
the west. In order to implement the geographic position of different regions, Fig.7 is displayed below for better understanding of different provinces.
Fig. 6. Average productivity change in tradable sector versus correlation between productivity change in tradable sector and inflation conducted by CPI for twelve provinces and China (1995-2014)
As we can see from Fig. 6 above, for all the provinces within the list, the correlation of product efficiency with CPI is distributed out of order. There are coastal regions such as Hainan and Shandong, which are considered to be fully developed having negative correlations just as developing countries and there are under developed inland provinces such as Chongqing and Guangxi, having a positive correlation with the CPI just as many developed countries.
The reason for explaining the correlation difference between provinces and nations relies on the difference in nature among them. For an independent country, no matter how small it is, it possesses independent monetary and fiscal base in face of
Liaoning Fujian Hainan Hebei Jiangsu Shandong Guangxi Qinghai Chongqing Henan Hubei Hunan China -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0 0.02 0.04 0.06 0.08 0.1 C or re la ti on b et w ee n pr od uc ti vi ty cha nge i n tr ad abl e se ct or a nd inf la ti on c ond uc te d by C P I
28
global economy. Province on the contrary, doesn't have self controlled monetary and fiscal policy no matter the size of the region and under t he effect of change in productivity, no rightful monetary and fiscal measures can be implemented as countries can do. Since China is a country consisted of many regions, it is not feasible to regard it as a combination of different small countries with the trend of convergence to the negative linear regression and the distribution difference with the other countries can now be explained.
Fig. 7.
With internal analysis above, the compatibility of Balassa-Samuelson hypothesis on China between 1995 and 2014 is revealed. However, as we can see from Fig. 6, the correlation of total economy is higher than every province within the measurement and this can't be verified using any explanation above. To answer this question and for
29
the consistency of the hypothesis on Chinese economy, a long term test as is stated in the previous section is implemented the results are as follows.
Fig. 8. Average changes for productivity in tradable sector, currency value and inflation conducted by CPI
As we can see from Fig. 8, the Balassa-Samuelson hypothesis holds in the long run since the substitute effect between the real exchange rate and price index is applicable for the Chinese economy. However, this is still too weak for the statement of a long term relationship and in Fig. 9, the two correlation for Chinese economy changes from time to time. To be more specific, the correlation change can be divided into three periods. The first period between 1980 to 2008 witnesses stable correlation of about 0.3 for both the real exchange rate and price index, the second period starting from 2009 to 2013 indicates tremendous increase for the correlation of real currency value and enormous decrease for the correlation of inflation. And the two coefficient correlations try to converge to the initial level in the third period starting from 2014.
-0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 A ve ra ge c ha nge s
Average productivity change in tradable sector Average change in currency value
30
Fig. 8. Correlation change between productivity change in tradable sector and change in real currency value versus correlation change between productivity change in
tradable sector and inflation conducted by CPI
For the better understanding of the correlation change s and for the comparison between countries, Fig. 8 is further transformed in to a diagram consistent with the structure of Fig. 5. and by comparing Fig. 9 with the countries and regression in Fig. 5, we find a stable correlation positions for China between 1981 and 2008 much closer to the negative trend line compared to the correlation result between 1995 and 2014. The outcome of this empirical research is quite interesting since according to previous research, it is the first few years of economic reform that may falsify the hypothesis whereas the correlation until 2008 remained constant. It is only when the data after 2008 are taken into account when the correlations suffer tremendous changes.
The reason for the severe vibration is obvious because the only issue that broke -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Correlation between productivity change in tradable sector and change in real currency value
Correlation between productivity change in tradable sector and inflation conducted by CPI
31
out in 2008 and had great impact on later economy is the economic crisis. Since China is growing into an open economy, it is now more easily affected by global economy than ever before and along with new monetary and fiscal policies implemented to conquer the crisis, it is not surprising that the correlation position of China fluctuate around the graph from 2008 to 2014.
Fig. 9. Correlation between productivity change in tradable sector and change in real currency value versus correlation between productivity change in tradable sector and inflation conducted by CPI for China between 1980 to 2014 on a twenty year bases
What's more, as the effect of new policies slowly reveal itself and economy gradually restore from global crisis, a trend of convergence occur at the year of 2014 for both coefficient correlations of China. Up till now, it is still uncertain whether the two correlations can converge to their initial position, much less to predict the time for convergence and the new position of steady state but still, the empirical test proves that the Balassa-Samuelson hypothesis is compatible with Chinese economy on the long run due to the consistency of the correlation between 1980 to 2008. The reason 1981-2000 1990-2009 1995-2014 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.4 -0.2 0 0.2 0.4 0.6 C or re la ti on be tw ee n pr od uc ti vi ty cha nge i n tr ad abl e se ct or a nd inf la ti on cond uc te d by C P I
Correlation between productivity change in tradable sector and change in real currency value
32
why stability of the correlation of Chinese economy instead of the numerical value should be paid much attention to is due to the fact that only under the situation of steady correlation can further models on expla ining the relationship be conduct and maintained. Under the situation of fluctuated correlations, the parameter of the model may vary from time to time and under more extreme conditions, the model itself may loss its feasibility and more efforts will be need to reconstruct a new model adaptive to the new relationship.
8. Conclusion
The evidence in the previous section shows the applicability of the Balassa-Samuelson hypothesis to China on the long run although violations were also evident. The long term feasibility of the Balassa-Samuelson hypnosis on Chinese economy comes from two factors. Firstly, the changes in real exchange rate and price index of China are consist with the tradeoff effect of the Balassa-Samuelson model. In addition, subsequent analysis prove the stability of the coefficient correlations between the Balassa-Samuelson hypothesis and Chinese economy from 1980 to global recession in 2008.
However, since China is a large economy consisted of many provinces without independent monetary and fiscal system, it can't be regard as a combination of different small countries with the trend of converging to the negative linear regression. As a result, the effect of the Balassa-Samuelson hypothesis for China is less applicable than for other small and open economy and the statistic result of China is diverged from the distribution of other countries. As for the period after 2008, the Balassa-Samuelson hypothesis is neither possible nor sustainable on explaining current Chinese economy since the two correlation results for China diverge year from year and only when the correlations become stable once more can the Balassa-Samuelson model be used once more for predicting Chinese economy.
33
Appendix 1. Data of eighteen countries including China (round to 3 decimals)
Country Name Average changes for productivity in tradable sector Correlation between productivity change in tradable sector and change in real currency value
Correlation between productivity change in tradable sector and inflation conducted by CPI Austria 0.006 -0.189 0.026 China 0.060 0.242 0.726 Colombia 0.000 0.242 -0.155 Czech Republic 0.013 -0.061 0.019 Denmark 0.006 -0.584 0.439 Dominican Republic 0.010 0.174 -0.235 Finland 0.006 -0.250 0.022 France 0.003 -0.100 0.064 Germany 0.006 -0.363 0.285 Italy 0.002 -0.406 0.434 Netherlands 0.004 -0.215 0.067 Norway 0.011 -0.010 0.153 Philippines 0.010 0.177 -0.372 Romania 0.006 -0.446 -0.548 Sweden 0.008 0.224 0.063 Switzerland 0.005 -0.241 0.282 United Kingdom 0.004 -0.162 0.003
Appendix 2. Data of twelve provinces including China (round to 3 decimals)
Province Average changes for productivity in tradable
Correlation between productivity change in tradable sector and inflation
34
sector conducted by CPI
China 0.060 0.726 Chongqing 0.088 0.196 Fujian 0.066 0.533 Guangxi 0.067 0.337 Hainan 0.047 -0.342 Hebei 0.069 0.439 Henan 0.072 0.587 Hubei 0.070 0.599 Hunan 0.069 0.196 Jiangsu 0.083 0.484 Qinghai 0.070 -0.202 Shandong 0.071 -0.613
Appendix 3. Average data for China between 1980 and 2014 measured by 20 years (round to 3 decimals) China Average productivity change in tradable sector Average change in currency value Average change in inflation conducted by CPI 1981-2000 0.052 -0.044 0.074 1982-2001 0.054 -0.037 0.073 1983-2002 0.055 -0.035 0.071 1984-2003 0.055 -0.038 0.071 1985-2004 0.054 -0.034 0.072 1986-2005 0.054 -0.026 0.068 1987-2006 0.056 -0.012 0.065 1988-2007 0.057 -0.003 0.064 1989-2008 0.057 -0.003 0.058
35 1990-2009 0.059 -0.009 0.048 1991-2010 0.068 0.004 0.048 1992-2011 0.068 0.011 0.049 1993-2012 0.065 0.016 0.047 1994-2013 0.063 0.016 0.041 1995-2014 0.060 0.028 0.030
Appendix 4. Correlation data for China between 1980 and 2014 measured by 20 years (round to 3 decimals)
China
Correlation between
productivity change in tradable sector and change in real currency value
Correlation between productivity change in tradable sector and inflation conducted by CPI
1981-2000 0.303 0.332 1982-2001 0.271 0.319 1983-2002 0.271 0.296 1984-2003 0.268 0.290 1985-2004 0.299 0.314 1986-2005 0.324 0.296 1987-2006 0.326 0.277 1988-2007 0.347 0.271 1989-2008 0.349 0.294 1990-2009 0.477 0.433 1991-2010 -0.162 0.816 1992-2011 -0.225 0.813 1993-2012 -0.201 0.856 1994-2013 -0.332 0.841 1995-2014 0.242 0.726
36 Reference
Balassa, B. (1964). The purchasing power parity doctrine: A reappraisal. Journal of
Political Economy, 72, 584−596
China Statistical Yearbook. (various editions). Beijing: National Bureau of Statistics of China.
Lawrence H. Officer (1976). The Purchasing-Power-Parity Theory of Exchange Rates: A Review Article. International Monetary Fund, Vol. 23, No. 1 (Mar., 1976), pp. 1-60.
N. Gregory Mankiw. (2009). Principle of Economics.(Fifth Edition). Beijing: Peking University Press.
N. Gregory Mankiw. (2010). Macroeconomics.(International Edition). New York: Worth Publisher.
Pilbeam, K. (2013). International Finance. Beijing: China Taxation Publishing House.
Qian Guo (2010). The Balassa–Samuelson model of purchasing power parity and Chinese exchange rates. China Economic Review, 21 (2010), 334–345.
Sylviane Guillaumont Jeanneney, Ping Hua (2002). The Balassa–Samuelson effect and inflation in the Chinese provinces. China Economic Review, 13 (2002), 134–160.
Takatoshi Ito, Peter Isard, and Steven Symansky. (January 1999). Changes in Exchange Rates in Rapidly Development Countries: Theory, Practice, and Policy Issues. (NBER-EASE volume 7). University of Chicago Press.
Taylor, A. (2002). A century of purchasing power parity. Review of Economics and
Statistics, 84, 139−150.
