masters essay final draft

Upload: ayaz-mahmood-ghani

Post on 05-Apr-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/2/2019 Masters Essay Final Draft

    1/39

    Testing Wagners Law forthe Canadian Economy during the period 1961-2009

    by

    Ayaz Mahmood Ghani

    An essay submitted to the

    Department of Economics

    in partial fulfillment of the

    requirements for the degree of

    Master of Economics

    Department of Economics

    Memorial University of Newfoundland

    6th December, 2010

    St. Johns Newfoundland

  • 8/2/2019 Masters Essay Final Draft

    2/39

    i

    Table of Contents

    Abstract iii

    List of Tables iv

    List of Figures v

    Section 1 Introduction 1

    1.1 Background of Study 1

    1.2 Purpose of Study 1

    1.3 Significance of Study 2

    Section 2 The Growth of Government Expenditure in Canada 4

    Section 3 Theoretical Background 5

    3.1 Wagners Law 6

    3.2 Criticizing Wagners Law 6

    3.3 Formulation for Wagners Law 7

    Section 4 Empirical Literature 8

    Section 5 Empirical Framework and Data 11

    5.1 Framework 11

    5.2 Data Analysis 12

    Section 6 Methodology 12

    6.1 Unit Root tests 12

    6.2 Engel and Granger Cointegration test 14

  • 8/2/2019 Masters Essay Final Draft

    3/39

    ii

    6.3 Granger Causality test 15

    Section 7 Empirical Results 15

    7.1 Unit Root tests 15

    7.2 Engel and Granger Cointegration test 17

    7.3 Granger Causality test 19

    Section 8 Wagners Law and Population Structure 21

    Section 9 Wagners and Unemployment 25

    Section 10 Conclusion 28

    References 30

  • 8/2/2019 Masters Essay Final Draft

    4/39

    iii

    Abstract

    The main purpose of this paper is to investigate the Wagners law by testing the long run

    and short run relationship between real government expenditure and real income for

    Canada from 1961 till 2009. The study carried out Augmented Dickey-Fuller andPhillips-Perron unit root test on the variables to find whether the variables are stationary.

    We used the two-step Engel and Granger cointegration methodology to find the long-run

    relationship between the two variables. The unit root tests show that the variables are

    non-stationary, but their first difference is stationary. Hence, both the variables are

    integrated of order one. The cointegration result gives evidence that real government

    expenditure and real income for Canada do have long-run relationship for the time period

    1961-2009. However, the causality examination, using the error correction terms, shows

    that growth of GDP does not cause the growth in government expenditure, but the growth

    of government expenditure does Granger cause the growth of GDP. From the empirical

    results, we can conclude that Wagners law is not valid for Canada. Wagners law is also

    tested when demographic variables, dependency ratio and unemployment rate, are taken

    into consideration.

    Keywords: Wagners Law; Augmented Dickey-Fuller; Phillips-Perron; Engel and

    Granger Cointegration; error correction terms; dependency ratio; unemployment rate.

  • 8/2/2019 Masters Essay Final Draft

    5/39

    iv

    List of Tables

    Table 1 Summary statistics of log ofreal GDP per capita and logof real government

    expenditure per capita.

    13

    Table 2 Test Statistics to determineunit roots in the variablesused in the study.

    19

    Table 3 Regression result ofEquation 14, 14, 15 and15.

    23

  • 8/2/2019 Masters Essay Final Draft

    6/39

    v

    List of Figures

    Figure 1 Projected G7 Total

    Government Net Debt-to-GDP

    Rates, 2009.

    2

    Figure 2 Fiscal Stimulus flowing in

    2009 and 2010, G7 countries.

    3

    Figure 3 Showing the upward trend of

    Government Expenditure of

    Canada.

    5

    Figure 4 Trend of GDP for Canada

    from 1961 till 2009.

    6

    Figure 5 Trend of Dependency rate for

    Canada from 1961-2009.

    25

    Figure 6 Trend of Unemployment Rate

    for Canada from 1976-2009.

    29

  • 8/2/2019 Masters Essay Final Draft

    7/39

    1

    I. INTRODUCTION

    The relationship between public expenditure and national income has been a center of

    attraction to economists for a very long duration (Peacock and Wiseman, 1961). A profound

    approach to clarify the growth of public spending is Wagners law which asserts that during

    economic development transition, government economic activity increases in relative to private

    economic activity. As Singh and Sahini (1984) states that the relationship between public

    expenditure and national income has been treated in different ways in two areas of economic

    study.

    Public finance analysis showed that growth in public expenditure is caused by growth in

    national income which signifies the Wagnerian approach. However, according to the Keynesian

    approach, the macroeconomic models tend to explain that growth in national income is due to thegrowth in public expenditure. In Wagners (1883) empirical study, public expenditure was kept

    as a dependent variable which indicates that economic growth is the cause of the growth in

    government expenditure. Keynes (1936) conducted his analysis treating public expenditure as an

    independent variable characterized to amend short-term cyclical fluctuations in aggregate

    expenditures. Thus, the causation moves from growth in government expenditure to growth in

    national income.

    This empirical study is based on the Canadian economy due to its cautious involvementin the world political arena. According to IMF, the Canadian economy ranks as the tenth largest

    and Canada is one of the wealthiest nations of the world. Canada has been grouped with

    countries that provide the highest level of economic freedom. The government sector in Canada

    has augmented steadily from the very beginning of the nineteenth century. Economists have

    carried out intensive research to answer the development in the government sector of Canada.

    One of the significant reasons has been the distinguished Wagners law which states that the

    development of an industrial economy will be accompanied by an increased share of public

    expenditure in gross national product (Wagner, 1883). According to the above theory, during

    the transition of economic development, government economic activity increases in relative to

    private sector economy; thus, the process leads to higher degree of public expenses. Hence,

    augmentation in public sector will act as a friction to advancing economic development.

  • 8/2/2019 Masters Essay Final Draft

    8/39

    2

    To test the validity of Wagners law has become intriguing, since the Canadian

    government expenditure share in GDP is on the rise even though the government has maintained

    subtle policy to tackle public sector economy. Examination of Wagners Law hypotheses has

    become important for Canada, since the Canadian government has changed from their past

    policy of deficit financing to an expenditure deficit policy. Government of Canada is trying to

    decrease its expenditure since the early 1990s in order to cope from the large budget deficits and

    growing public debt. The governments budget deficit, for the fiscal year 1996-1997, declined by

    $33.1 billion and that within a span of three years. According to the Economic and Fiscal Update

    (1997) of Department of Finance, the decline in deficit occurred due to increase in revenues,

    which resulted from an expanding and mounting economy, and also due to a reduction in

    government expenditure. According to OECD Economic Outlook (2009) as shown in Figure 1,

    the recent shape of Canadian economy puts them with lowest net debt-to-GDP rates among G7

    countries. The government of Canada maintained such strong position, during the downturn, due

    to its debt reduction efforts in the last decade. Even though holding onto such solid economic

    position, the Canadian government offered the largest economic stimulus package around the

    world as shown in Figure 2 (IMF, Update on Fiscal Stimulus and Financial Sector Measures,

    2009).

    To venture implications from change in policy where government expenditure and

    growth plays key role makes it necessary to conduct empirical studies on Wagners Law

    hypotheses. The analysis will help to find evidence to suggest approach to developing

    government policies which will direct ways to improve public expenditure composition and

    productivity. This study will examine and determine the relationship between government

    expenditure and GDP for Canada from 1961 to 2009. First, the analysis will examine the long-

    run relationship, if only, between government expenditure and GDP. Second, the study will aim

    at Wagners Law by testing the causality between public expenditure and GDP.

    There are ample studies to validate Wagners Law hypotheses for Canada. These inc lude

    Ahsan et al. (1996), Biswal et al. (1999), Kolluri et al. (2000), Chang et al. (2004), and

    Lamartina and Zaghini (2008). The present study is different, since it employs a more recent

    sample period in testing Wagners Law, and takes population structure and unemployment into

    consideration.

  • 8/2/2019 Masters Essay Final Draft

    9/39

    3

    Figure 1: Projected G7 Total Government Net Debt-to-GDP Rates, 2009.

    Figure 2: Fiscal Stimulus flowing in 2009 and 2010, G7 countries.

    The paper is organized as follows. Section II gives an analysis of public expenditure and

    GDP patterns in Canada. Section III outlines the theoretical and empirical models that are

    suggested in the literature. Section IV gives the overview of the empirical literature. Section V

    defines the data, its sources and the framework that will be used in this study. The empirical

    methodology is presented in section VI. Section VII discusses the empirical results of our study.

    Section VIII explains the empirical results when population structure is taken into account.

    Section IX carries out empirical study when unemployment rate is taken into consideration.

    Section X concludes and summarizes the results in the respective subject matter.

  • 8/2/2019 Masters Essay Final Draft

    10/39

    4

    II. THE GROWTH OF GOVERNMENT EXPENDITURE IN CANADA

    The percentage of GDP of government expenditure in Canada is similar to any of the

    industrialized countries. According to OECD statistics National Accounts at a Glance (2009),

    France devoted 28 percent, Germany used 37 percent, Italy allocated 13 percent, Japan devoted

    15 percent, United Kingdom allocated 73 percent and United States of America used up 55

    percent of GDP in the year 2008. G7 countries like Italy and Japan has been noted to apportion

    low government expenditure as a percent of GDP from the year 2006, primarily because of the

    lingering recession of this century. The government size of Canada has increased significantly in

    absolute and relative terms from 1961 till 2009. Real government expenditure increased from $

    61.3 billion in 1961 to $ 273.2 billion in 2009 taking in 2002 constant dollars. Canadian

    economy has maintained a stable increase in public expenditure for the given period with

    expenses allocated to agricultural and energy sector research and development, manufacturing

    sector and service sector.

    Figure 3: Showing the upward trend of Government Expenditure of Canada.

    1960 1970 1980 1990 2000 2010

    Year

  • 8/2/2019 Masters Essay Final Draft

    11/39

    5

    Figure 4: Trend of GDP for Canada from 1961 till 2009.

    0

    1960 1970 1980 1990 2000 2010Year

    Figure 3 show that the government expenditure in Canada has an upward trend over the

    period 1961-2008. However, between 1994 and 1997 the economy experienced decreases in

    public expenditure and slowdown of the economy during the respective period (see Figure 4).

    However, the Canadian economy was able to vitalize due to economys openness and internal

    economic policies, and maintained an upward trend thereafter.

    Figure 4 shows the trend of GDP for Canada from 1961 till 2009. According to the chart,

    the Canadian GDP rose over the years with an upward trend. According to OECD Economic

    Outlook (2009), the GDP in Canada expanded at an annual rate of 0.50 % in the last reported

    quarter of 2009. From 1961 until 2009, Canada's average quarterly GDP growth was 0.84 %

    reaching an historical high of 3.33 % in the last quarter of 1963 and a record low of -1.80 percent

    in the first quarter of 2009.

    III. THEORETICAL BACKGROUND

    In 1883, Adolph Wagner, a German economist, developed a principle called the Law of

    the Increasing Extension of State Activity. The hypothesis deals with the augmentation of

    government activity in the economy; the principle has come to be known as Wagners Law.

    According to Wagner, there are three important reasons to expect an expanding scope of public

    activity. First, as Wagner (1883) puts it, the administrative and protective functions of the state

  • 8/2/2019 Masters Essay Final Draft

    12/39

    6

    have to expand due to the rising complexity of legal relationship and communications. On top of

    that, higher government expenditure is needed on law and order and socioeconomic regulation

    when urbanization and population increases. Second, Wagner felt that the development process

    required larger amounts of capital than what the private sector could provide. Finally, Wagner

    believed that the income elasticity of demand for public commodities was greater than unity.

    Even though Wagner was not the first one to explain the above matter, he was indeed the first to

    try to explain the situation using empirical analysis.

    Wagner was ambiguous in explaining the formulation required for the law, and kept it

    subject to disagreement among economists. An empirical analysis of Wagners Law has been

    criticized, since it is not a precisely and vividly constructed theory. The study usually looks at the

    past and tries to explain the upward trend of public expenditure. Hence, the law is basically

    inclined towards particular variables and their assumed part in historical progress. As the

    assumptions are not distinctly stated, making it inexplicable to accept or decline the law based on

    the historical facts. To add further difficulty the law does not have specific and identical

    empirical counterpart. There are lot of socio-economic variables, not all of which are

    measurable, can be used to explain public expenditure. Economists are in doubt to understand

    what variables can be used to denote both economic development and state activity. However,

    major studies by Gupta (1967), Pryor (1968), Goffman (1968), Musgrave (1969), Bird (1971)

    and Michas (1975), regarding the subject matter, have all used income per capita to explain

    development, even though the proxy is not the only index of development nor is it the only

    agreeable explanation of the law. The respective studies have used government expenditure to

    explain state activity, since it is the most significant and practical measure.

    Bird (1971) criticizes Wagners hypotheses as not a theory, but an axiomatic explanation

    about state development. Bird (1971) state any attempt to test it necessarily does violence to the

    facts by adjusting them to preconceived theory. The author then goes on to signify the fact that

    the relevant variables in the defined formulation are not fixed enough over time to test this

    evolutionary proposition.

    Musgrave (1969:73), one of the critics of Wagners law, states that the law is unable to

    explain whether the principle is related to the relative share of government in national economy

    or only the absolute size of the economy. Peacock and Wiseman (1967) clarify, using political

  • 8/2/2019 Masters Essay Final Draft

    13/39

    7

    theory, the effects of political occurrence on government expenditure. The authors explain the

    situation with the idea of the displacement effect.

    Even though the Wagners law does not show itself to be empirically tested

    straightforwardly, but there are procedures through which tests can be performed to know thevalidity of the law. Researchers have taken the second best way to solve the problem regarding

    the law, since the optimal solution is surrounded by complexity and intricacy. Economists have

    sought for measurable proxies for state development and activity, and using econometric

    methods it is now possible to set apart the consequences of some variables of government

    spending. Moreover, the analysis will also examine the stationarity properties of the data and the

    stability of the variables that caused it. We have sought the appropriate tools and measurable

    variables for the test, but what is required is the exact functional form that is compatible with the

    analysis.

    Although a number of studies have been carried out to test the validity of the law, the

    nature of the original statement makes it harder to comprehend the relationship between

    economic progress and the growth of state activity. There are various interpretations of the

    hypothesis which result into six formulations given the ambiguous nature of Wagner s original

    description. They are as follows:

    Peacock and Wiseman (1967): G = f(Y) (1)

    Goffman (1968): G = f(Y/N) (2)

    Gupta (1967): G/N = f(Y/N) (3)

    Musgrave (1969): G/Y = f(Y/N) (4)

    Mann (1980): G/Y = f(Y) (5)

    Pryor (1968): GC = f(Y) (6)

    where G = real government expenditure, Y = real GDP, G/N = real government expenditure per

    capita, G/Y = ratio of real government expenditure to real GDP, Y/N = real GDP per capita and

    GC = government consumption.

  • 8/2/2019 Masters Essay Final Draft

    14/39

    8

    The first equation was used by Peacock and Wiseman (1967), Musgrave (1969), and

    Goffman and Mahar (1971). The second equation was employed by Goffman (1968) and Mann

    (1980). The third equation was suggested and formulated by Gupta (1967) and Michas (1975)

    and tested by Mann (1980) and Henrekson (1993). The fourth formulation was employed by

    Musgrave (1969), Murthy (1993) and Ram (1987). The fifth equation was suggested and tested

    by Mann (1980), and the formulation is known as Peacock-Wiseman share version. The sixth

    formulation was stated by and tested by Pryor (1968).

    Economists have used all of the six formulations to test for the validation of Wagners

    Law. Wagner stated that culture and welfare spending were income elastic, and he has also

    assumed that a major part of public commodities and services are luxury items and hence, the

    public expenses in national income is income elastic. This assertion has been formulated from

    Wagners basic thoughts and assertion of the literature about state development and activity.

    Therefore, analysts have expected the income elasticity of public expenditure to surpass unity.

    Goffman (1968) and Bird (1975) infer that Wagners law is valid if the income elasticity of

    demand exceeded unity. However, Michas (1975) states that elasticity does not require

    exceeding unity to validate the law but rather zero; this notion depends on the choice of

    functional form by the analyst. For Equations 1 and 2, the estimate of the elasticity need to

    overcome unity, but for Equations 4 and 5, the elasticity estimate need to be more than zero to

    verify Wagners hypothesis.

    IV. EMPIRICAL LITERATURE

    Wagners law has been examined under empirically for various countries using both time

    series and cross-sectional data. The results differ from country to country. Dhawan, Biswal and

    Paul (2003) test Wagners Law for Canada and its ten provinces using Johansens maximum

    likelihood technique of cointegration and the Granger-causality procedure for the period 1960 to

    2003. Their results support the law for Canada and three of its biggest provinces, implying that

    GDP and government expenditure possess a long run equilibrium relationship. However,

    Granger causality analysis shows that GDP Granger-causes the government expenditure in the

    case of Canada and seven of its ten provinces.

  • 8/2/2019 Masters Essay Final Draft

    15/39

    9

    Biswal, Dhawan and Lee (1999) empirically test Wagnerian versus Keynesian

    hypotheses for Canada from 1950 to 1995. The authors use Engel and Grangers two-step

    cointegration and error correction technique to find the relationship, if any, between government

    expenditure and national income. The results show that Wagners law is valid when tested with

    total current government expenditure and total current expenditure on goods and services. The

    research shows evidence of short-run causation between total current government expenditure

    and GDP.

    Ahsan et al. (1996) tests Wagners Law (using equations (1) and (4)) for Canada and

    obtained coefficient estimates of the income variables between 1.4 and 0.5, suggesting support

    for the hypothesis that GDP growth and the relative size of the public sector possess a positive

    relationship. Chang, Liu and Caudill (2004) analyzes Wagners law for ten countries based on

    cointegration and error-correction modeling techniques for the period 1951-1995. These authors

    includes three of the emerging industrialized countries of Asia, South Korea, Taiwan and

    Thailand, and seven industrialized countries, Australia, Canada, Japan, New Zealand, USA, the

    United Kingdom, and South Africa. The study, which also used cointegration analysis and

    unidirectional Granger causality between income and government spending, shows that the

    Canadian economy does not seem to provide evidence in support of Wagners Law.

    Lamartina and Zaghini (2008) tests Wagners law for23 OECD countries which includes

    Canada and using an autoregressive distributive lag approach to cointegration, shows empirical

    evidence supporting the existence of a long run positive correlation between public spending and

    GDP growth in 23 OECD advanced economies. Granger causality test showed that GDP causes

    government expenditure in the case of all the OECD countries.

    The test by Kolluri, Panik and Wahab (2000), using time series data from G7 countries,

    shows that the law holds for some components of the government expenditure. Lamartina and

    Zaghini (2008) tests a panel cointegration analysis of the joint development of governmentexpenditures and economic growth in twenty-three OECD countries. Their analysis provides

    evidence of a structural positive correlation between public spending and per capita GDP which

    is fitting to the Wagners law.

  • 8/2/2019 Masters Essay Final Draft

    16/39

    10

    A recent study by Mohave (2009) that examine Wagners law based on disaggregated US

    state-local government expenditure for the period 1957-2006 has shown support towards the

    hypotheses. The author has used the equation suggested by Musgrave (1969). The analysis

    includes two methods of cointegration; Engel and Grangers two step procedure and bound

    testing. Both the procedures show results supporting the Wagnerian hypotheses. Kalam and

    Aziz (2009) have carried out analysis testing the validity of Wagners Law for Bangladesh from

    1976 to 2007. The study supports Wageners law in the state in both the long run and short run.

    They conclude that population size plays a vital role for government expenditure to grow in the

    long and short run.

    Ziramba (2008) has examined the long-run connection between real government

    expenditure and real income for South Africa using the autoregressive distributive lag approach

    to cointegration, and then uses Granger non-causality test procedure improvised by Toda and

    Yamamoto (1995) to test for the link between the two variables. He found a long-run relationship

    between real government expenditure and real income, but different result in the short-run.

    Kumar, Webber and Fargher (2008) tests cointegration and causality tests to investigate the

    validity of Wagners Law for New Zealand for the period 1960-2007 between government

    expenditure on the one hand, and GNP and GDP on the other. The authors conclude that the

    Wagners law holds in the long run irrespective of the chosen output measure.

    Shelton (2007) tests several leading hypotheses on determinants of government

    expenditure, and one of them is Wagnerian hypotheses. The author uses Government Financial

    Statistics Data from IMF covering 100 countries for the period 1970 till 2000. Shelton (2007)

    analyzes that the correlation between income and government size is greatly driven my

    demographics, which includes population over 65 years, during the period 1970-2000. He gives a

    new explanation on Wagners Law that fragmentation leads to decentralization rather than

    decline in public expenditure. Shelton (2007) shows evidence that the evidence of the franchise

    affects the degree of redistribution in advanced democracies, and puzzle role of trade openness.

    Sinha (2007) tested the law using the total and per capita GDP and government expenditure data

    of Thailand. His empirical study suggests that there is no causality stimulating from either

    direction between GDP and government expenditure, and there is a weak relationship between

  • 8/2/2019 Masters Essay Final Draft

    17/39

    11

    the two variables in the long run. Hence, he concludes that Wagners law is not valid for

    Thailand.

    V. EMPIRICAL FRAMEWORK AND DATA

    In this paper, equation (3) suggested by Gupta (1967) will be considered as the model to

    test Wagners law for Canada from 1961 till 2009.

    )/(/ NYfNG (3)

    We assume that the function is log-linear specified as follows:

    uNYNG )/ln()/ln( (7)

    where G represents real government expenditure, Y is real income, N is the population, G/N =

    real government expenditure per capita, Y/N = real GDP per capita, and are parameters to

    be estimated, u is the random error term and ln denotes natural logarithm.

    The analysis is carried out using annual data from 1961 to 2009. The data for income,

    government expenditure and population were obtained from Statistics Canada using Canadian

    Socio-Economic Information Management System (CANSIM).1 Summary statistics of the

    variables used are presented in Table 1.

    Table1: Summary statistics of log of real GDP per capita and log of real government expenditure

    per capita.

    Log of Real GDP per

    capita

    Log of Real

    Government

    Expenditure per

    capita

    Mean 10.16946 8.685767Standard Deviation 0.2910891 0.2382136

    Minimum 9.553675 8.120816

    Maximum 10.58905 8.999946Count 49 49

    1 CANSIM is the Statistics Canada's computerized database of time series covering a wide variety of social andeconomic aspects of Canadian life.

  • 8/2/2019 Masters Essay Final Draft

    18/39

    12

    The series of real government expenditure and real GDP are at the annual rate. The real

    government expenditure and real GDP is measured at market prices and at constant (2000=100)

    prices. A main reason behind carrying out the analysis using annual data is because government

    expenditure is hardly sensitive to seasonal and even cyclical fluctuations, and therefore reference

    periods of less than a year may not be sufficient to register the response between government

    expenditure and national income. On the other hand, a reason that supports the use of time series

    framework was given by Bird (1971). He states that the formulation to test Wagners law has no

    ability to compare two nations given their government expenditure and the level of average per

    capita income. Bird shows evidence which shows that an increasing trend over time is quite

    different from a maximum and minimum value at a point in time. Therefore, Birds analysis

    shows that conclusion made from international cross sectional studies are not required to test

    Wagnerian hypotheses. Wagners law is unquestionably a time series incident. Hence, in this

    study we will be using time series technique to verify for the validity of Wagners law in Canada

    for the time period 1961-2009.

    VI. METHODOLOGY

    (a) Unit Root tests

    First, the stationarity of the time series, ln(G/N) and ln(Y/N) is investigated. An

    important choice is to determine, before conducting unit root tests, the most appropriate form ofthe trend in the data. If the data are trending, then some form of trend removal is required. Two

    common trend removal or de-trending procedures are first-differencing and time-trend

    regression. First-differencing is appropriate for integrated of order one (I(1)) time series and

    time-trend regression is appropriate for trend stationary integrated of order zero (I(0)) time series

    (Maddala, 1998). Unit root tests can be used to determine if trending data should be first

    differenced or regressed on deterministic functions of time to render the data stationary.

    Moreover, economic and finance theory often suggests the existence of long-run equilibrium

    relationships among non-stationary time series variables. If these variables are I(1), then

    cointegration techniques can be used to model these long-run relations. Hence, the need to carry

    out unit root test has become mandatory in our analysis.

  • 8/2/2019 Masters Essay Final Draft

    19/39

    13

    In the econometrics literature, there are several tests that can be carried out to detect the

    presence of unit roots in time series data. In this study, unit roots are tested using both the

    Augmented Dickey-Fuller (ADF) and Philips-Perron tests (Maddala, 1998).

    The ADF test is based on estimating the test regression:

    p

    j

    tjtjttt yyDy1

    1 (8)

    whereyt is the time series in question, tD a vector deterministic term, p is the number of lagged

    difference terms, and jty is used to approximate the structure of error, so as to ensure that the

    residual termt

    is not serially correlated. Under the null hypothesis 1 which implies the

    presence of unit roots.

    Phillips and Perron (Ng and Perron, 1995) develops a number of unit root tests that have

    become popular in the analysis of financial time series. The Phillips-Perron (PP) unit root tests

    differ from the ADF tests mainly in how they deal with serial correlation and heteroskedasticity

    in the errors. In particular, where the ADF tests use a parametric autoregression to approximate

    the ARMA structure of the errors in the test regression, the PP tests ignore any serial correlation

    in the test regression. The test regression for the PP tests is

    tttt uyDy 1 (9)

    wheret

    u is trend stationary and may be heteroskedastic. The PP tests correct for any serial

    correlation and heteroskedasticity in the errorst

    u of the test regression by directly modifying the

    test statistics 0t and T . Under the null hypothesis 0 which implies the presence of unit

    roots in the time series. One advantage of the PP tests over the ADF tests is that the PP tests are

    robust to general forms of heteroskedasticity in the error term tu , and the user does not have to

    specify a lag length for the test regression.

  • 8/2/2019 Masters Essay Final Draft

    20/39

    14

    (b) Engel and Granger Cointegraten test

    The following test is suggested by Engel and Granger (1987). The authors proposes that

    the test should run regression analysis on variables integrated of order one, or I(1). Suppose we

    are testing between Gt and Yt for the following cointegration regression equations:

    ttt uYG 11 (10)

    ttt vGY 22 (11)

    In small samples, the test for cointegration may not be invariant to the choice of

    normalization and therefore both cointegration regressions are run to examine if the results are

    sensitive to normalization. In the above equations, we assume that both the variables are

    integrated of order one using first differencing of the variables, and ut and vtis the error term of

    the formulation. Engel and Granger suggest running regression analysis on Gt and Yt. The

    asymptotic distribution of and are not standard, but the method required to estimate and

    by ordinary least square (OLS) procedure, and then test for unit roots in the equations stated

    below:

    (12)

    (13)

    The null hypothesis for the unit root of the residual value implies that there is no

    cointegration. However, the limiting distribution does not match the distribution tabulated by

    Dickey and Fuller. The limiting distribution of this test, however, resembles the Dickey-Fuller

    distribution even though you require a separate table for each dimension of the regressor. Hence,

    in this study we will allow for dynamics in the residual and carry out the augmented Dickey-

    Fuller (ADF) test and Phillip-Perron test, and then use the table generated by MacKinnon (1991)to find the critical value.

    If we find the two variables to be cointegrated, then we will move to the second step of

    Engel and Granger method. In this step, we acquire the error correction term (ECT) from the

  • 8/2/2019 Masters Essay Final Draft

    21/39

    15

    cointegrating equation, and use at least one period lag of ECT to test the direction of the

    causation of the following equation:

    ttittteuYGG

    111111 (14)

    ttittt uGYY 122122 (15)

    ttittt vYGG 133133 (16)

    ttittt vGYY 144144 (17)

    where Gtand Yt are the variables to catch the short run relationship, and et, t, t, and t are

    the new error terms of the error correction models.

    (c) Granger Causality test

    The first try at testing the direction of causality was by Granger in 1969. The intuition

    behind the Granger causality test is simple. If the variables are cointegrated, then the significance

    ofi (i = 1,2,3,4) of the equations 14, 15, 16 and 17 shows that the independent variable Granger

    causes the dependent variable. However, if the variables are not cointegrated, then the

    significance ofi (i = 1,2,3,4) shows the Granger causality among the variables.

    VII. EMPIRICAL RESULTS

    (a) Unit root tests

    In this study, we use traditional methods, which are Augmented Dickey-Fuller and

    Phillips-Perron tests, to test for non-stationarity of a time series. The variables examined are

    ln(G/N) and ln(Y/N). Table 2 shows the augmented Dickey-Fuller test and Phillip-Perron test

    results ran on ln(G/N) and ln(Y/N) to test the existence of unit roots in the time series.

    In our analysis, the null hypothesis is the presence of unit root. The critical values at the

    1%, 5% and 10% significance levels for the Dickey-Fuller test are -4.178, -3.512 and - 3.187

    respectively. The Augmented Dickey Fuller test statistic for log of GDP per capita is 1.905

    suggesting that the presence of unit root cannot be rejected, and that the time series for log of

    GDP per capita is non-stationary. Likewise, the Augmented Dickey Fuller test statistic for log of

  • 8/2/2019 Masters Essay Final Draft

    22/39

    16

    government expenditure per capita is 2.398. Therefore, we cannot reject the presence of unit

    root, and conclude that this time series is also non-stationary.

    Table 2: Test statistics to determine unit roots in the variables used in the study.

    Variables Augmented Dickey Fuller teststatistic

    Phillips-Perron test statistic

    ln (G/N) -2.398 -2.192

    ln (Y/N) -1.905 -1.956

    dependency rate 0.009 -0.581

    unemployment rate -2.649 -2.066

    ln (G/N) -3.758 -3.810

    ln (Y/N) -3.876 -3.785

    dependency rate -3.388 -3.217

    unemployment rate -4.330 -4.273

    For the Phillips-Perron test the critical value at the 1%, 5% and 10% significance levels

    are -4.178, -3.512 and -3.187 respectively. The test statistics, which yielded from log of GDP per

    capita, is -1.956 which lies at the acceptance region of 1%, 5% and 10%. Hence, this test also

    shows that we cannot reject the presence of unit root; therefore, the time series of log of GDP per

    capita is non-stationary. From Table 2, we found that Phillips-Perron test statistic of log of

    government expenditure per capita to be 2.192, which lies at the acceptance region of 1%, 5%

    and 10%. Therefore, we cannot reject the presence of unit root; thus, the respective time series is

    non-stationary.

    The results from the Dickey-Fuller and Phillips-Perron tests indicate that both variables

    are non-stationary. Therefore, if we employ classical econometric approach using ordinary least

    square (OLS) to carry the analysis with unit root presence, then the conclusion obtained will be

    misleading, and yield spurious result. To make the time series stationary, we take first

    differences of ln(G/N) and ln(Y/N).

    In our analysis, the null hypothesis is the presence of unit root. The critical values for the

    Dickey-Fuller test at 1%, 5% and 10% are 3.600, - 2.938 and 2.604 respectively. The test

  • 8/2/2019 Masters Essay Final Draft

    23/39

    17

    statistic of ))/ln((1tNG is 3.758 which do not lie at the acceptance region of 1%, 5% and

    10%. Therefore, we can reject the null hypothesis of the presence of unit root; hence, the time

    series data of log of government expenditure per capita is stationary when the first difference is

    taken into account. From the above test, we note that the log of government expenditure per

    capita is integrated of order one, or I(1). The Augmented Dickey Fuller test statistic of

    ))/ln((1tNY is 3.876 which do not lie at the acceptance region of 1%, 5% and 10%.

    Therefore, we can reject the null hypothesis of the presence of unit root; hence, the time series

    data of the log of GDP per capita is stationary when the first difference is taken into account.

    From the above test, we can clearly note that the log of GDP per capita is integrated of order one,

    orI(1).

    For the Phillips-Perron test, the critical values at 1%, 5% and 10% significance level are

    3.600, - 2.938 and2.604 respectively. The Phillips-Perron test statistic of ))/ln((1t

    NG is

    3.810 which do not lie at the acceptance region of 1%, 5% and 10%. Therefore, we can reject the

    null hypothesis of the presence of unit root; hence, the time series data of the log of GDP per

    capita is stationary when the first difference is taken into account. From the above test, we can

    clearly note that the log of government expenditure per capita is integrated of order one, or I(1).

    The Phillip-Perron test statistic of ))/ln(( 1tNY is3.785 which do not lie at the acceptance

    region of 1%, 5% and 10%. Therefore, we can reject the null hypothesis of the presence of unit

    root; hence, the time series data of the log of GDP per capita is stationary when the first

    difference is taken into account. From the above test, we can clearly note that the log of GDP per

    capita is integrated of order one, orI(1).

    (b) Engel and Granger Cointegraten Test

    Cointegraten suggests evidence of long term relationship between a nations activity and

    economic development. In this paper, we will verify the long run relationship between

    government expenditure and GDP. In our research, we apply Engel and Granger cointegration to

    find the establishment of the long term relationship between the respective variables used in the

    study. The primary requirement of Engel and Granger cointegration test needs the variables to be

    integrated of the same order. After carrying out Augmented Dickey-Fuller and Phillips-Perron

    test, we can infer that both the variables are integrated of order one, or I(1). The first step of

  • 8/2/2019 Masters Essay Final Draft

    24/39

    18

    Engel and Granger cointegration test requires us to run static ordinary least square (OLS)

    regression.

    tt NYNG ))/ln((7835204.07177868.0)/ln(

    R2 = 0.9167 CRDW=1.316853 (18)

    (2.05) (22.72)***

    The equation above shows that the log of government expenditure per capita and the log

    of GDP per capita are positively related. For every one percent increase in GDP per capita would

    lead to 0.7835204 % increases in government expenditure per capita. The t-statistic (in

    parentheses) of log of GDP per capita is 22.72, and also suggests that GDP per capita is

    significant even at the 1% significant level.

    tt NGNY ))/ln((169955.10075083.0)/ln(

    R2 = 0.9167 CRDW=1.227212 (19)(0.02) (22.72)***

    The equation above shows that the log of GDP per capita and the log of government

    expenditure per capita are positively related. For every one percent increase in government

    expenditure per capita would lead to 1.169955 % increases in GDP per capita. The t-statistic (in

    parentheses) of log of GDP per capita is 22.72, and also suggests that GDP per capita is

    significant even at the 1% significant level.

    The next step of Engel and Granger test requires us to obtain the estimated residuals of

    the OLS regression run above, and then carries out stationarity test on the estimated residuals.

    The MacKinnon critical value for the cointegrating equation of two variables without trend using

    forty-nine observations at the 1%, 5% and 10% levels of significance are4.1293, - 3.4637 and

    3.1321 respectively. After running stationarity test on the residuals of Equation 18, the ADF

    test statistic is 3.906 which do not lie at the acceptance region of 5% and 10%. The PP test

    statistic is6.967 which do not lie at the acceptance region of 1%, 5% and 10%. According to

    Dittmann (2002), PP test, applied to residual based cointegration determination, is more

    powerful than the ADF test. Hence, we can reject the null hypotheses of no cointegration, and

    the estimated residual is stationary. After running stationarity test on the residuals of Equation

    19, the ADF test statistic is 9.511 which do not lie at the acceptance region of 1%, 5% and

    10%. The PP test statistic is 7.923 which do not lie at the acceptance region of 1%, 5% and

  • 8/2/2019 Masters Essay Final Draft

    25/39

    19

    10%. Hence, we can reject the null hypotheses of no cointegration, and the estimated residual is

    stationary.

    From the following results, we can infer that government expenditure per capita and GDP

    per capita are cointegrated and do have a long run relationship. Our cointegration results confirm

    those by Ahsan et al. (1996), Biswal et al. (1999), Dhawan et al. (2003), and Lamartina and

    Zaghini (2008). From Equation 18, we can see that the coefficient of ln(Y/N) is positive and

    significantly different from zero. Hence, these studies yield results that support the notion that

    GDP growth and the relative size of government sector have a behavioral relationship with

    each other.

    In the second step, we are required to find the error correction term (ECT) since the two

    variables are cointegrated. The procedure requires the estimate the following two equations:

    (20)

    (20)

    (21)

    (21)

    where is the first difference operator, and and are the error correction terms, the first

    lag of error term in Equation 18 and Equation 19. If the error terms are not taken into

    consideration, then we will have a biased estimation in the respective procedure. The

    coefficients, , , and , are expected to illustrate the long run dynamics in Equations 20 and

    20. The coefficients, , , and , are expected to capture the long run dynamics in Equations

    21 and 21. The optimal lag length orders of the variables (M, S) are found to be three. The

    optimal lag length orders of the variables (K, L) are found to be two. Equations 20, 20, 21 and

    21 are then used to investigate the causal relationship between government expenditure and

    GDP to test the Wagnerian and Keynesian perspective.

  • 8/2/2019 Masters Essay Final Draft

    26/39

    20

    (c) Granger Causality test

    According to Biswal et al (1999), since the two variables are cointegrated, then the

    results of Granger causality are based on the significance of the coefficient of error correction

    terms, for Equations 20 and 20, and for Equations 21 and 21. When is significant then it

    implies that GDP Granger causes government expenditure which verifies Wagners hypothesis.

    When is significant then it implies that government expenditure Granger causes GDP which

    verifies the Keynesian paradigm.

    Table 3: Regression result of Equation 20, 21, 20 and 21.

    Variables (20) (21) (20) (21)

    0.1985218

    (1.18)

    -0.1173303

    (-0.64)

    0.2300227

    (1.36)

    -0.1087396

    (-0.59)

    0.2314485

    (1.45)

    -0.0887551

    (-0.53)

    0.2519814

    (1.55)

    -0.0857034

    (-0.51)

    0.1046356

    (0.71)

    _ 0.1295952

    (0.87)

    _

    0.3038138

    (2.16)**

    0.3101603

    (1.90)

    0.3522681

    (2.58)

    0.3519704

    (2.20)**

    0.1760605

    (1.10)

    -0.1259164

    (-0.71)

    0.1824272

    (1.13)

    -0.0901079

    (-0.51)

    -0.0789739

    (-0.52)

    _ -0.0629421

    (-0.41)

    _

    -0.0152288

    (-1.03)

    -0.0417051

    (-2.70)***

    _ _

    -0.0027062

    (-0.24)

    -0.0313511

    (-2.61)***R

    2 0.5350 0.3050 0.5226 0.3050

    AIC -5.45045 -5.05459 -5.45045 -5.05459

    *** shows that the variable is significant at 1%.

    ** shows that the variable is significant at 5%.

    * shows that the variable is significant at 10%.

  • 8/2/2019 Masters Essay Final Draft

    27/39

    21

    From Table 3, we can deduce that the error correction term of Equations 20 and 20 is not

    significant, even at the 10% significance level. Hence, the result provides us with evidence that

    GDP does not Granger cause government expenditure. In other words, the growth of GDP does

    not cause the growth in government expenditure; hence, the result shows that Wagners Law is

    not valid for Canada for the time period 1961 till 2009. According to Table 3, we can deduce that

    the error correction term of Equations 21 and 21 is significant at 1% significance level. Hence,

    the result provides us with evidence that government expenditure does Granger cause GDP. In

    other words, the growth of government expenditure causes the growth in GDP; hence, the result

    shows that Keynesian paradigm is valid for Canada for the time period 1961 till 2009. These

    results also show that normalization does not affect the result and the conclusion.

    VIII. WAGNERS LAW AND POPULATION STRUCTURE

    The research on Wagners Law rarely takes the population structure into consideration

    even though it affects expenditure. Shelton (2007) has found by carrying out research on cross-

    country panel data from 1970 to 2000 that Wagners Law is affected by demographics. A time

    series analysis by Durevall and Henrekson (2010) with larger samples for Britain and Sweden

    shows that the weight of the Wagners Law is lessen, since the expenditure associated to age

    structure is already integrated with the government expenditure.

    According to OECD Economic Outlook (2009), Canada is a welfare state. In other words,government provides the basic needs to the people. Hence, our research has undertaken the role

    of dependency ratio to verify Wagners Law. In the case of Canada, dependency ratio is

    calculated by summing up the population aged 0-17 and 65+, and then dividing by the

    population aged 18-64. According to Durevall and Henrekson (2010), the dependency ratio

    measure apprehends the expenditure on child care, school and pension, and also other spending

    of that sort.

    From Figure 5, we can deduce that the dependency ratio has a downward trend. The ratio

    has decreased following a constant downward trend from 1961 to 2009. The reduction in the

    dependency ratio explains that the public expenditure has declined as a share of GDP during the

    respective period of study. In this paper, the regressed equation below will be considered as the

  • 8/2/2019 Masters Essay Final Draft

    28/39

    22

    model to test Wagners law for Canada from 1961 till 2009 when age structure is taken into

    consideration.

    Figure 5: Trend of Dependency ratio for Canada from 1961-2009.

    .5

    .6

    .7

    .8

    .9

    1960 1970 1980 1990 2000 2010

    Year

    In our analysis, the null hypothesis is the presence of unit root. The critical values at the

    1%, 5% and 10% for the Dickey-Fuller test are -4.168, -3.508 and -3.185 respectively.

    According to Table 2, the Augmented Dickey Fuller test statistic for dependency ratio is 0.009

    which lies at the acceptance region of 1%, 5% and 10%. Therefore, we cannot reject the presence

    of unit root; thus, the time series for dependency ratio is non-stationary. To make the respective

    time series variable stationary, we will be employing the first-differencing approach. After

    carrying the first-difference, the critical value for the Dickey-Fuller test at the 1%, 5% and 10%

    significance levels are -4.178, -3.512 and -3.187 respectively. The ADF test statistic of

    ratiodependency _ is3.388 and the PP test is 3.217 and both the values do not lie at the

    acceptance region of 10%. Therefore, we can reject the null hypothesis of the presence of unit

    root; hence, the time series data dependency ratio is stationary when the first difference is taken

    into account. From the above test, we can clearly conclude that the dependency ratio is integratedof order one, orI(1).

    ratiodependencyNYNG _344798.1)/ln(2538529.0971322.6)/ln( R2=0.9655CRDW=1.362503

    (8.63) (3.66) (-8.07) (22)

  • 8/2/2019 Masters Essay Final Draft

    29/39

    23

    Equation 22 shows that the log of government expenditure per capita and the log of GDP

    per capita are still positively related, but dependency ratio is negatively related to log of

    government expenditure per capita. For every one percent increase in GDP per capita would lead

    to 0.2538529 % increases in government expenditure per capita which are lower than the

    coefficient found when OLS is run for ln (G/N) on ln (Y/N). The t-statistic of log of GDP per

    capita (in parentheses) is 3.66 and it is significant even at the 1% significant level. The t-statistic

    of dependency ratio is 8.07, and the variable is also significant at the 1% significance level.

    Based on the results reported in the previous section, we have already established

    cointegration for Canada from 1961-2009. Hence, we are using the dependency ratio to check the

    consistency during the respective period. The Engel and Granger two step cointegration analysis

    is carried out to check the long run relationship. In our analysis, the null hypothesis is no

    cointegration. The MacKinnon critical value for the cointegrating equation of three variables

    without trend using forty-nine observations at 1%, 5% and 10% level of significance are

    4.60176, - 3.9197 and - 3.58055 respectively. The ADF test statistic of 1u

    t is7.613, and PP

    test statistic is 5.802, and both the values do not lie at the acceptance region of 1%, 5% and

    10%. Hence, we can reject the null hypotheses of no cointegration, and the estimated residual is

    stationary. From the following result, we can infer that the respective variables do have a long

    run relationship.

    Since there is evidence of the presence of cointegration in the model, we now proceed to

    check the causality when dependency rate is a part of the model. Granger causality test is used to

    verify whether there is a causal relationship between real government expenditure, real income

    and dependency ratio. We have considered the following model to be used in our study to

    explain the government expenditure and national income:

    (23)

    Similar to the approach used in the previous section to check for the causality, we will

    rely on the significance of ofthe above equation. When is significant then it implies that GDP

  • 8/2/2019 Masters Essay Final Draft

    30/39

    24

    along with the dependency ratio Granger causes government expenditure which verifies

    Wagners hypothesis. The optimal lag length orders of the variables (M, S, P) is found to be four.

    From Equation 24, we can deduce that the error correction term is not significant even at

    the 10% significance level. Hence, the result provides us with evidence that GDP anddependency ratio does not Granger cause government expenditure. In other words, the growth of

    GDP along with the growth in the age structure does not Granger cause the growth in

    government expenditure; hence, the result shows that Wagners Law is not valid for Canada for

    the time period 1961 till 2009 even after the demographic variable, dependency ratio, is

    considered in the model.

    R2

    = 0.6929

    CRDW=2.003721

    AIC=5.40294

    (24)

    *** shows that the variable is significant at 1%.

    ** shows that the variable is significant at 5%.

    * shows that the variable is significant at 10%.

  • 8/2/2019 Masters Essay Final Draft

    31/39

    25

    Our conclusion differs to that of Shelton (2007) who states that the relation between

    government expenditure and GDP is formed due to the age structure in combination with social

    security spending. However, the establishment of Wagners law in Canada cannot be perceived

    when age structure is considered, since the age structure to a large extent is dependent on GDP

    per capita in the long run as demonstrated by the theory of demographic transition (Caldwell,

    1976). As a matter of fact, the current old age boom in Canada is a result of demographic

    transition. The result shows that the main implication of the Law is undermined by its

    characteristics, where dependency on age structure follows automatically when the government

    provides services that are dependent on age, such as pension, child-care, free education and so

    on.

    IX. WAGNERS LAW AND UNEMPLOYMENT RATE

    To my best knowledge, there are currently no empirical studies carried on Canada

    relating Wagners Law and another demographic variable, unemployment rate. However, the

    necessity of the respective variable to take a role in the model is seen, as it influences

    government expenditure side through unemployment benefits. Shelton (2007) has found by

    carrying out research on cross-country panel data from 1970 to 2000 that Wagners Law is

    followed when unemployment in combination with electoral rules and redistributive transfers

    affects the government spending.

    From Figure 6, we can note that unemployment rate has an upward trend. The

    recessionary period in the early 1980s and 1990s are captured in Figure 6 through the two peaks

    during the respective time. The recession brought on in the United States by the collapse of the

    dot combubble beginning in 2000, hurt the Toronto Stock Exchange but has affected Canada

    only mildly. It is one of the few times Canada has avoided following the United States into a

    recession. Hence, the slightly rise in the unemployment rate is noted in the early twenty-first

    century from the figure below.

    In our analysis, the null hypothesis is the presence of unit root. The critical values at 1%,

    5% and 10% significance level for the ADF test is -3.594, -2.936 and -2.602 respectively. From

    Table 2, we can deduce from the variables ADF test statistic that all three values lay at the

    acceptance region of 1%, 5% and 10%. Hence, we cannot reject the null hypotheses of the

    http://en.wikipedia.org/wiki/Early_2000s_recessionhttp://en.wikipedia.org/wiki/Dot_com_bubblehttp://en.wikipedia.org/wiki/Dot_com_bubblehttp://en.wikipedia.org/wiki/Dot_com_bubblehttp://en.wikipedia.org/wiki/Toronto_Stock_Exchangehttp://en.wikipedia.org/wiki/Toronto_Stock_Exchangehttp://en.wikipedia.org/wiki/Dot_com_bubblehttp://en.wikipedia.org/wiki/Early_2000s_recession
  • 8/2/2019 Masters Essay Final Draft

    32/39

    26

    presence of unit roots; thus, the variables are non-stationary for the period 1961-2009. After first-

    differencing the variable, the critical values at 1%, 5% and 10% obtained from ADF test is -

    3.600, -2.938 and -2.604 respectively. According to Table 2, test statistic values suggest that the

    variables are stationary in their first-difference. Hence, all three variables are integrated of order

    one, orI(1).

    Figure 6: Trend of Unemployment Rate for Canada from 1961-2009.

    4

    6

    8

    10

    12

    1960 1970 1980 1990 2000 2010

    Year

    Equation 25 represents the model after ln of government expenditure per capita is

    regressed against ln of GDP per capita and unemployment rate.

    ratentunemploymeNYNG _0197196.0)/ln(7125781.0288839.1)/ln( R= 0.9417

    CRDW=1.218635

    (3.99) (21.46)*** (4.45)*** (25)

    Equation 25 shows that the log of government expenditure per capita and the log of GDP

    per capita and unemployment rate are positively related. For every one percent increase in GDP

    per capita would lead to 0.7125781 % increase in government expenditure per capita which is

    lower than the coefficient found when OLS is run for ln (G/N) on ln (Y/N). . For every one

    percent increase in government expenditure per capita would lead to 0.0197196 increases in

    unemployment rate. The t-statistic for GDP per capita (shown in parentheses) is 21.46 and shows

  • 8/2/2019 Masters Essay Final Draft

    33/39

    27

    that GDP per capita is significant even at 1% significant level. The t-statistic for unemployment

    rate is 4.45 and also shows that unemployment rate is significant at 1% significance level.

    The MacKinnon critical value for the cointegrating equation of three variables without

    trend using forty-nine observations at 1%, 5% and 10% levels of significance are -4.60176, -

    3.9197 and -3.58055 respectively. The ADF test statistic for stationarity test of1

    u

    tis -8.507,

    and PP test statistic is -6.292, and both the values do not lie at the acceptance region of 1%, 5%

    and 10%. Hence, we can reject the null hypotheses of no cointegration, and the estimated

    residual is stationary. Since there is evidence of the presence of cointegration in the model, we

    now proceed to check the causality when unemployment rate is a part of the model. Granger

    causality test is used to verify whether there is a causal relationship between real government

    expenditure, real income and unemployment rate. We have considered the following model to be

    used in our study to explain the government expenditure and national income:

    (26)

    Similar to the approach used in the previous section, we will rely on the significance of

    of the above equation. When is significant then it implies that GDP along with the

    unemployment rate Granger causes government expenditure which verifies Wagners hypothesis.

    The optimal lag length orders of the variables (M, S, P) is found to be one.

    R2

    = 0.4386

    CRDW=2.373403

    AIC=5.3647

    (27)

  • 8/2/2019 Masters Essay Final Draft

    34/39

    28

    From equation 27, we can deduce that the error correction term is not significant even at

    the 10% significance level. Hence, the result provides us with evidence that GDP and

    unemployment rate does not Granger cause government expenditure. In other words, the growth

    of GDP along with the growth in the unemployment rate does not Granger cause the growth in

    government expenditure; hence, the result shows that Wagners Law is not valid for Canada for

    the time period 1961 till 2009 even after the demographic variable, unemployment rate, is

    considered in the model.

    X. CONCLUSION

    The aim of this research is to add to the literature on Wagners Law on Canada for the

    period 1961-2009. Among the theories of government growth, the preference towards Wagners

    Law has been given primarily importance because of the positive attributes of the hypothesisitself. The Law is provided with scientific simplicity, and amplified the long-run characteristics

    during the industrialization process. In our study, we have focused our attention at the two major

    macroeconomic variables, government expenditure and national income, to explain Wagners

    Law. We have implemented cointegration analysis through econometric procedure to explain the

    long-run relationship between the respective variables. The research also sought to find evidence

    of long-run relationship between these two variables when the demographic variables,

    dependency ratio and unemployment rate, were taken into consideration.

    We have used time series data to comprehend the purpose of our analysis. The time series

    properties of the data were assessed using Augmented Dickey-Fuller test and Phillips-Perron

    test. The tests found that both the variables along with the demographic variables were non-

    stationary in level, but stationary in first differences; hence, we can conclude that the log of

    government expenditure per capita and log of GDP per capita, and the demographic variables are

    integrated of order one, or I(1). The Engel and Granger two-step cointegration approach was

    employed to find evidence of long-run relationship existence between national income and

    government expenditure. At first, the analysis was carried using the formulation suggested by

    Gupta (1967), and later on checking the long-run relationship by controlling the particular

    demographic variables. In this study, cointegration tests find evidence of long-run relationship

    between income and expenditure for all the models in this paper.

  • 8/2/2019 Masters Essay Final Draft

    35/39

    29

    Granger causality test has been employed to check for evidence of relationship between

    the variables in the short run. When tested for Musgrave (1969) formulation only, we find that

    the relationship between government expenditure and national income is Keynesian where GDP

    per capita is dependent on government expenditure per capita, but not Wagnerian where

    government expenditure per capita is dependent on GDP per capita. One possible reason for the

    Law being rejected during the period 1961-2009 is the restrictive fiscal policies implemented by

    the government of Canada in order to cope with the augmenting national debt. Another possible

    reason as stated by Afxentiou and Serletis (2002) is that the rejection of Wagners hypothesis

    may be due to the changing structure of the Canadian economy which is changing the direction

    towards service oriented sector. To end our discussion, we can conclude from our empirical

    analysis that Wagners Law is not a conventionalized way that shows how the Canadian

    economy behave in the long-run, nor it explains how government activity should perform.

  • 8/2/2019 Masters Essay Final Draft

    36/39

    30

    REFERENCES

    Abdulnasser H. J. and Manuchehr I. (2002). On the causality between exchange rates and stock

    prices: A note.Bulletin of Economic Research, 54(2): 197-203.

    Afxentiou, P. and Serletis, A. (2002). Macroeconomic Policy in the Canadian Economy. Boston,

    Kluwer Academic Publishers.

    Ahsan, M., Kwan, A. C. C. and Sahni, B. S. (1996). Cointegration and Wagners hypothesis:

    time series evidence for Canada,Applied Economics, 28, 10551058.

    An, C. B., and Jeon, S. H. (2006). Demographic change and economic growth: An

    inverted-U shape relationship.Economics Letters. 92, 447.

    Banarjee, A, Dolado, J., Galbraith, J. W., and Hendry, D. F. (1993). Cointegration, errorcorrection and the econometric analysis of non-stationary data. Oxford: Oxford University Press.

    Bird, R. M. (1971). Wagners Law of Expanding State Activity. Public Finance 26(1), 126.

    Bird, R. M. (1972). The Displacement Effect: A Critical Note. Finanzarchiv 30(3), 454463.

    Biswal, B., Dhawan, U., and Lee, H. Y. (1999). Testing Wagner versus Keynes using

    disaggregated public expenditure data for Canada.Applied Economics. 31, 1283-1291.

    Caldwell, J. C. (1976). A Restatement of Demographic Transition Theory. Population

    and Development Review 2(3/4), 321366.

    Cameron, David (1978). The Expansion of the Public Economy: A Comparative Analysis.

    American Political Science Review 72(4), 12431261.

    Chang, T. (2002). An econometric test of Wagners Law for six countries based on

    cointegration and error correction modeling techniques,Applied Economics, 34, 1157-1170.

    Chang, T., Liu, W. and Caudill, S. B. (2004). A re-examination of Wagners law for ten

    countries based on cointegration and error correction techniques.Applied Financial Economics,

    14: 577-589.

    Chow, Y. F., Cotsomitis, J., and Kwan, A. (2002). Multivariate cointegration and causality tests

    of Wagner's hypothesis: evidence from the UK.Applied Economics. 34, 1671-1677.

    Dhawan, U., Biswal, B. and Paul, S. (2003). Testing Wagners Law for Canada and Its

    Provinces: A Cointegration Analysis. International Journal of Applied Economics and

    Econometrics, July-September 2003, v. 11, iss. 3, pp. 435-54.

  • 8/2/2019 Masters Essay Final Draft

    37/39

    31

    Dickey, D. and Fuller, W. (1979). Distribution of the Estimators for Autoregressive Time Series

    with a Unit Root.Journal of the American Statistical Association, 74, 427-431.

    Dipendra, S. (2007). Does the Wagners Law hold for Thailand? A Time Series Study.Munich

    Personal RePEc Archive. Paper No. 2560, posted 07.

    Dittmann, I. (2002): Residual-Based Tests for Fractional Cointegration: A Monte Carlo Study,

    Journal of Time Series Analysis, Volume 21 Issue 6, Pages 615647.

    Durevall, D., and Henrekson, M. (2010). The futile quest for a grand explanation of long-

    run government expenditure. Gteborg, Department of Economics, School of Business,

    Economics and Law at University of Gothenburg.

    Engle, R. F. and Granger, C. W. J. (1987) Co-integration and error-correction: representation,

    estimation and testing,Econometrica, 55, 25176.

    Goffman, I. J. (1968). On the empirical testing of Wagners law: A technical note. PublicFinance, 23: 359-364.

    Granger, C. W. J. (1969). Investigation causal relationship by econometric models and cross-

    spectral methods.Econometrica, 37: 424-438.

    Gupta, S. (1967). Public expenditure and economic growth: A time series analysis. Public

    Finance, 22: 423-461.

    Henrekson, M. (1993). Wagner's Law A Spurious Relationship? Public Finance 48(3), 406

    415.

    Islam, A. M. (2001). Wagners law revisited: Cointegration and exogeneity tests for the USA.

    Applied Economics Letters, 8: 509-515.

    Karagianni, S., Pempetzoglou, M. and Strikou, S. (2002). Testing Wagners Law For The

    European Union Economies. The Journal of Applied Business Research, 18(4): 107-114.

    Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. New York:

    Harcourt, Brace and Co.

    Kolluri, B. R., Panik M. J., Wahab, M. S. (2000). Government expenditure and economic

    growth: Evidence from G7 countries.Applied Economics, 32; 1059-68.

    Kumar, S., Webber, D. and Fargher, S. (2008). Wagners Law Revisited: Cointegration and

    Causality tests for New Zealand.Department of Business Economics, Auckland University of

    Technology, New Zealand, JEL: C22; H50.

  • 8/2/2019 Masters Essay Final Draft

    38/39

    32

    Lamartina, S. and Zaghini, A. (2008). Increasing Public Expenditures: Wagners Law in OECD

    Countries. Center for Financial Studies. CFS Working Paper No. 2008/13.

    Maddala, G. S. (1998). Unit Roots, Cointegration, and Structural Change. New York: Cambridge

    University Press.

    Mahdavi, S. (2009).A re-examination of Wagner's law based on disaggregated U.S. state-local

    government expenditure. San Antonio, Tex, UTSA, College of Business.

    Mann, A. J. (1980). Wagners law: An econometric test for Mexico, 1925-76.National Tax

    Journal, 33: 189-201.

    Michas, N. A. (1975). Wagners law of public expenditures: what is the appropriate

    measurement for a valid test, Public Finance/Finances Publiques, 30, 7784.

    Musgrave, R. A. (1969). Fiscal Systems. New Haven and London: Yale University Press.

    Ng, S., and P. Perron (1995). Unit Root Tests in ARMA Models with Data-Dependent Methods

    for the Selection of the Truncation Lag.Journal of the American Statistical Association. 90,

    268-281.

    Ng, S., and P. Perron (2001). Lag Length Selection and the Construction of Unit Root Tests

    with Good Size and Power,Econometrica, 69, 1519-1554.

    Peacock, A., and Scott, A. (2000). The Curious Attraction of Wagner's Law. Public Choice.

    102, 1-17.

    Peacock, A. T. and Wiseman, J. (1967). The Growth of Public Expenditure in the UnitedKingdom. London: Allen and Unwin.

    Pesaran, M. H., and Y. Shin 1999. An autoregressive distributed lag modeling approach to

    cointegration analysis. In Storm, S. (Ed.) Econometrics and Economic Theory in the 20th

    Century: The Ragnar Frisch Centennial Symposium. Cambridge University Press.

    Pesaran, M. H., Shin, Y. and Smith, R. J. (2001). Bounds testing approaches to the analysis of

    level relationships.Journal of Applied Econometrics, 16(3): 289-326.

    Pryor, F. L. (1968). Public Expenditures in Communist and Capitalist Nations. London: George

    Allen and Unwin.

    Ram, R. (1987). Wagner's Hypothesis in Time-Series and Cross-Section Perspectives: Evidence

    from "Real" Data for 115 Countries. The Review of Economics and Statistics. 69, 194-204.

    Singh, B. and Sahni, B. S. (1984). Causality between public expenditure and national income.

    Review of Economics and Statistics, 66(4): 630-644.

  • 8/2/2019 Masters Essay Final Draft

    39/39

    Singh, B. and Sahni, B. S. (1984). Causality between public expenditure and national income.

    Review of Economics and Statistics, 66(4): 630-644.

    Shelton, C. A. (2007). The size and composition of government expenditure.Journal of

    Public Economics. 91, 2230.

    Toda, H. Y. and Yamamoto, T. (1995). Statistical inference in vector autoregressions with

    possibly integrated processes.Journal of Econometrics, 66: 225-250.

    Liu, C. W., and Caudill, S. B. (2004). A re-examination of Wagners law for ten countries based

    on cointegration and error-correction modelling techniques.Applied Financial Economics, 2004,

    14, 577589.

    Verbeek, Marno (2000). A Guide to Modern Econometrics. West Sussex: John Wiley & Sons

    Ltd.

    Wagner, A. (1883). Three extracts on public finance. In R. A. Musgrave and A. T. Peacock (eds)(1958), Classics in the Theory of Public Finance. London: Macmillan.