Unemployment Convergence among the 36 States in Nigeria

Being a Revised Paper Presented by:

Michael O. Nyong
Department of Economics,
University of Calabar, Nigeria

At the

Finance and Economics Conference 2013

in Frankfurt am Main, Germany
Thursday July 4th to Saturday July 6th

1
Abstract:

Given the spate of bombings, kidnapping, armed robbery and terrorism in

Nigeria there has been renewed efforts to tackle the high unemployment rate in the
country. This is more so because vast differences in unemployment rates among
the states are indicative of regional economic inequality and poverty. This paper
investigates the dynamics of unemployment rates among the states in Nigeria and
the prospects for convergence with the national rate. We use more theoretically
defensible unit root tests with structural breaks and the autoregressive fractional
integrated moving average (ARFIMA) approach. Our results show that behavior
of unemployment rates of thirty(30) states are consistent with the
structuralists/NAIRU thesis given that they are stationary with breaks. We also
investigate the prospects for convergence in relative unemployment rate among the
states. We find some evidence of a process of conditional convergence among ten
states with the national unemployment rate. However we find some evidence of
non-convergence among 27 states. For these states the fractional integration
parameter d exceeded unity. The fact that ten states display stochastic convergence
suggests that at least part of Nigeria is converging. This means that poorer states
are on the average catching-up to the national unemployment rate. The non
convergence of relative unemployment differentials of most of the states is
indicative of the long lasting problem of unemployment which may not go away.
Policies aimed at reducing relative unemployment differences are likely to make
much impact when they are substantial. These include substantial investment in
infrastructural rehabilitation in electricity, road and rail network, development of
entrepreneurial skills among the youth and government provision of start-up grant
or loans.

2
 Unemployment Convergence among the 36 States in Nigeria

Introduction

Terrorism, kidnapping, armed robbery and the spate of bombings in

Nigeria of recent has brought to greater attention the challenge of unemployment
in the country. Some analysts (Mohammed 2012) have argued that the high rates of
unemployment is the grease that promotes and propagate not only the incidence of
poverty but also crime and insecurity. According to Mohammed the “the hallmark
of poverty and crimes in Nigeria” is the level of “unemployment among its active-
age citizens”. If Nigeria can tackle the high rate of unemployment in the country
by creation of productive employment, the battle against poverty and crime would
have been won. The view by Mohammed has rekindled a troubling reality of an
alarming employment crisis in Nigeria despite improved economic growth(i.e
non-inclusive growth). According to the National Bureau of Statistics (NBS 2010)
the high incidence of unemployment amid economic growth is indicative of major
socio-economic dislocations in the country.

Unemployment rate has witnessed persistent rise over the years in Nigeria.
Average unemployment rate in the past four decades (1970-2011) is about 9.67
percent, 5.52 percent during the period 1970-1989, and two-digit value at 13.46
percent during the period 1990-2011(CBN 2002, 2007; NBS 2010, 2009,2012;
World Bank 2005). The national unemployment rate masks significant disparities
or differences among the states’ unemployment rate. For instance, while the
national unemployment rate was 19.7% in 2009, the unemployment rate for
Bayelsa was 38.4%, Katsina 37.3%, Bauchi 37.1%, Akwa Ibom 34.1%, and
Gombe 32.1%. The sectoral decomposition of unemployment rate with respect to
age-group is even more revealing. The unemployment is highest for the youth 15-
24 (age –group) at 41.6%, followed by the 25 – 44 age group at 17%, 60 - 64 age
group at 16.7% and 45 - 59 age group at 11.5% . With respect to educational
group the records show that the unemployment rate is 20.1% for those who never
attended school, 22.3% for below primary school, 14.8% for primary school,
23.8% for secondary school and 21.3% for 21.3% for post secondary education.
According to International labour Organization Report (2010) graduate

3
unemployment which was 26.6% in 2003 rose to about 40.3% in 2009. This is
unacceptable.

Several ad hoc policies and programmes aimed at reducing the rate of

unemployment in Nigeria have been implemented over the past years with minimal
effect. These include the National Directorate of Employment (NDE), an agency of
government established in March 1986 to create jobs for the teeming army of the
unemployed by designing and implementing appropriate programmes to combat
mass unemployment; the Small and Medium Scale Enterprise Development
Agency of Nigeria (SMEDAN); the National Poverty Eradication Programme
(NAPEP), the National Economic Emancipation and Development Strategy
(NNEDS) of 2003, among others. Inspite of these programmes, unemployment
rates in the country and among the states have continued to rise, thereby casting
serious doubt about the effectiveness of the agencies and their programmes.

Persistency in subregional unemployment rate may reflect stable equilibrium

differentials in unemployment rates or may be attributed to the fact that shocks to
subregional unemployment rates have long-lasting effect (Martin 1997; Bayer and
Juessen 2006). Discriminating among the two cases is important because policy
interventions are more likely to be effective in the case of long-lasting effect than
in stable equilibrium differentials.

The objective of this study is to investigate the hysteresis or persistence effect

in the unemployment rates of the 36 States and Federal Capital Territory (FCT) in
Nigeria and determine the prospects for stochastic convergence of unemployment
at lower levels. To the best of my knowledge this is the first systematic attempt to
do so in Nigeria. This is necessary for five reasons. First, the national
unemployment rate is uninformative, and perhaps misleading as it masks vast
unemployment differentials among the states and regions. Second, wide and
growing disparities in unemployment rates among the states threaten socio-
economic cohesion, compromise national unity and economic integration.
Understanding the dynamics and persistence of subnational (states) unemployment
differences helps to assess their consistency with established labour market models
and the effectiveness of national and subnational economic policies. Third, high
unemployment rates have not only economic, but also political and social
consequences including labour market reforms (Layard et al., 2005). Fourth,

4
regardless of the integration order of the unemployment series, it is important to
study convergence to determine whether the states have something in common
which attract them in the long-run. Finally, there has been limited studies on
convergence of unemployment rates among the 36 states in Nigeria. The few that
exist such as Obadan and Odusola(2001), Muhammed et al. (2011), Bankole
(2002) and Abiodun (2010), Samuel et al (2011) Aiyedogbon and Ohwofasa(2012)
focused primarily on unemployment and productivity in Nigeria and graduate
unemployment at the national level. Aiyedogbon and Ohwofasa(2012) for example
investigate the relationship between poverty and youth unemployment at the
national level in Nigeria. They find a positive and statistically significant effect of
unemployment on the incidence of poverty. Our study is the first of its kind to
investigate the dynamics of sub-national unemployment rates and prospects for
convergence using disaggregated data. Furthermore the results of this study may
serve as a benchmark of what to expect about long-run behavior of unemployment
rates in the ECOWAS subregion. The study covers the period 1990 - 2011. The
choice of the period is dictated by the rising rate of unemployment during the
period, the availability of unemployment data at the State level, the need to cover
both the military dictatorial regime period (1990-1999) and the civilian democratic
governance (1999-2011). Quarterly data are used.

The rest of this study is organized in five sections. The first had been the
introduction and background of study. Section II provides the theoretical
framework and literature review. In section III an overview of Nigerian
unemployment rate at national and State levels is articulated. Section IV presents
the analytical methodology. In section V presents the empirical results and
analysis. We conclude the paper in section VI with a summary of main results and
the policy implications.

II Theoretical Framework and Literature Review

One of the basic assumptions underlying macroeconomic theories is the

existence of equilibrium in economic variables. Shocks in the economy are
assumed to have temporary effects and the variables are expected to converge to
their long-run equilibrium level. This has given rise to debates in labour market
theories concerning the behavior of unemployment rate. There are two dominant
schools to the debate. The structuralists perspective a la Phelps(1967) and

5
Friedman(1968) maintains that changes in economic fundamentals (shocks) may
bring about movement or shifts in unemployment around its natural rate and have
no permanent effect whatsoever on the natural rate. By implication unemployment
rate is stationary around the natural rate. According to structuralists although
economic fundamentals like interest rate, government capital expenditure and
productivity may alter the equilibrium rate, there is an inherent tendency to return
to equilibrium after some time. They insist that once structural breaks are
accounted for while testing for stationarity of unemployment rate, it will yield a
stationary process. The structuralists school perspective is a relaxed version of the
non-accelerating inflation rate of unemployment (NAIRU) thesis which states that
there is a long run unemployment rate and hence that shocks to unemployment
have only transitory effects. This is the view followed by Phelps(1994) and Phelps
and Zoega(1998). According to Phelps(1994) unemployment rate is a stationary
process around a small number of structural breaks (Lee, Strazicich and Tislau
2011) . Shocks to unemployment are temporary with autonomous movements in
the natural rate itself, due to changes in the structure of the economy.

In contrast to the structuralists school of thought the hysteresis school

argues that temporary shocks will have permanent effects on the equilibrium
unemployment rate, thereby leading to nonstationarity of the unemployment rate
(Blanchard and Summer 1986). The school insists that cyclical fluctuations
generated by business cycles have permanent effect on unemployment.
Unemployment is a random walk which never returns to equilibrium after shock.
While hysteresis insists on permanent effect, the persistence theory takes a more
relaxed view maintaining that unemployment does return to natural rate but needs
long periods to recover after a shock.

The policy implications for either school of thought are straightforward. For the
structuralists it suggests that government should focus on preventing short-run
departures from equilibrium. For the hysteresis school, it suggests the speed at
which unemployment can be reduced (Mikhail, 2002). It indicates the need for
macroeconomic stabilization policies , structural reforms, and design of social
safety nets to counter the negative shocks(Leon-Ledesma and McAdam, 2004). It
is well known that for countries with evidence of multiple equilibria, reforms
carried out to reduce unemployment benefits could constitute large and long-
lasting positive shocks if done during an employment recovery phase. However,
6
reforms carried during rising unemployment may not have the desired effect of
changing equilibrium unemployment as the positive labour market reforms could
be choked – off by the negative (demand or supply) shock affecting the economy.

When labour markets adjust towards equilibrium in the long run, there will be
convergence of subregional unemployment rates. Unemployed workers take jobs
in other areas because capital flows into low-wage regions to take advantage of
lower labour cost (Blanchard and Katz 1992, Bayer and Juessen 2006). However,
if the speed of adjustment is slow, disparities in unemployment may arise during
adjustment as a result of negative demand shocks affecting some regions more than
others (Armstrong and Taylor 2000).

Three main theories have been advanced to explain the sources and causes of
hysteresis - duration theory, insider-outsider theory and the capital stock theory.
Duration theory is concerned with the effect of unemployment duration on labour
market outcomes in terms of demand and supply of the unemployed. According to
duration theory, the longer the unemployment period, the less likely the
unemployed are likely to be offered jobs, as firms use their long period of
unemployment as a screening device to hedge them out of unemployment. The
insider-outsider theory maintains that insiders have bargaining powers which
enable them to negotiate and determine wages, retain their jobs and prevent
outsiders from being employed. However, Blanchard and Summers(1986) and
Gregory (1986) have shown that large negative shocks may lead to job loss among
insiders, thereby causing increase in unemployment rate. Indeed Gregory (1986)
was the first to argue that insider considerations could lead to sustained
unemployment in Austria. The capital stock theory suggests that adverse shocks
lead to contraction in firms’ capital and may eventually lead to business closure
and job losses. By implication, adverse capital stock shocks leads to increase and
persistent rise in unemployment rate.

The implication of the above analysis is that multiple equilibria may arise in
unemployment level due to any of the three sources indicated above. For example
in a study by Leon-Ledesma and McAdam(2004) they observe the presence of
high and low unemployment equilibria among Central and Eastern European
Countries (CEECs). The speed of adjustment is faster for CEECs than for the EU,

7
although CEECs tend to move more frequently between equilibria, indicating that
they do not remain locked into a new level after a shock.

A number of models have been developed to account for multiple

employment(unemployment) equilibria such as Aghion and Blanchard(1994)
expectation model in which the expectations of economic agents could lead to a
high unemployment equilibrium in developing countries and Boeri(2001)
microeconomic lock-in model in which skill specificity of workers and search
incentives could generate high persistence rise in unemployment rate. Inspite of
these models, it is the traditional inside-outsider model of unemployment
hysteresis of Blanchard and Summers(1986) that have been used in many studies.

Formally and in line with Leon-Ledesma and McAdam (2004), assume that
unemployment rate (y) follows a AR(K) process such that

m
Yt = ψ0 + ΣψkYt-k + ε . . . (2.1)
k=1
where t is time period, k is lag operator and ε is stochastic random shocks. It is
expected that unemployment will always converge to its natural or equilibrium
level which is defined as

ψ0
Y = ---------- . . . (2.2)
1 - Σψk

with the assumption that Σψk < 1. If this hold then unemployment is said to be
mean-reverting. Unemployment will follow random walk and hence non-
stationary, and exhibiting path-dependence if Σψk = 1. This is what is referred to
as pure hysteresis whereas partial hysteresis which occurs when Σψk ≈ 1 (see
Layard et al. 1991). Once unemployment follows random walk , then shocks ε
are assumed to have permanent effects on unemployment.

Given the fact many developing countries are characterized by shocks which
may lead to breaks in the trend or time paths of macroeconomic variables,
conventional unit root tests may not reveal the true nature of the data generating
process. Hence structuralists argue on the need to account for structural breaks in
testing for unit roots in macroeconomic data.

8
 Empirical studies on unemployment rate have concentrated on time series
econometrics by checking the order of integration of the unemployment rates.
Areti and Mariscal(1999), Lee and Strazicich and Tislau(2001) as well as
Camarero and Tamarit(2004) adopted this methodology, each one with some
peculiarities but all of them are based on OECD countries. Camarero and Tamarit
(2004) test the hysteresis vs natural rate of unemployment for a panel of 19 OECD
countries using sequential procedure based on two multivariate ADF panel unit
root test in SURE framework. The results confirm the validity of the natural rate
of unemployment hypothesis. Lee et al (2001) investigates the validity of the
hysteresis hypothesis using annual unemployment rate of 17 OECD countries for
the period 1955 – 1999. They adopted panel Lagrange Multiplier (LM) unit root
test which allows for heterogeneity in structural change (breaks) and find strong
rejection of the hysteresis hypothesis. Similar acceptance of NAIRU was obtained
in Camarero and Tamarit(2004) for a panel of 17 OECD countries. Blanchard and
Summers (1987), Brunello(1990), Mitchell(1993), Roed(1996), and Leon-
Lesdesma (2002) employ conventional univariate unit root tests to investigate the
behavior of unemployment rates in European Union (EU) countries. They find that
that unemployment exhibits hysteresis. However, empirical evidence on US
unemployment rates are inconclusive. Mitchell(1993), Breitung(1994), and
Hatanaka(1996) find non-convergence in US unemployment rate while Nelson and
Plosser (1982), Perron(1988), Xiao and Phillips (1997) report evidence in favour of
convergence.

Neudorfer et al. (1990) examine hysteresis or persistence effects in Austrian

unemployment based on unit root tests and model of wage-price formation. The
results show high persistence in unemployment caused by rising share of long-term
unemployed which leads to increase in overall unemployment rate.

Cheng et al. (2011) investigate the stochastic nature of unemployment rate

from a panel of 51 US state level data allowing for cross-section dependence for
the period 1976 Q1 to 2010 Q2. They find significant evidence of hysteresis
when more recent data are included and convergence when data from the Great
Recession are included. However, they question the practical usefulness of the
natural rate hypothesis given the fact that they obtain half-life point estimates
ranging from 6 to 14 years “after correcting for bias”. According to them the half–
9
life “appears quite long compared with typical duration of economic recession”.
Similarly Rowthorn and Glyn(2003) examine the convergence and stability in US
regional employment. They find that although regional labour markets in the US
are said to be highly flexible so that employment shocks have only transitory
effects on joblessness since induced migration quickly offset much of the initial
impact their empirical results show otherwise. The find that responsiveness of
employment rates to state specific shocks are rather weak implying that flexible
regional adjustment is not a factor accounting for “the success of monetary union
in USA”.

The importance of investigating subregional unemployment rates has been

documented by Marston (1985) based on the compensation theory which predicts
that in equilibrium all individuals have the same utility level, and so areas which
are more attractive could have larger unemployment rate. His empirical study
reveals that disturbances to equilibrium relationship in unemployment rates
among US metropolitan regions are likely to disappear because of mobility with a
particular year. Blanchard and Katz(1992) analyze the dynamics of the American
regional unemployment rate. Their findings indicate that when jobless individuals
move to other areas in search of work, there is adjustment of the labour market
towards long-run equilibrium – ie convergence in regional unemployment rates.
Specifically they find that both unemployment and labour force participation shock
returned to their original levels within about seven years of a state specific
employment shock and about four years after the change in unemployment had
reached its maximum. Thus, US absorb regional employment shocks in a highly
flexible manner with regional migration being the main channel of preventing
long-lasting effects of joblessness of state specific shocks in the US.

Wu (2003) studies the Chinese unemployment persistence through panel data

regression approach. The results show that (i) regional relative unemployment rate
show greater persistence when compared to nation-wide unemployment rate, (ii)
total unemployment is more persistent than youth unemployment, (iii) region with
the highest unemployment rate is the one with the lowest persistence of regional
unemployment, (iv) Chinese job seekers are usually driven to the private sector,
which has helped reduce the unemployment persistence.

10
 Oliveira and Careiro(2001) investigate the existence of equilibrium
relationship between states and national unemployment rates. Their findings
suggest that states and aggregate unemployment rate have similar dynamism, but
with permanent differences in the long – run. Assuming that states have different
attractiveness level, the result is consistent with Marston (1985) prediction that in
equilibrium all individuals have the same utility level, and so areas more attractive
could have larger unemployment rate.

Boeri and Terell(2002), Ferragina and Pastore(2005) inquire if the

differences in unemployment rate in transition economies could be explained by
the optimal speed of transition theory. Frek et al (2004) was more concerned about
the Czech Republic. In an influential study of the evolution of labour market in the
Czech Republic, Flek et al (2004) submit that over the period of ten years, a
transition from over-employment to underemployment may have taken place.

Katrencik et al. (2008) criticized previous studies in transition economies

arguing that they are not consistent with recent labour market developments
because they used fairly aggregate data rather than policy-relevant data. He also
submits that past studies failed to incorporate structural breaks in the labour
market condition and this failure might have compromised the results obtained.

Carmeci and Chies(2006) examine the role of structural breaks in Italian

unemployment hysteresis. The findings show that hysteresis is not sensitive to
with or without structural breaks and that the finding of hysteresis in national
unemployment rate in Italy is a result of aggregation of regional trend-stationary
unemployment rate characterized by heterogenous breaks.

In Brazil Gomes and da Silva(2006) examine the hysteresis hypothesis effect

of unemployment rates in six metropolitan areas using unit root test with structural
breaks. The results reveal that hysteresis is absent in all the regions except Rio de
Janeiro. Further analysis indicate the existence of stochastic convergence in all
the regions except Porto Alegre. Oliveira and Carneiro(1999) examine the short-
run and long –run relationship between national unemployment rate and
subregional unemployment rate and obtained mixed results. Corseuil, Gonzaga and
Issler (1999) investigate the short-run and long-run unemployment movements
across the six metropolitan regions in Brazil. The results were not conclusive. But

11
the study by de Figuerado(2010) were more definitive. De Figuerado examines the
dynamics of regional unemployment rates in Brazil using a fractional integration
approach. The results show that both persistence and non-stationary but mean-
reversion exist in Brazilian regional unemployment rates. His test of stochastic
convergence show that regional unemployment rates are convergent. There is no
similar study for Nigeria at subregional or state level. This is the point of departure
of this study.

Fallahi and Rodriguez(2010) investigates the degree of persistence of regional

unemployment rate in Canada using quarterly data for the period 1976:1-2005:4
based on endogenous structural breaks methodology. Their findings indicate that
the degree of persistence decrease with structural breaks and that the rigidity of
Canadian labour market, unemployment insurance benefits, inter-provincial
redistribution, attitude towards migration and industry structure are major causes of
persistence.

III Overview of Unemployment rate in Nigeria and the States

The Nigerian economy has experienced several and external shocks that have
resulted in severe distortions and caused structural changes. The resultant effect
had been fluctuations in major macroeconomic aggregates such as output,
inflation, balance of payments, consumption, investment, and unemployment rate.
Of major concern to policy makers is the upward trend in the unemployment rate
while other macroeconomic variables like real GDP growth and balance of
payments have shown significant improvement. Trends in unemployment rate had
assumed worrisome upward trend since 1997. With average national
unemployment rate of about 9.56 percent between 1990:1 and 1994:4 the
unemployment rate became double digits thereafter rising to 10.72 percent in the
next period (1995:1-1999:4), and to 18.03 percent in the period 2006:1- 2011:4 as
shown in Table 3.1. The movement in unemployment rate is better appreciated by
the quarterly trend as indicated in Figure 3.1. The figure reveals quite clearly the
rising and worrisome trend in the unemployment rate in Nigeria and the need for
realistic government policy intervention to stem the tide. National unemployment
rate attained its highest value in 2011:4 at 23.9 percent.

12
Table 3.1: Average National Unemployment Rate
Year Mean Maximum Average population unemployed

1990:1 - 2011:4 13.24 23.90 12.37 million

1990:1 - 1994:4 9.56 10.80 10.37 million

1995:1 - 1999:4 10.72 12.50 11.88 million

2000:1 - 2005:4 13.25 14.80 17.69 million

2006:6 - 2011:4 18.03 23.90 40.39 million

Source: computed from records of National Bureau of Statistics

Figure 3.1: Dynamics of national unemployment rate 1990:1-2011:4

NIGERIA
28

24

20

16

12

8
90 92 94 96 98 00 02 04 06 08 10

Fig 3.2:Dynamics of Disparities in Unemployment Rates among the States and FCT

13
 ABIA ADAMAWA AKS ANAMBRA BAUCHI
25 40 40 25 50

20 20 40
30 30

15 15 30
20 20
10 10 20

10 10
5 5 10

0 0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

BAYELSA BENUE BORNO CRS DELTA

250 30 30 25 30

200 25 20 25

20 20 20
150 15
15 15
100 10
10 10 10

50 5 5 5

0 0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

EBONYI EDO EKITI ENUGU FCT

24 40 25 30 12

20 20
30 8
16 20
15
12 20 4
10
8 10
10 0
4 5

0 0 0 0 -4
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

GOMBE IMO JIGAWA

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

Fig 3.2:Dynamics of Disparities in Unemployment Rates among the States and FCT continued

14
 KADUNA KANO KATSINA KEBBI KOGI
40 40 40 40 25

20
30 30 30 30

15
20 20 20 20
10

10 10 10 10
5

0 0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

KWARA LAGOS NASSARAWA NIGER OGUN

50 30 40 40 25

25
40 20
30 30
20
30 15
15 20 20
20 10
10
10 10
10 5
5

0 0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

ONDO OSUN OYO PLATEAU RIVERS

20 20 16 30 40

25
16
15 12 30
20
12
10 8 15 20
8
10
5 4 10
4
5

0 0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

SOKOTO TARABA YOBE ZAMFARA

40 50 40 80

40
30 30 60

30
20 20 40
20

10 10 20
10

0 0 0 0
1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

15
 Figure 3.2 indicates the dynamics of subnational (states) disparities in
unemployment rate. From the Figure we find that unemployment rate among the
states showed wide fluctuations with Bayelsa displaying the widest dispersion
with estimated standard deviation of 36.01, followed by Zamfara at 21.83, Bauchi
at 11.92, Taraba at 11.12, Katsina at 10.95, Gombe at 10.50 in that order. Others
include Yobe at 98.3, Jigawa at 9.19, Kano at 9.17 and Adamawa at 9.05.

Most of the states started with low rate of unemployment less than 10 percent in
first quarter of 1990. This rose with time to double digits becoming worse in the
2000s. However, six states started with higher unemployment rate above 10
percent in the first quarter of 1990 but grew to double digits and become worse in
the 2000s. These include Abia recording higher unemployment rate of about 20
percent in 2010, Ekiti with over 20 percent in 2010, Imo with about 30 percent in
2010, Nassarawa with over 38 percent in 2011, Kwara with about 38 percent in
2008 and Rivers with about 33 percent in 2010. A map showing the states in
Nigerian is provided at the appendix as Figure 3.3.

All the states experienced two or more structural breaks in their

unemployment trends. For instance Abia state experienced three structural breaks,
first in 1995, second in 2001 and third in 2006. Adamawa also experienced three
structural breaks (1995, 2005, 2008) but Akwa Ibom State (AKS) experienced two
structural breaks (1995, 2003). The rest are similarly interpreted. A common
structural break date for all the states is 1995. This may be indicative of effect of
the confusion in government policy - deregulation before 1994, regulation of the
economy in 1994, and back to deregulation of the economy again in 1995.

An interesting unemployment dynamics is that indicated by Bayelsa and

Zamfara States. Bayelsa consistently maintained low unemployment rate of less
than 5 percent until 2007 when the rate started rising hitting an all time high value
at 33 percent in 2010. Zamfara on its part displays a normal distribution in
unemployment rate with concentration of high unemployment rate between 1997
and 2007 recording the highest rate at 71.5 percent around 2003 and 2004.

Another interesting characteristic of the evolution of unemployment rate among

the states is the poor performance of the northern states which show higher
unemployment rates particularly in the period 2000-2011, an era of

16
democratization and civilian governance. It suggests inadequate investment by the
northern governors in infrastructure and in agriculture and lower proportion of
capital expenditure in total expenditure. Furthermore the higher level of
insecurity in the North may be responsible for increase in unemployment rate,
although it may also be argued that unemployment may also cause insecurity.

An important characteristics of state labour force is that the Southern States

have more educated labour force than the Northern states, and unemployment is
more severe in the North than in the South. The lack of education of appropriate
skills in the Northern States complicate the unemployment situation in the North.
The mismatch between skills and job availability has led to the appearance of long-
term unemployment for most of the Northern States.

The adoption of structural adjustment programme since 1986, the

retrenchment of civil servants in the public sector in the late 1980s and early
1990s, the distress in the financial system in the 1990s, the change from dictatorial
military regimes in the 1999 to civilian democratic governance, the consolidation
and recapitalization of banks in 2004/2005 as well as the implementation of the
national economic emancipation and development strategy (NEEDS) in the
2003/2004 means that the labour market has not been able to absorb these shocks
as expected by the creation of new jobs in the private sector and the improvement
of matching induced by more market friendly policies as indicated in NEEDS. This
had led to high unemployment observed in Nigeria and among the states with
persistence and long duration of unemployment spells.

It would be wrong to associate the persistence of unemployment with the

flexibility of labour market because the conventional factors used to explain
maladjustment such as union strength, minimum wages and employment
protection legislation are either weak or ineffectively implemented in Nigeria. The
absence of unemployment benefits acting as a wage floor may have discouraged
job reallocation. In addition, multiple equilibria which arise due to microeconomic
lock-in effects, owing to the excessive skill specificity would generate the
appearance of long spells of unemployment and regime shifts (Aghion and
Blanchard 1994).

17
Table 3.2 provides further characteristics of the structure of unemployment rate.
According to National Bureau of Statistics (2011) unemployment in Nigeria is
largely rural based, with high concentration of unemployment among the youth
and secondary school leavers. It indicates that about 67 million youths are
unemployed out of a population of 167 million in 2011. Worried by this trend the

Table 3.2: Structure of Nigeria’s Unemployment(2011)

Structure Unemployment Rate (%)

1999 2005 2009 2011

Educational level

No school 23.20 11.52 20.1 22.4

Primary 13.80 7.83 14.8 21.5

Secondary 57.60 11.70 23.8 33.4

Tertiary 5.40 9.52 21.3 20.2

Gender

Male 53.60 10.37 23.5

Female 46.40 14.65 24.3

Location

Rural 10.11 11.50 25.6

Urban 12.23 11.00 17.1

Age composition

15-24 59.40 26.01 41.6 37.7

25-44 27.70 10.83 17.0 22.4

45-59 7.60 7.10 11.5 18.0

Sources (i): National Bureau of Statistics: 2011 Socio-Economic Report.
(ii): National Planning Commission: Economic Performance Review 2006(April/July)

18
Federal Government recently embarked on The Youth Enterprise with Innovation
Scheme to generate jobs by encouraging entrepreneurship among the youths. It is
envisaged that over a 3-year period in which the second cycle will be completed
in September 2013, between 80,000 to 320,000 new jobs will be created at a cost
of N10 billion Naira. Government has also carried out women and youths
employment project out of the subsidy reinvestment and empowerment
programme(SURE-P) targeting employment of about 300,000 youths, women and
the disabled each year.

Various explanations for the rising trend in unemployment have been put
forward. These include the rural-urban migration in search of jobs in the urban
centers, retrenchment in the public sector in 1990s coupled with distress in the
financial sector, dilapidated infrastructure and epileptic electric power supply
leading to many industries to close shops, and low and sluggish economic
growth. Critical infrastructure in Nigeria had continue to decay across the nation.
The non-diversification of the economy away from the oil, means that the oil-
sector which is capital intensive enclave will continue to have little employment
generating potentials. As a result Nigeria is in danger of not achieving some of the
millennium development goals (MDGs).

Another dimension of the unemployment problem is the existence of high rate

of economic growth amidst high level of unemployment and incidence of
poverty. For example in 2011 the growth rate of real GDP was about 7.4 percent
but the unemployment rate was as high as 23.9 percent (AEO 2012). Thus, the
economy suffers from non-inclusive growth suggesting, the failure of the trickle-
down effect thesis in development policy and the need to diversify the economy
from oil to manufacturing and agriculture.

IV Analytical Methodology

4.1 Conventional Augmented Dickey-Fuller Test

19
 We first tested for degree of persistence in national and state unemployment
rates to determine the consistency of the behavior of Nigeria’s unemployment
rate with the structuralists or hysteresis school of thought. We adopted both the
traditional augmented Dickey Fuller test and endogenous unit root test with
multiple structural breaks indicated in equations 4.1 and 4.2-4.6 respectively.

Conventional ADF model takes the form:

∆Ut=ψ0 + ψ1T + ψ2Ut-1 + ∑bi ∆Ut-i + vt (4.1)

where Ut is the unemployment rate, T is the time trend, ∆ is the difference

operator, ψ0 ψ1 ψ2, bis are parameters to be estimated. The existence of unit root
is indicative of hysteresis and rejection of the unit hypothesis is confirmation of
the structuralists (or NAIRU) hypothesis.

4.2 Unit Root Test with Breaks

The Lee and Strazicich Langrange multiplier model of unit root test with two
endogenous structural breaks takes the form:

∆Ut=δ’ΔZt + Āt-1 + ∑ψiΔĀt-I + ut (4.2)

where Āt is a detrended series such that Āt = Yt – Ŝx – Zδ*, t=2,3,…T; δ* are the

estimated coefficients in the regression of ∆Yt on ΔZt; Ŝx is given by Yt - Ztδt; and
Y1 and Z1 represent the first observations of Yt and Zt respectively. Zt is a vector is
a vector of exogenous variables defined by the data generating process.
Corresponding to two-break equivalent of Perron of Perron’s(1989) model C, with
two changes in level and trend, Zt is described [1, t, D1t, D2t, DT1t, DT2t]’. where
Djt=1 for t>=TBj + 1, j=1,2 and zero otherwise; DTjt= t -TBj for t ≥TBj +1, j=1,2 and
zero otherwise. Note that the test regression (2) involves ΔZt insead of Zt so that
ΔZt becomes [1, B1t, B2t, D1t, D2t]’, where Bjt= ΔDjt and Djt= ΔDTjt, j=1,2. The unit
root null hypothesis is described in equation (2) by =0. And the test statistics are
defined as follows:

p=T. ’ . . . (4.3)

20
 τ = t-statistic for testing the unit root null hypothesis that =0.
(4.4)

To endogenously determined the location of the two multiple breaks (λj = Tbj/T,
j=1,2) the minimum LM unit root test uses a grid as follows:

LMp =Infλp(λ), (4.5)

LMτ =Infλτ(λ). (4.6)

The location of the structural break (TB) is determined by selecting all possible
break points for the minimum t-statistic. The search is carried out over the
trimming region (0.1T, 0.9T), where T is sample size. We set the maximum lag
length equal to eight (Kmax=8) and use the 10 per cent asymptotic critical value of
1.645 to determine the significance of the t-statistic on the last lag. After
determining the optimal lag length at each combination of breakpoints, we
determine the breaks where the endogenous two-break LM t-test statistic is at a
minimum. Critical values for the one break case are tabulated in LS(2004), while
critical values for the two break case are tabulated in LS (2003). This technique
has been shown to perform well as compared to other data-dependent
procedures that select the number of lagged augmented terms as seen in Phillips-
Perron (1988), Kwaitkowski et al. (1992), Ng and Perron(1995). It is also an
improvement over Zivot and Andrews (1992) single break and Perron (1997),
Lumsdaine and Papell (1997) double breaks methods.

4.3 Fractional Integration and Convergence Tests

After examining the unit root properties of unemployment rate we then

tested for convergence of state unemployment rate to the national average, given
Federal Government declared objective to fight unemployment in Nigeria. The
methodology adopted is stochastic convergence proposition of Carlion Mills
(1993) who study income convergence across U.S regions, and Bayer and Juessen
(2006) who investigated convergence in West German Regional unemployment
rates. We decided to adopt the theoretically defensible fractional integration
analysis which is more suited to analysis of convergence and dispenses with the
rigid knife-edge I(0)/I(1) framework.
21
 The fractional integration model adopted in is the parametric fractionally
integrated moving average (ARFIMA) model which is a more flexible testing
approach where the fractional integration parameter d can take any value, not
necessary 0 or 1 as in the traditional augmented Dickey-Fuller (1979), or Phillip-
Peron (1988) tests. It was introduced by Granger (1980), Granger and
Joyeux(1980), Hosking (1981) and belong to long memory process. It takes the
form:

(1-L)dyt =ut . . . (4.7)

(1 - L)d = 1 – dL + d(d-1)L2/2! - d(d-1)(d-2)L3/3! + . . . (4.8)

Invertibility is obtained whenever -1/2<d< 1/2. The process is stationary for

values of d in the interval (-1/2 1/2) (Notini and Lima, 2006). For values of d in
the interval (1/2 1) the process is non-stationary, but mean reverting. In
summary, for values of d less than unity unemployment shocks die out, even
when the yt process is non-stationary. We use the maximum Likelihood Estimate
due to Robinson (1994) and available in OX ARFIMA package. The Lagrange
Multiplier (LM) test due to Robinson (1994) is of the form:

Ho: d - d0 . . . (4.9)

in a model given by

yt=b1X1t + b2X2t . . . + bkXkt + wt . . . (4.10)

and (4.7), for any real d0, where yt is the time series we observe; b1, b2, . . . bk,,
are unknown parameters to be estimated, wt is a (k x 1) vector of deterministic
regressors that may include, for example an intercept, e.g wt=1, or an intercept
and a linear trend, e.g wt=(1,t).

The use of ARFIMA approach in this study is an improvement over traditional

time series unit root tests such as augmented Dickey-Fuller (ADF), Phillip-
Perron(PP)), Kwaitkwoski-Phillip –Schmidt- Shin (KPSS) and Ng Perron because it
accepts a continuum of integration orders not necessarily zero or unity (Dufrenot
et al., 2003; Beyaert, 2004; Cunado et al., 2004). This approach enables us to
discriminate between fractional long-run convergence and fractional catching-up
22
and hence to contribute to solving the puzzling negative or inconclusive results
about convergence obtained in time series context. Fractional integration is a
long memory approach in which deviations from the long-run mean decays more
slowly (hyperbolic decay) than exponential decay.

The ARFIMA model is estimated with constant α and trend( trend parameter β)
thereby providing a rich variety of nonlinear dynamics.

Case 1. Conditional Convergence (α ≠ 0, β = 0).

Case 1.1: Strict conditional convergence -0.5< d ≤ 0

Case 1.2: Long memory conditional convergence 0<d<0.5

Case 1.3: Long memory mean-reverting convergence 0.5<d<1.

Case 2.: Non-stationary process, absolute divergence d≥1

Case 3: Catching-up dynamics (α ≠ 0, β < 0).

Case 3.1: -0.5<d≤ 0 rapid catching –up or short- memory catching-up

Case 3.2: 0 < d <0.5 long memory catching-up. This occurs when a state spends
too much time on the transition path to the common long –run deterministic
trend.

Case 3.3: 0.5< d< 1 long-memory mean-reverting catching-up.

The model was estimated by maximum likelihood method using James Davidson
2002-2012 Time Series Modelling (TSM) version 4.35 computer software. We test
for convergence in unemployment rate among the 36 states and FCT for the
period 1990-2011 using quarterly data.

V Empirical Results and Analysis

The empirical results are presented in Tables 4.1, 4.2 and 4.3. Results from
traditional linear augmented Dickey-Fuller unit root tests are presented in Table

23
4.1. From the results we find that the national unemployment rate has double
unit roots thereby confirming the validity of the hysteresis hypothesis. Out of the
37 states including Abuja/FCT only 7 states namely Benue and FCT (North
Central), Cross River (South South), Ekiti and Osun (South West), Katsina and
Sokoto (North West) exhibit stationary unemployment rate and hence are
consistent with the structuralists school of thought while 31 states are non-
stationary and hence are characterized by hysteresis in their unemployment rate.

Table 4.1: Results of ADF unit root test of Hysteresis

T-
SN Country/State STATS I(d) Decision SN State T-STATS I(d) Decision
1 NIGERIA*** -2.769 I(2) Hysteresis 20 KADUNA* -9.552 I(2) Hysteresis
2 ABIA** -3.18 I(2) Hysteresis 21 KANO*** -2.84 I(1) Hysteresis
3 ADAMAWA** -3.319 I(1) Hysteresis 22 KATSINA* -4.43 I(0) Structuralists
4 AKS ** -3.027 I(1) Hysteresis 23 KEBBI* -6.827 I(2) Hysteresis
5 ANAMBRA* -8.064 I(2) Hysteresis 24 KOGI** -3.315 I(1) Hysteresis
6 BAUCHI** -3.046 I(1) Hysteresis 25 KWARA** -3.207 I(1) Hysteresis
7 BAYELSA** -3.014 I(2) Hysteresis 26 LAGOS* -5.248 I(I) Hysteresis
8 BENUE* -4.144 I(0) Structuralists 27 NASSARA* -9.168 I(2) Hysteresis
9 BORNO** -3.273 I(1) Hysteresis 28 NIGER* -9.123 I(2) Hysteresis
10 CRS** -3.349 I(0) Structuralists 29 OGUN* -6.889 I(2) Hysteresis
11 DELTA* -5.548 I(2) Hysteresis 30 ONDO** -3.073 I(1) Hysteresis
12 EBONYI** -3.014 I(1) Hysteresis 31 OSUN** -2.914 I(0) Structuralists
13 EDO* -6.496 I(2) Hysteresis 32 OYO*** -2.626 I(1) Hysteresis
14 EKITI*** -2.723 I(0) Structuralists 33 PLATEAU* -9.143 I(2) Hysteresis
15 ENUGU** -2.957 I(1) Hysteresis 34 RIVERS*** -2.87 I(1) Hysteresis
16 FCT* -3.863 I(0) Structuralists 35 SOKOTO** -2.711 I(0) Structuralists
17 GOMBE * -9.346 I(2) Hysteresis 36 TARABA* -6.443 I(2) Hysteresis
18 IMO*** -2.793 I(1) Hysteresis 37 YOBE* -8.872 I(2) Hysteresis
19 JIGAWA*** -2.794 I(1) Hysteresis 38 ZAMFARA* -8.008 I(2) Hysteresis
Notes: * significant at 1%, ** significant at 5%, *** significant at 10%
Critical Values of ADF without trend 1% -3.508, 5% -2.896 10% -2.585
NASSARA is NASSARAWA state, CRS is CROSS RIVER State, AKS is AKWA IBOM State.

Table 4.2: Results of unit root tests with endogenous structural breaks

24
 SN Country/State τ-STATS Break Date SN State τ –STATS Break Date
1 NIGERIA -6.562* 1996:4 20 KADUNA -5.975** 2001:3,2004:3
2 ABIA -5.247 1996:3 21 KANO -5.604*** 2008:2,2001:2
3 ADAMAWA -5.765** 2008:3 22 KATSINA -5.815** 1998:4
4 AKS 11.745* 1999:4, 2007:4 23 KEBBI -5.097 1999:3
5 ANAMBRA -5.386*** 1995:4 24 KOGI -5.604*** 2000:3
6 BAUCHI -5.79** 1998:4 25 KWARA -5.366*** 2006:3
7 BAYELSA -12.996* 2010:1 26 LAGOS -8.322* 2002:3
8 BENUE -5.705* 2001:2 27 NASSARA -5.338*** 1998:1
9 BORNO -5.912* 2007:3 28 NIGER -6.601* 1998:3,2004:3
10 CRS -5.475*** 1995:4,2001:4 29 OGUN -4.916 1996:3
11 DELTA -7.929* 1999:3,2004:3 30 ONDO -4.363 1999:4
12 EBONYI -5.041 2003:3,2005:2 31 OSUN -6.437** 2007:4
13 EDO -9.079* 2003:2,2006:3 32 OYO -6.134** 2007:4
14 EKITI -6.463* 1999:3 33 PLATEAU -4.407 2002:3
15 ENUGU -6.319* 1995:4,2006:3 34 RIVERS -5.936** 2000:2,2008:1
16 FCT -5.185 2005:3 35 SOKOTO -6.019** 1996:4
17 GOMBE -6.513* 2007:3 36 TARABA -7.361* 1999:3,2004:4
18 IMO -4.506 1996:3 37 YOBE -5.599*** 2009:4
19 JIGAWA -5.121 1996:4 38 ZAMFARA -6.518* 1999:4,2006:4

Notes: * significant at 1%, ** significant at 5%. *** significant at 10%

Critical Values : 1% -6.450 5% -5.670 10% -5.310. The critical values are taken from Lee &
Strazicich(2003), Model C, Table 2.

Results from unit root tests with breaks are presented in Table 4.2. We find that the
national unemployment rate and those of 30 states reject the unit root hypothesis
with breaks and hence are confirmatory of the structuralists thesis. Only the un-
employment rates of eight( 8) states are found to exhibit unit root in the presence
of structural breaks. The 8 states include Abia, Ebonyi,and Imo (South East),
Jigawa and Kebbi(North West), FCT and Platea( North Central), Ogun and
Ondo(South West) states . The behavior of their unemployment rates are
consistent with the hysteresis school. This means that cyclical fluctuations
induced by business cycles have permanent effect on unemployment in these eight
(8 )states. For these states, individuals who have been unemployed for long
periods are likely to remain unemployed, since their skills eventually decline.
25
 However the behavior of unemployment rates of the majority of the states
(29) and national unemployment rate of Nigeria is consistent with the
structuralists thesis. Shocks to unemployment rate have no permanent effect.
They are temporary and tend to converge to the natural rate of unemployment in
the long run. Compared to the results in Table 4.1 the results in Table 4.2 reveal
that the degree of persistence of the unemployment rate decreases when breaks are
allowed.
From the results we can identify four main clusters of regions indicated by
presence of common structural breaks which are evidence of existence of
interrelationship in their respective local labour markets.
Cluster A: 1995-1996 Nigeria, Anambra, Enugu, Cross River (CRS).
Cluster B: 1998-2001 Akwa Ibom, Bauchi, Benue, Cross River, Delta,
Ekiti, Kaduna, Kano, Katsina, Kebbi, Kogi, Ondo, Nassarawa, Niger, Rivers,
Taraba, Zamfara.
Cluster C: 2003-2005: Edo, Ebonyi, Enugu, Delta, FCT, Kaduna, Kwara,
Lagos, Niger, Plateau, Taraba, Zamfara.
Cluster D: 2008-2010: Adamawa, Akwa Ibom, Bayelsa, Borno, Gombe,
Kano, Osun, Oyo, Rivers, Yobe.
These break points are associated with four important events, the movement
from regulation to deregulation in 1995, the transition from military dictatorship
to civilian democratic government in 1999, the adoption of national economic
emancipation and development strategy in 2003 and recapitalization and
consolidation in banking of 2004/2005 which led to the collapse of many banks
and attendant retrenchment of staff in the sector, and finally the global financial
crisis of 2007/2008 which impacted negatively on the Nigerian economy.

Table 4.3: Convergence Results with Fractional Integration

26
SN State Intercept Trend FI(d) AR1 AR2 MA1 MA2
1 ABIA -0.059 0.508 1.144 -0.516 0.109 -1.116
(-1.688) (3.062) (7.017) (-4.068) (7.93) (-83.52)
ARFIMA (2,d,2) LogLKL -79.53 SBIC -95.12
2 ADAMAWA 9.592 -0.208 -0.44 1.718 -0.894 -0.061
(42.75) (-34.17) (-2.083) (18.73) (-11.33) (-0.294)
ARFIMA (2,d,1) LogLKL -99.43 SBIC -115.02
3 AKWA IBOM 5.938 -0.137 -0.394 1.921 -1.005
ARFIMA(2,d,0) (10.73) (-7.77) (-1.681) -27.91 (-13.82)
ARFIMA (2,d,0) LogLKL -85.63 SBIC - 98.996
4 ANAMBRA 5.163 -0.043 1.822 -0.895
8.747 (-2.560) 19.96 (-10.87)
ARFIMA(2,d,0) LogLKL -63.43 SBIC -74.565
5 BAUCHI 5.802 1.05 0.665
(773.63) (4.692) (12.56)
ARFIMA(1,d,0) LogLKL -160.04 SBIC -168.97
6 BAYELSA -2.029 1.106 0.671
(-33.29) (3.070) (11.34)
ARFIMA(1,d,0) LogLKL -323.8 SBIC -332.73
7 BENUE 9.132 1.073 0.686
(23.85) (4.87) (12.97)
ARFIMA(1,d,0) LogLKL -129 SBIC -137.96
8 BORNO 2.98 0.816 0.756
(65.20) (2.558) (7.224)
ARFIMA(1,d,0) LogLKL -87.415 SBIC -99.55
9 CROSS RIVER 2.988 1.097 0.725
(260.29) (9.747) (15.498)
ARFIMA(1,d,0) LogLKL -60.99 SBIC -69.92
10 DELTA 2.95 -0.058 1.026 0.66
(20.3) (-0.19) (5.529) (21.84)
ARFIMA(1,d,0) LogLKL -106.72 SBIC -117.86
11 EBONYI 7.8 1.058 0.699
(2108.2) (5.644) (15.312)
ARFIMA(1,d,0) LogLKL -87.71 SBIC -96.04
12 EDO 3.026 1.908
151.74 28.44
ARFIMA(0,d,0) LogLKL -88.39 SBIC -95.1
13 EKITI 1.017 1.75
35.96 7.689
ARFIMA(0,d,0) LogLKL -112.64 SBIC -119.36
14 ENUGU 7.863 1.109 0.737

27
 866.88 9.748 16.968
ARFIMA(1,d,0) LogLKL -90.73 SBIC -99.65
15 FCT 2.872 1.834
(27.15) (13.025)
ARFIMA(0,d,0) LogLKL -106.23 SBIC -112.95
16 GOMBE 5.801 1.057 0.665
(775.9) (6.766) (17.054)
ARFIMA(1,d,0) LogLKL -146.97 SBIC -155.9
17 IMO 0.027 1.759
(0.235) (7.049)
ARFIMA(0,d,0) LogLKL -129.68 SBIC -136.398
18 JIGAWA 1.883 -0.476 0.805 0.744
13.11 -1.052 2.326 13.62
ARFIMA(1,d,0) LogLKL -86.796 SBIC -100.19
19 KADUNA 3.581 -0.118 0.753 0.768 -0.273
2.44 -0.586 1.742 14.71 -0.819
ARFIMA(1,d,1) LogLKL -102.51 SBIC -121.44
20 KANO 1.985 0.958 0.612
(5.851) (2.025) (16.45)
ARFIMA(1,d,0) LogLKL -181.27 SBIC -192.44
21 KATSINA -0.0191 1.051 0.677
(-0.869) (4.629) (14.063)
ARFIMA(1,d,0) LogLKL -154.97 SBIC -163.9
22 KEBBI 4.941 1.064 0.677
(30.674) (3.431) (15.77)
ARFIMA(1,d,0) LogLKL -141.77 SBIC -152.94
23 KOGI -0.018 1.125 0.679
(-1.026) (7.896) (16.08)
ARFIMA(1,d,0) LogLKL -94.3 SBIC -103.23
24 KWARA -0.026 1.05 0.659
(-1.235) (3.467) (18.3)
ARFIMA(1,d,0) LogLKL -175.65 SBIC -184.58
25 LAGOS 2.979 1.077 0.67
(145.76) (5.471) (22.17)
ARFIMA(1,d,0) LogLKL -126.76 SBIC -135.69
26 NASSARAWA 0.999 1.146 0.806
(199.85) (13.35) (5.91)
ARFIMA(1,d,0) LogLKL -76.93 SBIC -85.86
27 NIGER 5.997 1.152 0.738
(663.39) (17.44) (6.587)
ARFIMA(1,d,0) LogLKL -119.76 SBIC -128.69

28
 28 OGUN 6.047 1.148 0.646
(92.03) (6.57) (13.23)
ARFIMA(1,d,0) LogLKL -85.21 SBIC -94.14
29 ONDO 2.038 1.11 0.716
(37.56) (8.01) (7.469)
ARFIMA(1,d,0) LogLKL -67.877 SBIC -76.81
30 OSUN 6.01 1.213 0.714
(307.33) (23.45) (9.29)
ARFIMA(1,d,0) LogLKL -88.29 SBIC -97.22
31 OYO 6.221 0.774 0.7226 -0.283
(12.033) (1.868) (24.421) (-0.892)
ARFIMA(1,d,1) LogLKL -76.43 SBIC -87.6
32 PLATEAU 2.094 0.9097 0.7599 -0.182
(9.263) (1.938) (18.8) (-0.537)
ARFIMA(1,d,1) LogLKL -92.289 SBIC -103.45
33 RIVERS -1.801 1.008 0.629
-78.22 3.59 12.52
ARFIMA(1,d,0) LogLKL -145.11 SBIC - 154.05
34 SOKOTO 6.025 1.143 0.735
(178.15) (10.77) (12.52)
ARFIMA(1,d,0) LogLKL -109.79 SBIC 118.72
35 TARABA 9.071 1.0696 0.6697
(89.99) (4.034) (19.61)
ARFIMA(1,d,0) LogLKL -180.04 SBIC -188.98
36 YOBE 4.044 1.114 0.698
(58.34) (5.46) (13.23)
ARFIMA(1,d,0) LogLKL -77.04 SBIC -85.97
37 ZAMFARA 4.995 1.12 0.718
(320.59) (8.017) (10.26)
ARFIMA(1,d,0) LogLKL -179.18 SBIC -188.11
Notes: values in brackets are t-values; LKL=loglikelihood ratio; SBIC=Schwarz bayesian information
criterion

Table 4.3 presents the results for convergence test. First we find that two (2)
states (Adamawa and Akwa Ibom) display rapid catching-up or short memory
catching-up given that the trend coefficient is negative and statistically significant
and the fractional integration parameter is less than zero. One state, Anambra
exhibits strict conditional convergence. But seven states (7) namely Abia, Borno,
Kaduna, Kano, Jigawa, Plateau and Oyo show long memory but mean-reverting
conditional convergence with the national unemployment rate. The implication of

29
conditional convergence is that each state approaches its own steady-state
unemployment rate. Keeping the differentials with national rate constant. Twenty
seven (27) states show absolute divergence from the national unemployment
rate. These include: Bauchi, Yobe and Taraba (North-East), Sokoto and Zamfara
(North-West), Bayelsa, Cross River, Delta, Edo, Rivers (South-South), Ebonyi
(South-East), Ekiti, Lagos, Ogun, Ondo and Osun (South-West), Benue, Kogi,
Kwara and Niger(North-Central). These states cut across the six geo-political
zones in the country. Overall the results show that the unemployment rate
among most of the Nigerian states do not show any tendency of convergence
towards the national average. One of the reasons for this is the uniform level of
minimum wage which is relatively high compared to labour productivity in
economically lagging poor states. Policies aimed at reducing relative
unemployment differences are not likely to make much impact in Nigeria unless
they are sufficiently large.

VI Concluding Remarks

Given the spate of bombings, kidnapping, armed robbery and terrorism in

Nigeria there has been renewed efforts to tackle the high unemployment rate in the
country. This is more so because vast differences in unemployment rates among
the states are indicative of regional economic inequality and poverty. This paper
investigates the dynamics of unemployment rates among the states in Nigeria and
the prospects for convergence with the national rate.

We use more theoretically defensible unit root testing with structural breaks
and the autoregressive fractional integrated moving average (ARFIMA) approach.
Our results show that behavior of unemployment rates of thirty(30) states are
consistent with the structuralists/NAIRU thesis given that they are stationary with
breaks. We identify four clusters of break points 1995/96, 1998-2001, 2003-2005
and 2008-2010 which correspond respectively with confusing policy signals of
deregulation, regulation and deregulation between 1993 and 1995 and the attendant
distress in the financial sector, the change from military dictatorship regime to
civilian democratic governance in 1999, the failure of NEEDS and the

30
implementation of banking policy of recapitalization and consolidation in
2004/2005, the global financial crisis in 2008 which affected Nigeria negatively.

We also investigate the prospects for convergence in relative unemployment

rate among the states. We find some evidence of a process of conditional
convergence among ten states with the national unemployment rate. The states
include Anambra, Adamawa, Akwa Ibom, Abia, Borno and Kaduna. Others
include Kano , Jigawa, Plateau and Oyo states. Adamawa and Akwa Ibom states
display rapid catching-up or short memory catching-up. However we find some
evidence of non-convergence among 27 states. For these states the fractional
integration parameter d exceeded unity. The divergence pattern observed in
unemployment differentials among most of the states is a consequence of
hysteresis in the unemployment rates of some states. The fact that ten states display
stochastic convergence suggests that at least part of Nigeria is converging. This
means that poorer states are on the average catching-up to the national
unemployment rate.

The non convergence of relative unemployment differentials of most of the

states is indicative of the long lasting problem of unemployment which may not
go away. Policies aimed at reducing relative unemployment differences are likely
to make much impact when they are substantial. These include substantial
investment in infrastructural rehabilitation in electricity, road and rail network,
development of entrepreneurial skills among the youth and government provision
of start-up grant or loans to build private businesses.

References
Abiodun, S.O. (2010):Analysis of mismatch between supply of skills and university graduate
unemployment in Nigeria. Unpublished M.Ed Dissertation, Lagos State University, Nigeria

31
African Economic Outlook (AEO)(2012): Nigeria.
www.africaneconomicoutlook.org/en/countries/west-africa/nigeria/

Aghion, P. and O.J. Blanchard(1994):On the speed of transition on Central Europe. NBER
Macroeconomic Annual, 283-320.

Arestis, P. and Mariscal, I. (1999);Unit roots and structural breaks in OECD unemployment.
Economic letters 65, 149-156.

Armstrong, H. and J. Taylor (2000): Regiona Economics and Policy. 3rd edition, Blackwell.

Bankole, O. (2002): Graduate unemployment in Nigeria: A report by National Institute for Social
and Economic Research. NISER

Bayer, Christian and Falko Juessen (2006):Convergence in Wst German Regional

Unemployment Rates. Mimeo University of Dortmund. Available online:
econwpa.wustl.edu.8089/eps/urb/papers/0411007.pdf

Bernard, A. and S. Durlauf (1995): Convergence in international output. Journal of Applied

Econometrics 10, 97-108.

Bernard, A. and S. Durlauf (1996): Interpreting tests of the convergence hypothesis. Journal of
Econometrics 71, 161-73

Beyaert, A. (2004):Fractional Output Convergence with an Application to Nine Developed

Countries. Econometric Society 2004 Australian Meetings 280, Econometric Society.

Blanchard O.J. and L. Katz(1992): Regional evolution. Brookings Papers on Economic Activity
1, 1-75.

Blanchard, O.J. and L.H. Summers(1986): Hysteresis and the European unemployment problem.
NBER Macroeconomics Annual (ed. S. Fisher) Cambridge, MA, 15-78

Blanchard, O.J. and L.H. Summers(1987):Hysteresis in unemployment. European Economic

Review 31, 288-95.

Boeri, T. (2001):Transition with labour supply. IZA discussion Paper 257

Boeri, T. and K. Terrell(2002): Institutional determinants of labour reallocation in transition.

Journal of Economic Perspectives 16, 51-76

Breitung, J. (1994): some simple tests of the moving average hypothesis. Journal of Time Series
Analysis, 15, 31-59

Brunello, G. (1990):Hysteresis and the ‘Japanese unemployment problem’: a preliminary

investigation. Oxford Economic Papers 42, 483-500.

32
Camarero, M. and C. Tamarit(2004): Hysteresis vs. natural rate of unemployment: new evidence
for OECD countries. Economic Letters 84, 413-417

Carmeci, Gaetano and Laura Chies (2006): Hysteresis in unemployment: Do structural breaks
and aggregation matter? Working paper 104, Universita degli studi di Trieste. Departimento di
Scienze Economiche e Statistiche.

Carlino, g. and I. Mills(1993): Are US regional Income converging? Journal of Monetary

Economics 32, 335-346

Central Bank of Nigeria (CBN)(2002): Statistical Bulletin. Abuja

Central Bank of Nigeria (CBN)(2007): Statistical Bulletin. Abuja

Cheng,Ka Ming; Nazif Durmaz, Hyeongwoo Kim, and Michaerl L. Stern (2011): Hysteresis vs.
Natural Rate of US Unemployment. Economic Modelling 29(2) 428-434. See also Auburn
Economics Working Paper Series AUWP2011-01, Department of Economics, Auburn
University.

Corseuil, C.H., G. Gonzaaga, and J.V. Issler(1999):Desemprego Regional no Brasil: uma

abordagem empirica. Economia Aplicada v.3- n. 3 Julho-Septembro

Cunado, J., Gil-Alana, L.A. and F. Perez de Gracia(2004a):Real Convergence in some emerging
countries: A long memory approach. Working paper, Department of Economics, Universidad da
Navarra, Spain.

De Figueiredo, Erick Alencar(2010): Dynamics of regional unemployment rates in Brazil:

Fractional behavior, structural breaks, and Markov switching. Economic Modelling 27(50 900-
9008. Also available at http://www.sciencedirect.com/science/article/B6VB1-50DXD5M-

Dickey, D. and W. Fuller (1979):Distribution of the estimates of autoregressive time series with
unit root. Journal of the American Statistical Association 74, 422-431

Dufrenot, Gilles; Valerie Mignon and Theo Naccache(2009):The Slow Convergence of per
capita income between the developing countries: growth resistance and sometimes growth
tragedy. CREDIT research paper No. 09/03. Centre for Research in Economic Development and
International Trade, University of Nottingham.

Fallahi, Firou and Gabriel Rodriguez(2010); Persistence of unemployment in the Canadian

provinces. Documento de Economia No. 286, July. Retrieve at
http://www.pucp.edu.pe/departamento/ economia/images/documentos/DDD286.pdf

Ferragina, Anna Maria and Francesco Pastore(2005): Mind the gap: unemployment in new EU
regions. IZA Discussion papers No. 1565

33
Friedman, Milton. (1968); The role of monetary policy. American Economic Review 58, 1-17

Gomes, Fabio Augusto Reis and Cleomar Gomes da Silva(2006):Hysteresis vs Nairu &
convergence vs divergence: The Behaviour of reginal unemployment rates in Brazil.
www.cepe.ecn.br/seminarioii/material/paper/cleomar.pdf

Granger, C. J. W. (1980). Long Memory relationship and the aggregation of dynamic models.
Journal of Econometrics, 14, 227-238. http://dx.doi.org/10.1016/0304-4076(80)90092-5

Granger, C. J. W., & Joyeux, R. (1980). An introduction to long memory time series and
fractional differencing. Journal of Time Series Analysis, 1, 15-29.
http://dx.doi.org/10.1111/j.1467-9892.1980.tb00297.x

Gregory , R.G. (1986): Wages Policy and Unemployment in Australia, Economica, 53: S53-S74.

Hatanaka, M. (1996): Time –Series-Based Econometrics, Oxford, Oxford University Press.

International Labour Organization(ILO)(2010): Labour force statistics

Hosking, J. R. M. (1981): Fractional differencing. Biometrika, 68, 165-176.

Katrencik, David; Joana Tyrowicz, Piotr Wojcik(2008): unemployment convergence in

transition. RepEc, pra: mprapa

Krugman, Paul Robin (1996):POP Internationalism. Cambridge, Mass. MIT Press

Kwaitkowski, D., P.C.B. Phillips; P.Schimdt and Y. Shin(1992):Testing the null hypothesis of
stationarity against the alternative of a unit root. Journal of Econometrics 54, 159-178.

Layard, R., S. Nickell and R. Jackman(1991):Unemployment: macroeconomic performance and

the labour market. Oxford University Press.

Lee, J. and M.C. Strazicich(2003):Minimum Lagrange multiple unit root test with two structural
breaks. Review of Economics and Statistics 85(4), 1082-1089.

Lee, J., M.C. Strazicich, and M. Tislau(2001);Hysteresis in unemployment? Evidence from

panel unit root tests with structural change. www.econ.unt.edu/research/pdf/01082-1089

Lumsdaine,R. and D. Papell (1997);Multiple trend breaks and the unit root hypothesis. Review
of Economics and Statistics 79, 212-218.

Leon-Ledesma, Miguel A. (2002):Unemployment hysteresis in the US states and the EU.: a

panel approach, Bulletin of Economic Research 54, 95-103.

Leon-Ledesma, Miguel A. and Peter McAdam(2004): Unemployment , Hysteresis and

Transition. Scottish Journal of Political Economy. 51(3) 377-401

34
Marston, S. (1985); Two views of geographic distribution of unemployment. Quarterly Journal
of Economics, 100(1) 57-79

Martin, R. (1997):Regional unemployment disparities and dynamics. Regional Studies 31, 237-
252

Mitchell, W. F. (1993): Testing for unit roots and persistence in OECD unemployment. Applied
Economics 25, 489-501

Mohammed, Abubakar (2012): NDE’s failure to reduce unemployment in Nigeria. Daily

Independent, August 2012.

Muhammad, S.A., N.D. Oye and I. Inuwa(2011):Unemployment in Nigeria: Implications on the

Gross Domestic Product (GDP) over the years. International Journal of Economic Resources,
2(1) 66-71

National Bureau of Statistics (NBS)(2010): National manpower stock and employment

generation survey. Abuja.

National Bureau of Statistics(NBS)(2011): Socio-economic report

National Bureau of Statistics(NBS)(2012): Labour force survey.

National Planning Commission(NPC)(2006):Economic performance review (April/July).

Nelson, C.R. and C.I. Plosser(1982);trends and random walks in macro-economic time series:
some evidence and implications, Journal of Monetary Economics 10, 139-162.

Notini, Hilton Hostalacio and Luiz Renato de Oliveira Lima(2006):Empirical evidence on real
convergence across Brazilian States. Graduate School of Economics, Getulio Vargas Foundation,
Rio de Janeiro, Brazil

Obadan, M. and A. Odusola(2001):The Economy and Poverty in Nigeria: A paper presented at

national conference on Law and Poverty NIALS, Kaduna, June 26-28.

Oliveira, C.W.A. and F.G. A Carneiro(2001): The behavior of Brazilian unemployment rate.
Revisita Brasileira de Economia 55(4) 493-516

Perron, P. (1988): Trends and random walks in macroeconomic time series: further evidence
from a new approach. Journal of Economic Dynamics and Control. 12, 297-332.

Perron, P. (1997):Further evidence on breaking trend function in macroeconomic variables.

Journal of Political Economy 76, 678-711

Phelps, E. S.(1967): Phillips curves, expectations of inflation and optimal unemployment over
time. Economica 34, 254-81.

35
Phelps, E. S.(1994): Structural slumps; the modern equilibrium theory of unemployment,
interest, and assets. Cambridge, MA. Harvard University Press.

Phelps, E.S. and G. Zoega(1998):Natural rate theory and OECD unemployment. Economic
Journal, 108, 782-801.

Reuter, M.(2000): Hysteresis in West Germany Unemployment reconsidered. CESifo Working

Paper 240.

Roed, K. (1996): Unemployment hysteresis – macro evidence from 16 OECD countries.

Empirical Economics 21, 589-600

Roed, K. (1997):Hysteresis in unemployment. Journal of economic Surveys, 11, 389-418.

Rowthorn, R. and A. Glyn(2003): Convergence and stability in US regional employment.

Mimeo, Cambridge University.

Samuel, Akinyemi; Igot Bassey Ofem; S.O. Ikuenomore(2012): Graduate turnout and graduate
unemployment in Nigeria. International Journal of Humanities and Social Sciences 2(14) 257-
265

World Bank (2005): World Bank Development Indicators.

Wu, Z. (2003):The persistence of regional unemployment: evidence from China

Xiao, Z. and P.C.B. Phillips(1997): An ADF coefficient test for a unit root in ARMA models of
unknown order with empirical applications to the US economy. Manuscript, Cowles Foundation.

Zivot, E. and D.W.K. Andrews (1992):Further evidence of the great crash, the oil price shock
and the unit root hypothesis. Journal of Business and Economic Statistics 10, 251-270

36
Figure 3.3: Map of the 36 States and Abuja in Nigeria

37