The relationship between wages, productivity and
unemployment through time in China
Romy van As
(10186042)
Bachelor Thesis Economics and Business
Specialization: Economics and Finance
Faculty of Economics and Business
Supervisor: Stephanie Chan
Academic year: 2014 – 2015
Statement of Originality
This document is written by Romy van As (student number 10186042), who declares to take full responsibility for the contents of this 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.
Abstract
Since the economic reforms of China, the Chinese economy shifted from a planned to a market-‐ oriented economy. Real wages, labour productivity and unemployment represent an important connection within labour markets and the economic reforms caused big changes in these variables. There are many economic theories that state that there is a relationship between the variables. The connection between these variables received a sizable amount of attention in other countries, but not in China. This study examines the relationship between wages, productivity and unemployment in China through time. Time series econometric methods show that there is a long-‐run equilibrium relationship between real wages and productivity, and that productivity positively causes real wages in the short-‐run, which is supported by the marginal productivity theory.
Content
1. Introduction 5 2. History 8 2.1 The Chinese economic reforms 8 2.1.1 Unemployment 8 2.1.2 Wages 10 2.1.3 Productivity 11 3. Literature review 13
3.1 Wage equation 13
3.2 Literature review about the wage-‐productivity-‐unemployment nexus 13
3.3 Findings of other studies 16 4. Methodology 18 5. Data 21 5.1 Defining the variables 21 5.2 Data information 22 6. Results 24 6.1 Preliminary data analysis 24 6.2 Cointegration test for 1980 – 2009 27 6.3 Cointegration between wages and productivity, 1994 – 2009 31 7. Conclusion 34 8. References 36 9. Appendix 38
1. Introduction
More than 30 years ago, China started the long process of embracing free-‐market principles. China achieved a gross domestic product (GDP) growth from 226 dollar per capita to 6091 dollar per capita in the period from 1978 until 2012. China’s export has grown about 13.6% per year from 1980 to 2000, which is very remarkable since the world average export growth per year was 5.9% over the same period (Yue and Hua, 2002). The export growth of manufacturing goods was even higher, namely 18.8%. In the period from 2000 to 2005, manufacturing accounted for 89% of China’s merchandise exports and 32% of its GDP. This made China more specialized in the manufacturing sector than other large developing economies (Hanson and Robertson, 2008)1. China has become a
major source of supply in labour-‐intensive manufacturing goods and China has a comparative
advantage in producing these goods, which ensures that world product prices are pushed down (Goh and Wong, 2010).
China is now called the world second largest economy and the “Reform and Opening Up” policy, which started in 1978, was a significant turning point that contributed to this current position. The economic reforms of China during this period contributed to the fact that China’s economy moved from a planned to a market-‐oriented economy (Lin, Cai, & Li, 1996). The shift to this economy contributed to a better exploitation of the comparative advantage of China in labour-‐intensive manufacturing (Lin, Cai, & Li, 1996).
The increased Chinese trade has led to higher labour productivity in China (Young, 2000). Significant changes in the structure of wages and the labour market have occurred in the last twenty years because of the economic reforms (Yueh, 2004). These changes caused that wages, labour productivity and unemployment changed a lot through time. Labour productivity increased a lot because of the growth of the manufacturing GDP, wages increased in the last 15 years and unemployment also made a lot of changes. There has been a political debate about the impact of globalization on the developed world economies. Low wages and rapid increases in productivity in “emerging” economies like China are putting high wage countries at a cost disadvantage (Krugman, Melitz & Obstfeld, 2012).
Real wages, productivity and unemployment represent an important connection within labour markets and it received a sizable amount of attention in the literature. There have been put forward many economic theories that state that there is a relationship between these variables. An example of such a theory is the efficiency wage hypothesis, which is a theory about the relationship
between wages and productivity (Gärtner, 2009). Conventional economic theory suggests that higher productivity in firms is rewarded with higher wages, but conversely the efficiency wage hypothesis proposes that higher wages cause greater productivity (Wakeford, 2004). This proofs that
theoretically the causal relationship between wages and productivity is interpreted in different ways. This makes the development of real wages and unemployment and the strong surge in productivity in China an interesting case to re-‐examine the relationship between wages, productivity and unemployment. Because China is highly specialized in the manufacturing sector and this made China’s emergence potentially disruptive (Hanson and Robertson, 2008), the focus in this thesis will be on the manufacturing sector.
This leads to the following research question:
"How did the relationship between wages, productivity and unemployment in China behave through time?"
This thesis will look at how the productivity-‐wage-‐unemployment relationship in China moves over time and how this relationship reacts to the economic reforms. To do this macro-‐economic data for China from 1980 to 2009 will be used.
In this thesis, the relationship between productivity, wages and unemployment will be examined by applying time-‐series econometric methods. There will be a focus on several sub-‐questions in order to assess the research question. First, there will be examined if there is evidence of structural breaks and trends, which will be identified by plots of real wages, productivity and unemployment. From this we can examine if the variables are responsive to the economic reforms mentioned above. Second, there will be looked at if there is a long-‐term equilibrium (co-‐integrating) relationship between real wages, unemployment and productivity. Thirdly, there will be examined if a Granger causality test enlightens the short-‐term directions of causality between the variables. Based on the results arising from the analysis, this thesis discloses how and to what extent the real wage, productivity and unemployment affect each other.
Section 2 contains the history of the Chinese economic reforms. In section 3, the literature review, attention will be paid to a couple of things. The section will begin with a review of the labour market theories about the relationship between wages, productivity and unemployment together with a discussion of the possible causal relationships between wages, unemployment and productivity. A selection of previous empirical research about the relationship between these variables in other countries will follow. Section 4, the methodology, outlines the method that will be used in this thesis. Section 5, the data section, first pays attention to the definitions of the three variables and then
describes the data that will be used. Section 6, the results, analyses the data, reports the empirical results of the tests that have been used and will state the interpretations that became apparent. In section 7, the thesis will be concluded with an answer to the research question.
2. History
This section discusses a history of the Chinese economy that outlines the movements of the wage, productivity and unemployment through time caused by the economic reforms of China according to the literature. Figure 1 gives an overview of all the events during the economic reforms that are relevant for the discussion of the wage, productivity and unemployment over time.
2.1 The Chinese economic reforms
In the period between 1978 and 1998, the GDP per capita in China grew 8% per year. This
performance makes that China is the most rapidly growing economy during this period in the world (Young, 2000). These rapid economic developments were caused by economic reforms to a great extent. In December 1978, at the third plenum of the eleventh central committee, the authorities of China decided to make the development of the economy a top priority (Yueh, 2004). The economic reform program that they started was initiated to seek welfare gain and to tackle urgent problems such as the inefficiency of the planned economy (Cai & Wang, 2010). It led to a transition to an economy that is market-‐oriented (Liu, 2012). The reform program was characterized by the four modernizations of industry, agriculture, national defence, and science and technology (Yueh, 2004). The entry of China into the World Trade Organization (WTO) in 2001 was accompanied by an increase in key economic indicators such as the GDP (Lee & Warner, 2007).
2.1.1 Unemployment
The high growth of the GDP during the reforms was accompanied by increasing unemployment rates (Liu, 2012), as will become clear in this part.
Before 1978, in the planned economy of China, three key components characterized the employment system. First, the government allocated the labour force by intervention in the overall urban employment to secure that urban workers had stable and life-‐long employment (Cai & Wang, 2010). Enterprises were not allowed to choose their own employees and employees were not allowed to choose their own jobs (Liu, 1998). Once a worker was allocated to a job, he did not have a chance to switch jobs and he also could not be dismissed, so there was no fear of unemployment, which is why unemployment rates before the reforms in urban areas were low (Liu, 1998). The second component that characterized the employment system before 1978 is the household registration (hukou) system (Cai & Wang, 2010). This system separates the labour and the population of China between urban and rural areas. This harms the functioning of the integrated labour market, because fewer rural surplus workers than was expected were absorbed by the
(industry-‐oriented) economic growth because of the restriction between urban and rural areas (Cai & Wang, 2010).
Third, the employment system of China before the reforms was characterized by a host of welfare policies (such as urban exclusive social security policy), which deterred the labour mobility and the equal treatment of the population between urban and rural areas (Cai & Wang, 2010). These three institutional arrangements had a negative effect on the efficiency of the employment allocation (Cai & Wang, 2010).
From 1978, labour mobility increased because of the economic reforms (Liu, 1998). This was
implemented by: 1) reducing the role of the government in the allocation of labour and 2) setting up a new employment system that was based on contracts (Liu, 1998).
First, the labour reform program with the aim of increasing labour mobility reduced the influence of the state regarding labour allocation (Liu, 1998). This was done by providing general guidelines and by leaving the allocation to other agencies such as the enterprises themselves or to labour service companies (Liu, 1998). From the mid 1980s, enterprises have been legitimated to select and dismiss workers. With this change and with increasing competition pressure, employment has become more market-‐oriented (Fang & Zhao, 2009).
Second, the new system that was based on contracts was implemented. It began by implementing the contract responsibility system. During the economic reforms, this system was implemented into the rural areas at the end of the 1970s (Cai & Wang, 2010). Individual households and local managers were held responsible for the profits and losses of their enterprise by contracting the former collective-‐owned land to them (Cai & Wang, 2010). This caused improving incentives in the rural sector, which was accompanied by a labour surplus in this area. As a result, many labourers migrated from rural to urban areas.
In 1983, the contract responsibility system was implemented in the state enterprises. This was the most significant labour reform program that was investigated in the study of Liu (1998). Under this system, the contract between the enterprise and the employee had to meet all kinds of conditions such as the duration of the agreement and the wages. On the expiration date both parties could decide either to renew the contract or to end the agreement. This has strengthened the powers of managers over their work force but it also has given freedom to the workers to quit their job and to find other employment opportunities.
In 1986, the Temporary Regulations on Labour Contract System of State-‐owned Enterprises was issued by the administration. All State Owned Enterprises (SOEs) were required to recruit new workers based on voluntary contracts. Under this new system, workers were to be contracted and
During the reforms, the restrictive regulations and welfare policies mentioned before were removed, which increased labour mobility since the 1980s (Cai & Wang, 2010).
The changes in the policy during the reforms mentioned above have had impact on the labour market and unemployment rates in the last three decades (Liu, 2012).
One factor that contributed to the high unemployment rates in the beginning of the
economic reforms is the inflow of people from rural areas who were seeking jobs in urban areas (Liu, 2012). The inflow of people was caused by the labour surplus and thus unemployment in the rural sector because of the contract responsibility system. Liu (2012) confirmed that urban unemployment has risen strongly between 1988 and 2002.
Unemployment started to rise from the mid 1980s because enterprises were legitimated to dismiss workers and all SOEs were required to recruit new workers based on voluntary contracts. Under this new system, workers were to be contracted and re-‐chosen on the basis of their performance (Fang & Zhao, 2009). This caused that a lot of workers were dismissed.
Another factor that had major impact on the labour market and unemployment is the massive layoff of workers in the urban areas due to the retrenchment reform in the state-‐owned sector (Liu, 2012). The restructuring and the privatization of urban collective-‐owned and state-‐owned enterprises caused declining numbers of manufacturing workers (because of the layoffs) (Banister, 2005). In this sector there were high deficits, which caused that many SOEs went bankrupt in the 1990s (Liu, 2012). This was a reason for the government to initiate the retrenchment reform. This urban radical reform was intended to solve the inefficiency problem in the state sector and this was done by the layoff of approximately a quarter of the workers in this sector in the period between 1997 and 2000. The laying off of workers was accompanied by unemployment and a decline in urban employment (Cai & Wang, 2010).
The East Asian financial crisis in the end of 1997 also had a negative impact on the Chinese economy, which was experiencing a downturn, in turn causing managers to lay off a lot of workers (Cai & Wang, 2010), which contributed to the rise of the unemployment rate.
2.1.2 Wages
An important part of the reforms was the reform program that aimed to improve the linking of wage levels with enterprise and individual performance (Liu, 1998). Since 1978, there have been a few sets of reforms regarding wages, specifically in 1985, 1992, 1994 and 1996 (Yueh, 2004).
redistributed by the state (Yueh, 2004). Wages and bonuses were unrelated to the individual and enterprise performance (Liu, 1998). Because of this, there was little incentive to be profitable, so it reduced the motivation and initiative of employers, which in turn caused low productivity.
In 1985, the Ministry of Labour determined that there was going to be a link between wages and the economic performance of collectively owned enterprises and state-‐owned enterprises. The economic performance of these enterprises was to be measured by enterprise profitability or by a combined indicator of economic returns (Yueh, 2004). The goal of linking wages to the economic performance was to provide incentives that were profit-‐oriented (Liu, 1998).
After this, in 1992, the State Council of China permitted enterprises to set their own wage structure if these structures were within the boundary of the overall wage budget that the government had established (Yueh, 2004). The introduction of a new Labour Law in 1994 gave management more discretion over wage determination. Because of these reforms, the share of bonuses in total wages rose for all enterprises (Brooks & Tao, 2003). Also in 1994, publicly listed companies were allowed to set their own wages, but subject to standards determined by the government (Yueh, 2004).
In 1996, the floating wage system was introduced, which was part of the five-‐year plan that ran from 1996 until 2000. This was the most widely reform program that was implemented to better link wages and performance (Liu, 1998). Wages are since then divided in two components: fixed and variable wage. The variable wage includes bonuses, which are based on both enterprise profitability and individual productivity (Yueh, 2004).
Furthermore, the non-‐state sector has grown since the economic reforms. This has created
employment opportunities in this sector (Liu, 1998). The enterprises in this sector have actively tried to recruit managerial and technical workers from state-‐owned enterprises that are highly skilled. They did this by offering higher wages and salaries compared to the state-‐owned enterprises. This caused that many state-‐owned firms lost valuable personnel to the non-‐state sector. This is why the state-‐owned enterprises were under high pressure to increase their wages and salaries (Liu, 1998). Liu (1998) found positive and significant relationships between the reform programs mentioned above and earnings, which indicates that the programs have increased earnings.
2.1.3 Productivity
While in the past, economic growth was expressed through large inputs of new labour and capital into the production process, during the economic reforms the government rather wanted to achieve growth through gains in productivity (Kuan, Hongchang & Yuxin, 1988). Between 1978 and 1988,
include agriculture, retail commerce, personal services and small-‐scale manufacturing operations that are controlled by individuals and collectives (Kuan et al., 1988).
The economic reforms in the late 1970s, which resulted in the miraculous growth, have attracted the attention of many economists. The change in labour productivity caused by a series of technological and institutional changes that resulted from the economic reforms has been crucial and controversial (Wu, 1995). Major efforts have been made to examine what the role of
productivity and efficiency was in this economic growth to investigate whether this growth is sustainable or not (Wu, 2000).
Productivity growth can be seen as the sum of technological progress and technical efficiency change (Wu, 2000). Technical efficiency change refers to catching up on the frontier (i.e.
improvement in efficiency), which means increasing efficiency with the existing resources that are utilised in production. Technological progress refers to shifts in the frontier (i.e. innovation). The economic reforms aimed to improve the technical efficiency. Centrally planned economies, like the Chinese economy was before the reforms, often produce below their best practice outputs. The shift to a market-‐oriented economy aimed to raise the production close to the frontier, so it increased productivity (Wu, 2000). According to international standards, planned economies have faced low levels of technological progress, so again the shift of a planned to a market-‐oriented economy contributed to the stimulation of innovation (Wu, 2000).
Empirical studies about East Asian economies have been paying attention to the contribution of labour productivity to the economic growth of the country. According to Kuan, Hongchang and Yuxin (1998), there can be seen a mild trend in productivity through time, with an upward break that coincides with the efforts of the economic reforms starting in 1978. Banister (2005) argues that sharp increases in manufacturing labour productivity ensued from the restructuring and privatization of state-‐owned and urban collective-‐owned enterprises because those reorganized enterprises are competitive in the domestic and global economies.
Figure 1: timeline Chinese economic reforms 1978: Start
economic reforms
End 1970s: contract responsibility system
in rural areas
Begin 1980s: regulation of government
1983: contract responsiblity system in state
enterprises -‐1985: linking wages and performance -‐1986: SOEs dismissed workers 1992: enterprises set own wages 1994: publicly listed companies set own wages 1996: Gloating wage system 1997: East Asian Financial crisis 2001: entry into WTO
3. Literature review
This section gives a review about the wage-‐productivity-‐unemployment relationship. It first discusses the wage equation and then the possible different causal relationships between labour productivity, real wages and unemployment. The section ends with a part that pays attention to the findings of other studies about the wage-‐productivity-‐unemployment nexus in other countries.
3.1 Wage equation
In recent literature about the wage-‐productivity-‐unemployment nexus, wage equations have been derived. According to these different wage equations, productivity is not the only factor that influences wages. Wages are also influenced by other factors, for example unemployment (Goh & Wong, 2010). The following specification is suggested by Blanchard & Katz (1999):
𝑙𝑛𝑤!− 𝑙𝑛𝑝!! = 𝛼 + 𝛽𝑙𝑛𝑝𝑟𝑜𝑑!+ 𝜆 𝑙𝑛𝑤!!!− 𝑙𝑛𝑝!!! + 𝛾𝑢!+ 𝜀!.
In this formula, 𝑤! is the nominal wage rate in time t and 𝑝! is the price level, where 𝑝!! is the
expected price level in time. 𝑝𝑟𝑜𝑑! Is the level of productivity and 𝑢! is the unemployment rate. The
lagged term of real wage is expressed as 𝑤!!!− 𝑝!!! and proxies for the expected term of real wage
𝑤!− 𝑝!! (Blanchard & Katz, 1999). The coefficient on 𝑢
! is expected to be negative because a high
unemployment rate will cause the wages to decrease, as will become clear from the literature review about the wage-‐productivity-‐unemployment nexus. The coefficient on 𝑝𝑟𝑜𝑑! is expected to be
positive.
3.2 Literature review about the wage-‐productivity-‐unemployment nexus
In recent international macroeconomic literature there are several economic theories that justify a relationship between wages, unemployment and productivity. The efficiency wage theory, the marginal productivity theory and the wage bargaining theory argue that there is a causal relationship between the variables. On the other hand, the insider-‐outsider approach casts doubt on a direct relationship between wages and unemployment. Table 1 summarizes the hypothesized causal relations between the variables.
3.2.1 Relationship between wages and productivity Efficiency wage theory
The efficiency wage theory states that real wages positively affect productivity (Alexander, 1993). The theory rests on the assumption that workers are more efficient (i.e. productive) when they receive higher wages (Gärtner, 2009). There are various arguments that explain why the efficiency and hence labour productivity may increase when real wages rise, such as shirking, and turnover (Gärtner, 2009).
When employees are getting caught shirking and are getting fired because of this, this is not really a punishment as long as the firm pays the market clearing wage, because the employee can find a job with the same wage elsewhere (Gärtner, 2009). The rise of the firm’s wage rate above the market wage will provide an incentive to reduce shirking and in turn to increase productivity. Labour turnover, which is the exchange of incumbent workers for unemployed outsiders, is costly to firms. Those costs include for example hiring and firing activities and on-‐the-‐job training for new workers. If the probability that a worker quits is negatively related to the real wage, than firms may profit from paying a higher wage than other firms because they face lower turnover costs (Gärtner, 2009).
If real wages rise, higher productivity will be established because higher wages raise the costs of job loss (Wakeford, 2004). So according to the efficiency wage theory, by paying their employees a higher wage, firms can increase the efficiency or productivity of their employees.
Because the cost of labour (i.e. the sum of all wages paid to employees, including the cost of employee benefits and payroll taxes paid by the employer) will rise when wages increase, the threat of increasing unemployment (as a result of factor substitution from labour to capital) and in turn the threat of being fired will also stimulate workers to be more efficient (Alexander, 1993). Goh and Wong (2010) argue that high wages will strengthen the employment relationship in the long-‐term, which leads to loyalty from employees to employers and hence to higher effort.
Insider-‐outsider approach
In contrast to the efficiency wage theory, the insider-‐outsider approach does not assume that wages directly affect productivity and unemployment (Goh & Wong, 2010). The insider-‐outsider approach rests on the assumption that insiders are incumbent workers who have more advantageous employment opportunities than the outsiders (Dobbie, 2005). Outsiders are those currently out of employment but who are seeking employment (Gärtner, 2009). The idea is that insiders resist competition with outsiders (i.e. potential new employees) in wage setting, because they refuse to cooperate with outsiders if they try to underbid the wages of the insiders. The theory argues that, because of turnover costs and specific knowhow that can only be acquired within the firm, insiders are able to extract a higher wage than the wage for which outsiders are willing to work (Gärtner, 2009). The impact on the employment is that this behaviour will ensure that there is absence of wage underbidding while there are outsiders who are unemployed and are willing to work for a lower wage. Because insides have the power to ensure that there is absence of wage underbidding, this theory does not assume a direct effect of wages on productivity and unemployment.
Marginal productivity theory
The marginal productivity theory states that the higher the labour productivity of the worker, the higher the worker is paid (Goh & Wong, 2010). When individuals’ pay is performance related, wages will rise if productivity increases, so this will result in a positive causality from productivity to wages (Alexander, 1993).
Besides these theories, if labour unions are bargaining for increases in the real wage on the basis of improvements in productivity in the past, wages will also rise if the past changes in productivity were positive.
3.2.2 Relationship between productivity and unemployment
Changes in labour productivity may cause increases but also reductions in the unemployment rate, i.e. the effect is ambiguous (Wakeford, 2004). If productivity rises, the demand for labour could reduce because greater efficiency in the workforce will ensure that less labour is needed to produce the same output. Assuming that other factors that affect unemployment remain constant, this will cause an increase in the unemployment rate. On the other hand, when productivity rises, the impact on the employment can be positive because of the output effect (Alexander, 1993). This means that the unemployment rate will decrease because workers create more output which in turn creates more employment (Wakeford, 2004).
The impact of unemployment on productivity is hypothesized to be positive, because a higher unemployment rate may be associated with higher effort and hence higher average productivity among the workers that will remain (Wakeford, 2004).
3.2.3 Relationship between wages and unemployment Wage bargaining theory
The effect of unemployment on the real wage can be explained by the wage bargaining theory. The unemployment rate affects wages because of changes in bargaining power (Alexander, 1993). When the unemployment rate rises because of other factors than wage and productivity, union bargaining power is expected to weaken (Wakeford, 2004). This in turn will cause real wages to decrease.
Marginal productivity theory
The unemployment rate will also affect wages by incentivise workers to increase effort to secure their jobs if the unemployment rate rises (Wakeford, 2004). If the workers’ effort and hence productivity rises (according to this theory), wages are expected to increase.
Efficiency wage theory
The efficiency wage theory states that when real wages rise, factor substitution will occur from labour to capital, which in turn will cause a higher unemployment rate (Goh & Wong, 2010).
Involuntary unemployment is also explained by the efficiency wage theory. Involuntary unemployed workers are willing to work at a lower wage than the market-‐clearing wage (Yellen, 1984). But when lower wages are paid, the productivity of the employees who are already working will become lower (according to this theory), which is why firms will not lower the wages (Yellen, 1984).
Causal relation Effect Reason
Real wages à productivity Positive Efficiency wage theory
Real wages à unemployment Positive Higher labour costs will cause substitution from labour to capital
Productivity à real wages Positive (1) Marginal productivity theory: payment based on performance
(2) Bargaining
Productivity à unemployment Positive
Negative
Less labour is needed to produce the same output Output effect
Unemployment à productivity Positive Increase effort to secure jobs
Unemployment à real wages Negative Wage bargaining theory: rise in unemployment will weaken bargaining power
Table 1: Hypothesized causal relations between the variables
3.3 Findings of other studies
Goh and Wong (2010) examined whether there exists a relationship between productivity, wages and unemployment in Malaysia. Their main findings show that unemployment is separated from the equilibrium relationship in the long run between real wages and labour productivity, so there exists a long-‐run equilibrium (cointegrating) relationship between real wages and labour productivity in the period from 1970 to 2005. They found that labour productivity is a meaningful long-‐run factor in determining wages, while the effect of unemployment on the real wage rates in negligible. Real wages are very responsive to a change in productivity in Malaysia (the productivity elasticity of real wages is greater than 1). They found that in the long run, for every 1 per cent rise in labour
productivity, real wages rise by 1,223 per cent.
Goh and Wong (2010) also found that productivity Granger causes real wages (in the short run), which supports the marginal productivity theory. In their study, changes in productivity are not immediately reflected in the real wages, which can be seen by the relatively long lags that they used. Conversely, the real wage has no impact on productivity in the short run.
Jeremy Wakeford (2004) investigated the relationship between real wages, labour productivity and the unemployment rate in South Africa between 1983 and 2002. He found strong evidence of a break in 1990. After this year, all the variables increased rapidly. He found a long-‐term equilibrium
(cointegrating) relationship between real wages and labour productivity for the period 1983-‐2002, but unemployment was not connected to the system. There is also strong evidence found by Wakeford that productivity and real wages are cointegrated in the period between 1990 and 2002. The long-‐term wage-‐productivity elasticity that is found in this study is 0,58, so a 1 per cent rise in productivity is accompanied with a 0,58 per cent rise in real wages. This indicates that productivity has grown faster than wages in South Africa.
Wakeford found the following dynamic (short-‐term) causal relationships: real wages Granger-‐cause productivity negatively. Productivity does not Granger-‐cause real wages, so productivity has no effect on real wages in the short term.
Alexander (1993) used time series methods to analyse the relationship between unemployment, real wages and productivity in the United Kingdom for the period between 1955-‐1991.
The results of this study show evidence of a structural break in the late 1970’s. Before 1979, the central variable was unemployment, which was caused by both productivity and real wages. Unemployment was cointegrated with wages and (separately) with productivity. There was a negative causality from wages to unemployment but conversely there was no causal relationship from unemployment to wages. Alexander found no evidence of dynamic (short-‐term) causality from wages to productivity or vice versa in this period.
In the period from 1979 to 1991, there was a bivariate causal relationship between wages and productivity and unemployment was almost divorced from the system. The long-‐run wage elasticity of productivity was about 0,50 in this period. There was a negative causality from wages to productivity (which contradicts the efficiency wage theory). Alexander found no direct relationship between wages and unemployment for the period 1979-‐1991.
4. Methodology
In this thesis, time series data are used. Time series data are data for a single entity, such as a person, a firm or a country and the data are collected at multiple time periods (Stock & Watson, 2012). Time series data can be used to examine the changes of variables over time and to make a forecast about the future values of those variables. With time series data the dynamic causal effects between variables can be studied.
Prior in performing a cointegration test, it is a common practice to determine the stationarity of the time series. Stationarity means that a time series 𝑌! has a probability distribution that does not
change over time (Stock & Watson, 2012). A stationary time series has the property that the mean, the autocorrelation and the variance structure do not change over time. The most important types of non-‐stationarity are breaks and trends (Stock & Watson, 2012). A trend is a long-‐term persistent movement of a variable through time. In time series, there can be seen two types of trends: a deterministic and a stochastic trend. A deterministic trend is a non-‐random function of time, whether a stochastic trend varies over time and is random. A deterministic trend can be a linear function and a stochastic trend is for example a trend that exhibits a long-‐lasting period of increase followed by a long-‐lasting period of decrease (Stock & Watson, 2012). If the root of an AR(p) model equals 1, then the time series is said to have a unit root. A time series contains a stochastic trend if it has a unit root. A break is another type of non-‐stationarity and it arises when the regression function changes over time. If the variables in a time series are non-‐stationary, then the statistical tests can be unreliable.
A series that is stationary and does not have a stochastic trend is integrated of order zero: I(0). If a series has a random walk trend, then the series is integrated of order one: I(1). It means that the series has a unit autoregressive root (i.e. is stationary in levels), and because a non-‐stationary series needs to difference for it to be stationary, then the first difference of the variable is stationary. If the first difference of series has a trend, then the series is integrated of order two: I(2). This means that the first difference of the variable has an autoregressive unit root and the second difference is stationary. So in general, if 𝑌! is I(d), then 𝑌! is integrated of order d and 𝑌! must be differenced d
times to be stationary and to eliminate the stochastic trend. This means that Δ!𝑌
! is stationary.
To detect trends in time series, several methods can be used. An informal method is for example inspecting a plot of a time series (Stock & Watson, 2012). If doubt remains from this method, then a formal method can be used to test the hypothesis that there is a stochastic trend against the alternative hypothesis that there is no stochastic trend. An example of such a method is the Augmented Dickey-‐Fuller (ADF) test. This test is an augmented version of the Dickey-‐Fuller test for a larger set of time series.
If two or more time series variables have the same stochastic trend (i.e. they appear to have a common trend), this is referred to as cointegration (Stock & Watson, 2012). If this is the case, regression analysis can reveal long-‐run relationships between the variables. The definition of cointegration is as follows: When 𝑋! and 𝑌! are both integrated of order one, then if for the
coefficient 𝜃, 𝑌!− 𝜃𝑋! is integrated of order zero (i.e. is stationary), then 𝑋! and 𝑌! are cointegrated
and 𝜃 is the cointegrating coefficient.
When there are three time series variables, 𝑤𝑎𝑔𝑒!, 𝑝𝑟𝑜𝑑!, and 𝑢𝑛𝑒𝑚𝑝!, then they are
cointegrated if 𝑤𝑎𝑔𝑒!− 𝜃!𝑝𝑟𝑜𝑑!− 𝜃!𝑢𝑛𝑒𝑚𝑝! is stationary. When there are three time series
variables, there can be more than one cointegrating relationship. A test for multiple cointegrating relationships is the Johansen cointegration test (1988). For the Johansen test, variables must be non-‐ stationary at level but stationary in first differences (Johansen, 1988). This test determines the number of cointegrating relations of a Vector Error Correction Model (VECM). For this test, the null hypothesis and the alternative hypothesis are the following:
𝐻!: 𝑡ℎ𝑒𝑟𝑒 𝑎𝑟𝑒 𝑛𝑜 𝑚𝑜𝑟𝑒 𝑡ℎ𝑎𝑛 𝑟 𝑐𝑜𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑛𝑔 𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠 𝑣𝑠. 𝐻!: 𝑡ℎ𝑒𝑟𝑒 𝑎𝑟𝑒 𝑚𝑜𝑟𝑒 𝑡ℎ𝑎𝑛 𝑟 𝑐𝑜𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑛𝑔 𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠.
If a cointegrating relation is found, then a VECM can be estimated to account for the long-‐term equilibrium and short-‐term dynamics in the system (Stock & Watson, 2012).
An application of the F-‐statistic that is very useful is to test if the lags of one of the variables have predictive content that is useful, above and beyond the other variables that are included in the VAR (Stock & Watson, 2012). If one argues that a variable does not have predictive content, then this corresponds to the null hypothesis that the coefficients on all of the values (lags) of that variable are zero. The Granger causality statistic is the F-‐statistic testing this null hypothesis and the test of this null hypothesis is called the Granger causality test (Stock & Watson, 2012). Alexander (1993, p. 87): “If X and Y are two jointly covariance stationary processes, then X is said to ‘Granger cause’ Y if past Y and past X better predicts current Y than past Y alone”. So if X granger causes Y, then X is said to be a useful predictor of Y (given the other variables in the regression). However, this does not necessarily mean that a change in X will cause a subsequent change in Y, but it does mean that the past values of X are useful for forecasting the changes of X (beyond that contained in the past values of Y). For this test, the null hypothesis and the alternative hypothesis are the following:
𝐻!: 𝑛𝑜 𝐺𝑟𝑎𝑛𝑔𝑒𝑟 𝑐𝑎𝑢𝑠𝑎𝑙𝑖𝑡𝑦 𝑙𝑎𝑔𝑠 𝑎𝑟𝑒 𝑗𝑜𝑖𝑛𝑡𝑙𝑦 𝑧𝑒𝑟𝑜 𝑣𝑠. 𝐻!: 𝐺𝑟𝑎𝑛𝑔𝑒𝑟 𝑐𝑎𝑢𝑠𝑎𝑙𝑖𝑡𝑦
The productivity-‐wage-‐unemployment nexus will be investigated as follows. First, a preliminary data analysis will present some graphs, growth rates and correlation coefficients. The variables are plotted
to see the behaviour of the wages, unemployment and productivity over time. When the variables are presented graphically, there can be found evidence for trends and breaks.
To determine the stationarity of the time series or the order of integration, Augmented Dickey-‐Fuller tests are applied to the three series to detect trends in the series.
After a preliminary data analysis, assuming that the variables are found to be non-‐stationary, the Johansen’s multivariate cointegration test will be applied. This procedure starts with selecting an appropriate lag order for the VAR and then applies trace and maximal eigenvalue tests to determine the number of cointegrating vectors.
Finally, to determine the direction of dynamic short-‐term causal relationships among the variables, Granger causality tests will be performed.
Tests and estimations in this thesis are performed using STATA.
5. Data
This section starts with defining the definitions of the variables and then discusses information about the data that are used.
5.1 Defining the variables 5.1.1 Defining real wages
Real wages are in the literature defined either as real product wages, or as real consumption wages (Wakeford, 2004). We speak of real consumption wages when nominal wages are deflated by the consumer price index (CPI). Wages are then measured in terms of consumption goods (which means the workers’ real purchasing power). The real product wage is the wage rate in terms of the labour cost of production (Wakeford, 2004). For this definition, the nominal wage is deflated by the producer price index to provide a measure in terms of output.
The choice between real consumption wages and real product wages depends on the relationship that is being investigated. If the concern is about real purchasing power of the workers, it is the most appropriate to use real consumption wages as wage measure. Following the study by Goh and Wong (2010) and Alexander (1993), the real consumption wage is used in this thesis. Because the focus will be on the manufacturing sector, the nominal manufacturing wage is deflated by the consumer price index. In this analysis, the average real wages (i.e. per worker) are under consideration, which involves a ratio between the sum of wages of all employed persons and total employment.
5.1.2 Defining labour productivity
In general, the definition of labour productivity is the extent to which a firm’s labour force is creating output in an efficient way (Datta, Guthrie & Wright, 2005). To measure productivity, it is the most appropriate to use marginal productivity, which is the value added (i.e. the contribution to the output) of the last worker hired; only such a measure is not easy to obtain. The second best concept of productivity in economics would be a measure of output per hour of labour input (Wakeford, 2004). However, this is also not easily obtainable. The use of average labour productivity is more common in practice (Wakeford, 2004). This measure can be calculated in several ways, for example by dividing total output by total employment (i.e. total labour inputs) (Datta et al., 2005).
Following the study by Alexander (1993), Wakeford (2004) and Goh and Wong (2010), the measure for productivity is GDP divided by total employment. Again, because the focus lies on the manufacturing sector, the data for GDP and employment are from the manufacturing sector.