INTRODUCTION
Due to particular ecological condition and its location in quake-prone
zone, Iran has experienced natural disasters, intermittently. Nine destructive
earthquakes have been occurred in Iran during 1931-2004. Occurrence of
intermittent droughts during 1992-2006 and overflowing of Caspian Sea
during 1992-1997 have also brought significant damages to the country.
These natural disasters can go even more than the above-mentioned occurrences.
Natural disasters cause three main types of impacts: Direct, indirect
and macroeconomic impacts (McKenzie et al., 2005).
Natural disasters, in the intermittent occurrence, have direct impacts on physical
and human capital stock. These incur direct damages including total and partial
destruction of residential areas, buildings and installations, machinery, production
equipment, warehouses, farm lands and crops, dams, irrigation systems. Calculation
of direct damages arising from natural disasters is difficult especially in
the countries that have no proper recording system of assets as well as damages
on environment (including erosion and sedimentation) and human capital. In the
subsequent period of natural hazards, indirect losses include decline in the
production of goods and services due to destruction of equipments, decline in
agricultural outputs due to flooding of arable lands or decline in grain production,
decline in industrial products due to damages incurred on factories or lack
of institutions and increase in transportation expenditure due to destruction
of roads or other transportation infrastructure. The third effect is the changes
occurred in the macroeconomic variables that are created through the direct
and indirect impacts arising from natural disasters. The natural disasters affect
mostly the gross investment, balance of payments, public finance, inflation
and unemployment of the country. All these affects can be reflected in the real
GDP, the variable that shows the changes of general level of the economic activity
of the country. However, in the case there would have been increasing demand
for the renovation activities, a natural crisis can have positive impacts on
GDP growth.
A natural disaster abruptly decreases the quantity of the physical capital
of the economy that in turn leads to decline in production. As a result,
natural disasters have negative effects on the economic growth in general
and may decline the agricultural production or ruin the industrial production
capacity in different cases.
The ways of financing the needed budget for
the reconstruction during post-disaster period have particular effect
on the GDP growth. Table 1 shows the effects of natural
disasters on macroeconomic variables.
Major natural disasters have both intense short-term negative impacts and long-term
negative consequences on the economic growth and development as well as poverty.
The fundamental question relating to natural disasters is that whether the natural
disasters are essentially negative economic shocks or they can also have positive
economic effects (due to affluence in construction in post-disaster period and
recovery of infrastructure and technology). Answer to this question depends
on sequencing of the effects, kinds of the natural calamities (earthquake, hurricane,
flood, drought, etc.), degree of vulnerability against natural hazards and other
effects related to economic performance (Benson and Clay, 2004).
From economic point of view, a natural disaster is a shock that leads to decline
in the human, social and physical capital stocks as well as fall in economic
activities in the form of decline in production, investment and employment.
Distributive consequences of these disasters are also important because the
poor probably incur maximum damages from these calamities. Necessarily, segregation
should be maintain between direct impacts like physical damages and decline
in the level of production (like grain production) and indirect impacts like
changes in the flows of economic activities, but direct and indirect impacts
are often put together for calculating total cost of a natural disaster. For
instance, total damage of Tsunami in 2004 was US $ 4.45 billion, which included
US $ 2.9 billion direct effects and US $ 1.53 billion as indirect effects (Benson
and Clay, 2004).
As a whole, relation between level of development of one country (according
to its per-capita income) and influence of a natural disaster is complex and
undefined. This complexity indicates the fact that the development itself is
a non-linear process and there exists various routes for the development (Benson
and Clay, 2004). Meanwhile, under-developed countries have the most vulnerability
against natural calamities because in these countries, building codes and land
use by private sector are executed in weak manner and public infrastructures
are also not built according to the standards laid down to resist against disasters.
The only study has been conducted in Iran about the effects of natural disasters
on the economy is the work of the United Nations in cooperation with Iran. After
the earthquake of north of Iran in 1991, which has been one of the most destructive
earthquakes of the world, a team from the United Nations along with Housing
Foundation of Iran took step to assess the damages and evaluate reconstruction
needs of the affected area. Apart from US $ 50 million of damages incurred on
industrial foundations, the study evaluated the total economic damages from
this earthquake about US $ 1.2 billion, 76% of which was related to housing,
5.8% related to means of living of the affected households and the rest was
related to production sectors, infrastructure and other economic activities
of the region (UNDP et al., 1996).
Regarding the literature on the macroeconomic effects of the natural disasters,
we can refer to the following studies. Toya and Skidmore
(2007) studied the damages of natural disasters. The findings of the research
showed that countries with high per-capita incomes, education, degree of opening,
financial systems, smaller administration, experience lower intensity of damages
from those calamities. Noy (2006) also studied the effect
of natural disasters on macro economy during short term and found that the countries
with high literacy rate, higher per-capita income, jumbo government and high
trade freedom have better resistance power compare to initial shock. On the
other side, the countries that have dependent capital accounts, high level of
foreign reserves and higher level of internal credit, are powerful against the
shocks and resist against the penetration of effects to the GDP growth rate.
Anbarci et al. (2005) linked the earthquake fatalities
to the per-capita income and the level of domestic inequality (based on Land-Based
Gini Coefficient) as well as evaluated and predicted other factors affecting
on destructive power of earthquakes such as their magnitude, depth and proximity
to the population centers. Based on a theoretical model and keeping suitable
control variables constant (like magnitude, population, land area, distance
from epicenter, frequency of big earthquakes and other regional factors), they
predicted that number of fatalities must be a decreasing function of the levels
of per-capita income of a country and equality and confirmed the predictions
using negative binominal estimation method with coincidental and constant estimators.
Kirigia et al. (2004) in accounting economic effects
and human damages arising from natural disasters in 46 African nations, who
are the members of World Health Organization (WTO), showed that death of a single
person under such disaster has decreased the GDP to the rate of 0.01828 dollars.
Benson (2003) in his study about the effects of natural
disasters on long-term growth found that countries, which repeatedly experienced
the natural disasters during 1960-1993, accumulated lower rate of growth compare
to the countries experiencing less natural disasters. Narayan
(2003) about the damages resulting from hurricane to substructures, industrial
and agricultural activities in Fiji indicated that import and export declined
(exports with high percent) with the effect of 2003 hurricane and caused deficit
balance of payment. Likewise, actual GDP, private consumption, income, investment,
saving, national welfare declined. Skidmore and Toya (2002)
studied the long-term effects of natural disasters on economic growth. They
acquired abundance of natural disasters for each country during 1960-1990 and
have normalized them with respect to the extent of each country. They have determined
the correlation of this standard with economic growth, human and physical capital
stocks and exploitation of total factors. Used damages of natural disasters
for estimating the effects on the rate of family savings and as such found that
with assuming other factors determining saving, enhancement of damages arising
from geological and environmental atrocities distinctively with the rate of
savings of the families have correlation. Selcuk and Yeldan
(2001) in their assessment of the 1999 earthquake in Turkey on distinct
macroeconomic variables in short and long term applying General Equilibrium
(GE) model showed that the effects of the primary earthquake on GDP, taking
into account the policies followed by the government and international assistance,
have fluctuation from 4.5% negative to 0.8% positive GDP. Benson
and Clay (2004) and Cochrane (1994) studied the effects
of natural disasters on the countrys debt. Using a growth model of Kenzy and
identifying negative shocks in the form of lessening public and private capitals
and augmenting government expenditure for emergency helps, he reached to the
conclusion that natural disasters can reduce the confidence level of a country,
enlarge the debt rate of foreign loans and increase the debt stockpile with
declining investment and long-term growth. Albala-Bertrand
(1993) in his experiment about the relations between natural crisis and
its potential effects on the rate of growth output, reached to the conclusion
that the emergence of natural disasters cause minimum GDP growth in short-term,
no change in inflation, increase of investment and trade deficit as well as
growth in budgeting.
MATERIALS AND METHODS
To explain the effects of natural disasters on GDP in Iran during 1978-2004,
we first assume that desirable capital stock is related to level of production
in the following way:
where, desirable capital stock Kt is desirable capital stock
in the time (t), Yt is production in time (t) and Ut
is residual term. From the view of natural disasters, shocks incurred
on economy can be written as:
where, DMSt is the damage arising from natural disaster in
time (t) and thus we can write:
Thus, based on the adjustment hypothesis of existing capital, changes
of capital stock take place according to the following model:
where, δ is adjustment coefficient and Kt-Kt-1
is equal to gross investment which is shown with INV:
by solving the above equation in terms of Yt, the final equation
will be as:
in which the effect of the other variables is summarized in the intercept
(α0) and ηt shows the residual term with
identically independent distribution (i.i.d).
Equation 6 based on the size of damage of natural disasters
is on the situation that the developing countries like Iran at the time
of occurrence of disaster are confronted with basic deficiencies and shortages
(especially in the affected zone). As a result, level of economic activities
declines at least for short-term but in the later stage (during renovation)
with the allocation of enough financial resources, economic prosperity
is created and GDP is increased. Results of this process show that gap
between expected level of growth and actual level of growth become wider
after natural calamity.
Therefore, effect of natural calamities on national economy initially
is diminished and thereafter it is increasing, which can be observed in
Fig. 1, in which natural disasters have occurred in
the periods t1 and t2:
Inclusion of gross fixed capital formation (gross investment) is due
to positive compensatory effect of this variable after the emergence of
the natural disaster on GDP.
Rewriting Eq. 6 in per capita form gives rise to the
final model:
where, PGDP, PINV and PDMG indicate per capita GDP, per capita investment
and per capita natural damages, respectively.
According to Fig. 1 positive and encouraging effect
of investment growth on GDP, it is expected the following restrictions
on the coefficients of model (7) :
The statistics of GDP and gross fixed capital formation variables during
the study period (1978-2004) have been extracted from WDI, 2006 and as
such data about natural disasters damages have been extracted from EM-DAT
(www.em-dat.net) database
affiliated to the University of Louvain, Belgium. Per capita GDP, PGDP,
(in constant prices of 2000) is defined as Gross Domestic Product divided
by labor force. Per capita damages of natural disasters, PDMG, (in constant
price level of 2000) is calculated as damages divided by labor force.
|
Fig.1: |
Impacts of natural disasters on economic growth |
Table 2: |
Results of unit root test of actual research variables |
|
McKinnon critical value of tests at the level 5%: -3.69 |
Per capita gross fixed capital formation or per capita investment, PINV,
(in constant price of 2000) is measured as gross investment divided by
labor force.
In order to prevent the spurious regression, it is necessary to conduct
the required unit root tests. Table 2 explains the results
of unit root test:
According to Table 2, PGDP and PDMG are integrated
of degree one, i.e., I (1), but PINV is integrated of degree zero, i.e.,
I (0). Therefore analysis through ARDL method, in which the short-term
dynamism is considered in the model, leads to more exact coefficients
compare to other methods. The dynamic model is as follows:
To decrease the estimation bias of regression coefficients in small samples,
it is advised to use sufficient lags. Then the above Auto-Regressive Distributed
Lags (ARDL) model is used:
in which:
where, L is lag operator, w is a vector of constants including intercept,
dummy variables, time trend or exogenous variables having constant lags
and p and q are maximum number of lags for Y and X variables. Equation
3 is estimated for all possible orders of p and q in (m+1)(k+1)
stages using Microfit software in which m is the maximum lag determined
by researcher and k is the number of explanatory variables.
At the later phase, one equation is selected by using Akaike Information
Criterion (AIC), Schwarz-Baysian Criterion (SBC), Hannan-Quinn Criterion
(HQC) or adjusted coefficient (
).
Usually in the samples less than 100, Schwarz-Baysian Criterion is used
to conserve maximum degree of freedom.
For calculating long-term coefficient model, the same dynamic model is
used. Long-term coefficients related to x variables are given by the following
equation:
Then, to overcome the spurious regression of the long-term relationship,
the following hypothesis is tested:
The null hypothesis (H0) indicates the lack of co-integration
or long-term relationship among model variables. Since the condition that
short-term dynamic relation approaches to long-term equilibrium is that
sum of coefficients of lagged dependent variables (Φi)
will be less than one. To test the co-integration relation, the sum of
Φis minus 1 is divided by sum of standard deviation of
them (Σ Φi) :
If absolute value of t in (13) is greater than the absolute value of critical
value presented by Banerjee et al. (1998), then
the null hypothesis will be rejected and existence of long-term relations will
be accepted. In the substituting process, the method proposed by Pesaran
and Yongcheol (1996) and statistics F calculated by them would be applied
to study the long term relations between variables. Relying on the co-integration
among a set of economic variables, Error Correction Model (ECM) can be used.
In this model, fluctuations of short-term variables are related to long-term
equilibria. This model is a special from of Partial Adjustment Models (PAMs),
in which with the entrance of stationary residual terms resulting from long-term
relationship, speed of closeness of short term quantities of variables to long-term
equilibrium is measured.
Error Correction Model (ECM) is estimated in two stages: first, long
term relation is estimated to be sure that it is not spurious. Second,
when residual lag of long-term relation is used as error correction coefficient,
the following equation is regressed:
Error correction coefficient (c) with negative sign (which is also expected)
would indicate error correction speed and desire to long term equilibrium.
This coefficient shows that in each period, how many percentage of disequilibrium
of dependent variable were adjusted and become closer towards long-term
relation.
RESULTS AND DISCUSSION
As indicated, with reference to the differences in degrees of integration,
it is necessary to use auto-regressive distributed lags (ARDL) model for
estimating the experimental model (7) because in most of the economic
studies, macroeconomic variables represent their impacts with time lags.
Results of dynamic equation estimation (where dependent variable with
one lag appears in the line of explanatory variables) are shown in Table
3 in which ARDL (1, 0, 0) is applied and optimum lag is 1 based on
SBC criterion.
According to Table 3, in the short term PGDP with one
lag affects the level of same variable with coefficient +0.31. Likewise,
per capita gross investment has positive impact on per capita GDP and
PGDP that is compatible with economic theory. On the other side, per capita
damages of natural disasters, PDMG, in constant prices, negatively affect
the level of economic activities and thus, the relationship between PGDP
and PDMG is negative. Except for PDMG variable that is significant at
10% level, rest of the variables are significant at the level of 5% and
statistic F indicates whole significance of the regression. Before the estimation of long-term coefficients and Error Correction
Model (ECM), diagnostic tests are necessary for the accuracy of the model
(including tests on lack of serial correlation, functional form specification,
normality of error terms and heteroscedasticity). In Table
4 and according to F and LM statistics at 5% significant level, the
estimated model do not encounter the problem of auto-correlation of error
terms and heteroscedasticity, the specified form is correct and error
terms (residuals) are normally distributed.
Table 3: |
Autoregressive Distributed Lag Estimates: ARDL (1, 0,
0); Dependent variable: Per capita Gross Domestic Production = PGDP) |
|
Note: R2 refers to identification coefficient,
to
adjusted identification coefficient, F to Fisher statistics, AIC to
Akaike information criterion, SBC to Schwarz Bayesian criterion and
SD to standard error |
Table 4: |
Results of diagnostic tests for PGDP |
|
Note: LM refers to lagrange multiplier statistic and
F to Fisher statistic |
Considering the differences in the integration degrees of the model variables,
the condition of tendency of the dynamic model towards long-term equilibrium
and the existence of co-integration relationship among variables is that
sum of coefficients of lagged dependent variable should be less than one.
Here, the t-test based on relation (13) is:
Since absolute value of t is greater than the critical absolute value
in Banergee, Dolado and Mestres Table (with 50 observations, 2 explanatory
variables and inclusion of intercept, except for lagged dependent variable)
is -4.43. So, the existence of long-term equilibrium relations among variables
is confirmed. Table 5 shows the estimation results of
long-term equilibrium model.
Although the per capita damage of natural disasters (PDMG) affects, at
10% significant level, the per capita GDP, PGDP, but as a whole, presence
of long-term relation between these variables is confirmed. In other words,
if all conditions are constant, a long-run equation is calculated as:
PGDP = 3507.3+0.928*PINV-76074.4*PDMG |
(15) |
For estimating the error correction model, error terms of co-integration
regression in long-term equation form with one lag is considered as an
explanatory variable along with first difference of other variables. Table
6 shows the estimated results of such model.
According to the Table 6, the coefficient of lagged
error correction term is equal to -0.685 which is significant statistically
at the 5% confidence level. This coefficient indicates that in each period
only 17% of disequilibrium is adjusted; therefore adjustment speed is
relatively slow.
Table 5: |
Estimated long run coefficients of per capita gdp:ardl(1,
0, 0) |
|
Table 6: |
Error correction model for per capita GDP: ARDL (1,0,0) |
|
Note: R2 refers to identification coefficient,
to
adjusted identification coefficient, F to Fisher statistics, AIC to
Akaike information criterion, SBC to Schwarz Bayesian criterion and
SD to standard error |
CONCLUSION
With the help of a method proposed by Pesaran and Yongcheol.
(1996) for analyzing co-integration as well as auto-regression distributed
lag model (ARDL) and theoretical and experimental studies about the effects
of natural disasters on different economic variables, the present article studied
the damages incurred by natural disaster on the non-oil GDP as a negative shock
on the whole level of economic activities during 1978-2004. Explanatory variables
in this study included damages from natural calamities and fixed gross capital
formation as strong and powerful incentive in the movement of the level of total
economic activities.
Estimation results of ARDL (1, 0, 0) confirm the existence of long-term
equilibrium between non-oil GDP and explanatory variables. On the other
side, results of short-term dynamic estimation and long-term model indicated
the positive effects from the investment and negative effects of damages
incurred by natural disasters on the non-oil gross domestic product are
accepted. In other words, with the emergence of natural disasters, initially
GDP growth declines and then in the stages of renovation and reconstruction,
it begins to increase. Negative impact of natural disasters on per capita
GDP in the first stage is due to decline in physical capital but during
renovation, the government enters into the operation and allocates the
budget to remedy the damages incurred upon the capital (infrastructures,
residential and industrial units deserted or the products lost). Likewise,
estimated results of error correction model indicate the relatively slow
speed of adjustment in the disequilibria in a way that in each period
only 17% of disequilibria is adjusted.