The effect of climate fluctuation on output in developing countries

This paper has analysis the effect of climate change on the aggregate output particularly with two typical elements: temperature and precipitation in developing countries.. Besides, cont

VIETNAM- NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE EFFECT OF CLIMATE FLUCTUATION ON OUTPUT IN DEVELOPING COUNTRIES

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGO KHANH LIEN HUONG

Academic supervisor

Dr PHAM KHANH NAM

Ho Chi Minh City, July 2012

ACKNOWLEDGEMENTS

This paper has could not be started and completed without the help of several individuals who supported me directly and indirectly First of all, I appreciate my supervisor Dr Pham Khanh Nam so much for his enthusiastic assistance He has not only given me genuine intellectual guidance in academy but also encouraged me a lot through the analysis process It is so hard for me to complete this research without his profound advice Thus, I am very grateful to him I am also thankful to

Dr Nguyen Trong Hoai for sharing his knowledge and practice experiences in researching which are very useful for this study I also thank all my classmates in class MDE17 for sharing their advantage discussions on econometric techniques, especially Nguyen Thi Thuy Thanh Last but not least, I would like to thank my parents, and all my other family members for their concern and invaluable moral support

ABSTRACT

Climate change is one of the critical problems in recent years on over the world

It impacts on many aspects of human life, that has a lot of researches and evident This paper has analysis the effect of climate change on the aggregate output particularly with two typical elements: temperature and precipitation in developing countries Besides, control variables proxy unobserved effects at country level and country fixed effects are also counted in the relationship of temperature and precipitation with the country's output The finding shows that both temperature and precipitation have significant influence on the GDP of developing countries It

is specially that temperature affects the GDP in U-shaped, while precipitation affects the GDP in hump-shaped Moreover, this paper also conducts the relationship of temperature and precipitation with the output by two characteristics:

"hot country" and "agricultural country" The results are also significant; however, the relationship share cannot be identified Further, this research suggests some adaptive recommendations for production to changing of climate

TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION I

1.1 Problem statement 1

1.2 Research Objective 3

1.3 Chapter summary 4

CHAPTER 2: LITERATURE REVIEW 5

2.1 The theory 5

2.2 Empirical studies 6

2.3 Conceptual framework 14

2.4 Chapter summary 15

CHAPTER 3: METHODOLOGY 16

3.1 Econometric model 16

3.2 Data 20

3.3 Chapter summary 22

CHAPTER 4: RESULT 24

4.1 Descriptive statistics 24

4.2 Non parametric analysis 26

4.3 Econometric analysis 29

4.4 Chapter summary 39

CHAPTER 5: CONCLUSION 41

5.1 Conclusion 41

5.2 Recommendation 42

REFERENCES 44

APPEND IX 46

List of Tables

Table 3.1: Variable descriptions 22

Table 4.1: Descriptions Statistic 25

Table 4.2: Correlation matrix 26

Table 4.3: Temperature and Precipitation - Estimation Result 33

Table 4.4: Estimation result with hot countries dummy 36

Table 4.5: Estimation result with agriculture countries dummy 38

List of Figures Figure 2.1: The way climate fluctuation affects the country's output 15

Figure 4.1: Histogram of temperature -precipitation - GDP per capita 25

Figure 4.2: Average temperature trend 27

Figure 4.3: Average precipitation trend 27

Figure 4.4: Average temperature vs Average GDP per capita 28

Figure 4.5: Average precipitation vs Average GDP per capita 29

CHAPTER 1: INTRODUCTION

1.1 Problem statement

Increasing output is one of the most important targets of most of the countries, especially developing countries The output of a country has been identified in the function of elements such as human capital, physical capital, and technology However, the changing of climate belongs to environment elements in recent years, which is considered that it has a critical effect on country's output Further, the climate change has economically effected on human health, life and production in agricultural countries, which makes the studies more attractive and interesting Human activities have already changed the atmospheric characteristics such as rainfall, global warming, sea level rise, increasing global carbon dioxide (C02) emissions and the thinning of ozone layer Those climate changes happening constantly as a big problem all over the world have impacted upon many aspects of human life such as natural ecosystem, biodiversity, human health, landscape, socio-economic, etc There are obvious evidences that climate change affects human life (IPCC 2007) For example, expanding water stress, increasing temperature but reducing the rainfall has caused a decline in agricultural production in Asian developing countries while demands for production are increasing more and more Climate change has affected the supply, demand, and quality of water resources For instance, Northern Pakistan has been facing water shortages because of expanding winter period over the Himalayas during last 40 years; China is facing up to the drying up of rivers and lake in the dry weather while water need is increasing Likewise, In India, Bangladesh, and Nepal, water shortage is aggravated by climate change; consequently it has leaded to rising urbanization, and using water inefficiently Contrary to the water shortage, glaciers are melting rapidly in Central Asia, which has threatened human settlement, cities, forest ecosystem, and the rising of water level caused the disasters of landslides The 1997/98 El Nino event

in Philippines and Indonesia fired around 2 million ha of peatlands, 9 7 million

hectares of forest (IPCC 2007) That is the extreme climatic event caused substantial fatalities and economic losses Climate change puts over 50% biodiversity of the Asia at risk, so many species could be fragmented or extirpated such as pine forest in Korea or beech tree in Japan (IPCC 2007) Other critical sector relating to climatic influence is human health Siberia ·has been reported a big number of deaths associated with heatwaves, while there is a spreading disease like diarrhoeal, cholera, malaria, etc due to poor drinking water and sewerage system in flood season Malnutrition and Diarrhoeal are posed as substantial risk of human health in year 2000 at Bangladesh, India, Nepal, Maldives, and Myanmar under climate change (IPCC 2007) Specially, the effects of climate change on the countries having high temperature climate and agriculture share has been considered profoundly Those heavy effects are explained that hot climate countries are sensitive to any small change of temperature, while agricultural countries are sensitive to the changing of precipitation Development activities of human are continuously creating greenhouse gas, one of the causes of climate chance, although they also contribute to enhancing adaptation greatly and decreasing vulnerability of vulnerable sectors to the changing of climate In general, climate change affects human life through different channels such as societal (e.g migration, civilization, culture), psychological (e.g mental illness, cognition), physiological (e.g nutrition, starvation), and economic (e.g energy, manufacturing) aspect

In recent years, there are a lot of researches working on influences of climate change The most interesting sector is the climate change's economic consequences Climate change is the result of greenhouse gas emission from human activities Hence, finding out the relationship between climate change and output will help suggest the policies or adaptation strategies in socio-economic sectors to control the effects of climate change within each country Furthermore, we can adjust the changes of climate at our will toward increasing output For instance, the Kyoto Protocol's main target is to balance greenhouse gas emission (C02, CH4 , N20,

and improvement of the human beings This required developed countries to reduce greenhouse gas, and submit annual activity reports; meanwhile, developing countries are encouraged to joint this protocol

Existing studies about the relationship between climate change and the aggregate output have been quite rare until now Most of them were conducted for the countries of OECD (Organization for Economic Cooperation and Development), which are strongly impacted by temperature because they belong to the relatively low temperature areas For some other countries, studies mainly focused on the suffering of separate sectors under climate change such as food production, fish catching, forest, or agriculture Moreover, other published researches estimate the effect of climate change on economic growth based on C02 level My thesis mainly examines the effects caused by the fluctuation of climate, temperature and precipitation on aggregate GDP of the developing countries with the data from 1990 to 2006

1.2 Research Objective

The overall research objective of my study is to investigate the impact of climate fluctuation on country's aggregate output with individual national characteristics within developing countries I am going to consider the economic impact of climate change in the short run on output under two typical elements: temperature and precipitation Besides, my study also discusses some national characteristics, called control effects such as Inflation, GDP per capita growth, population density, urban population, and unemployment Moreover, according to Del (Dell et al, 2008), developing countries move toward to be more agricultural and hotter The reason is pointed out that developing countries focus on agriculture which requires low capital, technology, labor skill level (Tol, 2002) Furthermore, high temperature is advantage condition for increasing pests, pathogens, etc which depreciates agricultural (Masters and McMillan, 2001; and John, 2009) Therefore, my

objective also conducts the effect of climate fluctuation on developing countries with two special characteristics "hot" and "agriculture"

My research questions are "Is the relationship between temperature, precipitation and the output of developing countries significant? - Do the impacts of temperature and precipitation vary across "hot country" and "agriculture country"? This thesis considers the relationship between temperature, precipitation and output, which is divided into five chapters Chapter 1 introduces the problem statement and research objective The Reviewing of empirical studies about the influence of climate change on the output is presented in Chapter 2 - Literature review Chapter 3 shows how I collect data for this paper and set up the model Chapter 4 reports the economic result of regression, and the discussion about comparing with other literature The last chapter, Chapter 5 is used for conclusion and recommendation

CHAPTER 2: LITERATURE REVIEW

2.1 The theory

To demonstrate the relationship between climate and output, I follow the research of John (2009) and Mankiw et al (1992), which is based on the Solow-Swan model According to this model, the climate is considered as "natural resource" input in the Cobb-Douglas production function:

- ka h~ T"Y Yt- at t t t with Yt is output per capita, kt is physical capital per capita, ht is human capital per capita, and Tt represents the climate in average temperature at year t at is all exogenous country-specific variables, which may be composed of temperature in history T0• Thus we specify at= AtTo-0 with At is considered as an technology input

of exogenous elements The exponents a, ~, y, and 0 are assumed positive and identical across countries The y and 0 capture the impact of historical and contemporaneous temperature Let's suppose that the saving rates for physical and human capital per capita: sk and sh are constant; there are constant depreciation: 8 and constant population growth: n When Tt = T 0 = T and if steady-state output per capita exists, y* will be formed:

ln y* = \jllnT- (a +~)B ln(n+g+ 8) + aB ln(sk) + ~e ln (sh) + B lnAt (1) with \jl =- (y + 0)6 and e = 11(1 -a-~) Thus, average temperature changes one percent cause \jl percentage change of steady-state output It is expected \jl < 0 The equation (1) will be reformed by:

Bt is country-specific elements include saving rates, population growth, depreciation, and technology changes

2.2 Empirical Studies

The researches about the relationship between climate shock and economic activities are expanding in many approaches Generally, empirical studies can be divided into three groups as summary of Tol (2009) First group is the studies which measure damaging cost or influence cost function of climate change by statistical method Another category of studies concentrate on effect of climate on separate sectors, such as agriculture, fish catching, forest Final group analyses the effect of climate on aggregate income with many different approaches In this section, I will summarize some empirical literatures in the order of time

Masters and McMillan (2001) is one of the studies that consider temperature's effect on particular sectors, not on aggregate income They used frost- free days as proxy for temperature to test a role of frost frequency in overall economic performance Their regression includes dependent variable - growth of average income of a year, and independent variables such as the scale of economy, seasonal frost, and control variables The scale of economy represents three different dimensions involving the size of domestic population, direct of aggregate trade as one part of GDP, and linguistic barriers in people's communication The frost variable was measured by average frost-day number of each month in winter Finally, other variables capture investment rate of the countries (investment over GDP), human capital accumulation (school enrollment), trade policy, and the quality of domestic institutions Masters and McMillan (200 1) have made conclusions that the income level of temperate countries is higher than tropical countries, which related to the expanding the market and economic scale Their reasonable explanation is that maybe climate assisted temperate regions in growth since they transferred from agriculture into productivity sectors across countries; but it is contrast with tropical countries They also found out that frost can reduce pathogens and pest, which is missed by considering under average temperature Moreover, they identified that the number of frost-free days significantly affects

productivity and population density However, they went on to look at the effect of temperature, or frost on incomes

Like Masters and McMillan (200 1 ), Maddison (2003) also considers the impact

of climate on single sector, not on overall GDP; however, his topic is "the amenity value of climate change" with approaching the household production function He applied the Household Production Function (HPF) theory which combines all marketed goods by a fixed production technology The HPF is commented that it can be the best motivation in implication determining· varieties of surveyed expenditures by climatic variables Moreover, the advantages of the procedure incorporating environmental variables and demand equations are the nature of environmental amenities' role in defining consumption patterns clearly, and limitations of applying initiated utility functions is already understood The approach used two assumptions First, based on the levels of the environmental amenities, the fixed costs are not only translated but also added to or subjected from the household operation; second, the effective prices of goods are increased follow demographic scaling Although these assumptions are required, they are not necessary to be linear because it is a certainly offensive in the condition of climate variables The cross-sectional data of 60 countries is used to examine the climate role in determining consumption then calculate price for a climate variables range

of each country Madison (2003) concluded that climate change benefits on cold countries, while in hot countries any temperature rising decreases welfare and changes in the cost of living in those countries enormously For further exercise, this author suggests more discussions on this paper's limitations For example, under environmental amenities, what is the different effect of demand system and commodity aggregations on prices; applying other function forms which are more flexible to answer the climate is normal, luxury, or inferior good; or using other variables relating to climate such as ratio of heating days, precipitation

There was one other type of studies about the impact of climate change on economic measures and the range of climate effects on social welfare in a given period known as Fankhauser and Tol (2004) They analysed the climate change dynamic effects and social welfare with particular capital and saving formation, then pointed out how to maximize future welfare To set up their model, they based

on DICE model which is considered as the typical model of the world economic DICE does not separate four channels: non market, market, health, capital- through which growth is affected by climate, but it excludes the emission reduction and fix temperature scenario However, they run their model by adjusting and combining four different models to make their regression be available for their purposes First, the Solow-Swan model is used to compare the climate change impact directly on GDP with saving and technology are both exogenous Second, the Ramsey-Cass-Koopmans model which is the primary model structure of DICE is used to compare the important of saving effect with the Solow-Swan model This specification has get saving rate is endogenous, but technical progress is exogenous Third, in the Mankiw-Romer-Weil growth model, savings are fixed and technology is exogenous Comparing this model with the Solow-Swan indicates unlike dynamic effects since climate change influence human capital formation, but the saving rate

is the same in the Solow-Swan model Finally, the Romer model is similar to the Mankiw-Romer-Weil specification without human capital: savings are exogenous, technological progress is endogenous Fankhauser and Tol (2004) assumed that all saving rate and parameter in the Romer model is equal in the Solow-Swan model Generally, the first two models highlighted physical capital formation, but climate change has less been vulnerable; meanwhile it contrasts with the last two models They have paid attention that with a given savings rate, when climate change reduces output, it will also reduce investment, which will lead to decrease capital accumulation or future production Although this paper is more representative for growth model, it still has limitations such as not suggest policy for climate change, used closed economy Therefore, they draw open directs for further researches such

as there are other channels through which changing of climate impact growth, for example: structure economies, not yet saving and capital accumulation; climate would affect the capital supply as well as investment ratio

The study of Nordhaus (2006) directly estimates the impact of welfare, called the statistical approach by using observed variations in prices and spending within single country He estimated the aggregate impact of climate on income for the world with assuming that under respect of climate and geography, the economies are equilibrium in long run It is called "climate-economies equilibrium" However, the climate-economies equilibrium of those areas in which climate shock change very slowly is not reasonable His objective is examining that the farther from equator, the higher output per capital Other studies which used the observation unit are countries, meanwhile Nordhaus (2006) tested the relationship between economic and geography by grid cells of each country It means that economic activity of the world is measured at one degree latitude by one degree longitude scale The reasons he explained for selecting are plentiful global environment data, and statistical independent economic data Although he developed "gridded output" data, or called GCP whose concept is the same as gross domestic production, he still had to use various methodologies to collect data for various group of countries such

as the United State, the high-income countries, the middle-income countries, and the lowest-income countries He run a multivariate equation including dependent variable that is logarithm output per kilometer square, and independent variables that are precipitation, temperature, geography variables such as soil categories, coastline famess One more difference between the studies of Nordhaus (2006) and other economic growth theory is omitting endogenous variables: coastal density, capital formation, technology, health status, education, etc; but focus on exogenous variables: population, geography, etc He compared the existing economic productivity with two scenarios of climate change to measure the climate change impact The first scenario is assumed that there is only temperature change 3°C, precipitation is unchanged; and the result is that global warming reduce world

output 0.93% GDP Second scenario is which gets both assumptions: temperature change as in the first, and precipitation decrease 15% in mid-latitude regions, but precipitation raise 7% in others The consequence of the second indicates that climate change decline world output about 1.05% GDP, however, assumptions of the second scenario are hard to establish because of lack of reality He also proved that the far from equator, the much rising of income per capital Although Nordhaus (2006) conducted the globe into a gross cell product which relies on one degree latitude and one degree longitude by cross-sectional regression include temperature, country-specific variable dummies, the result is need to be considered with caution First, approach weakens linkage between economic data and geographic data Second, follow his consequence, Africa's geography actually has enormous economic disadvantage with respect to temperature, but it is true in comparing with low-latitude countries Third, the negative impact of global warming and greenhouse gas was estimate significantly larger than existing studies base on the G-Econ data Fourth, the result pointed out that there was positive relationship between economic activities and geographic condition, especially on elements: temperature, coastal proximity, and precipitation

In the same data collection manner with Nordhaus (2006), Choiniere and Horowitz (2006) also used cross-section from 97 countries to conduct the income-temperature correlation Their specific objectives are discussing proposition that temperature will low accumulate capital level in hot countries, and what will happen

as global becomes warmer in the future In this paper, they added one more input temperature- in the Cobb-Douglas function; and according to Fankhauser and Tol (2004) they also adopt the Mankiw-Romer-Weil and the Solow-Swan model to set

-up their regression The scale, human capital, function of labor, and physical capital are assumed to be constant Besides, they fixed country - specific such as saving rate, depreciation, and population growth Moreover, in this function, technology is

an exogenous input The author estimated the effect of temperature on output per capital, human and physical capita both singly and jointly To measure physical

capital stock, they collected data on total GDP, rates of capital depreciation, investment, and population; while measuring human capita, they took average school attainment of population aged older 25 from each country This setting up human capital data is different with the Mankiw-Romer-Weil model About the climate data, they based· on capital city's average annual temperature data of each country for 30 years by calculating minimum and maximum monthly temperature data then sum them up to take average They have concluded that the relationship between income and temperature is explicit and powerful since temperature affects more than 45% country's income stock; and the answer for their second objective is that 1% raising temperature will lead to reduce 2-3.5% GDP per capita However, this model is too simple to show a strong general conclusion about the impact of global warming on income or growth, because temperature has just affected capital and labor of production function, but it might influence many other aspects such as savings, population growth, or technological progress Thereby, we need further research developed more sophisticated models about this problem to provide enough exact evidences

While Nordhaus (2006) used cross-sectional data, Dell et al (2008) used panel data to study climate change affects growth They examined the variation of precipitation and annual temperature of 136 countries over 50 years through world economic activities with adding country-specific dummy in equation Their dynamic approach was effective in capturing influence of climate in long run They began estimating simply without lag and based on dynamic growth form which includes country fixed influence, time fixed influence, and climate variables: temperature and precipitation; then they were continuous with 10 temperature lags Furthermore, Dell et al (2008) also conducted political instability effects on the relationship of temperature and growth They divided the research into two parts: in the short run, to measure climate fluctuation effects reduce approximately 1% income per capita since temperature raised 1 °C, and in long run to consider historical climate change effects They found three primary conclusions First,

higher 1 °C of temperatures significantly reduce total GDP around 11% in poor countries, but slightly in rich countries Therefore, the income gap between poor and rich countries will be increase by clime change in future Second, the negative effect of temperatures on growth rate is more than output Final, higher temperatures reduce agriculture output, industrial output, aggregate investment, and expanding political instability in poor countries However, Dell et al (2008) run regression by measuring effect of climate less than weather, so they pointed out issue that weather effect is not the same with climate effect on economic Moreover, although world income mainly contributed by rich countries, their estimate showed that temperature has just increased income of rich countries slightly In the other hand, there are some precise distinctions between this paper and other traditional studies They used aggregate data which can catch and reflect interactions between important channels; then they considered climate effects on rich and poor countries differently Finally, they examined the possibility of temperature affects on both growth rate and level of output The authors also suggest further analysis about particular causal the climate-economy mechanisms

Both previous studies had been relevant to the research of John (2009) He also conducted the relationship between income and temperature by top-down approach with mortality data The reason is that he thinks that mortality data is representative for temperature's historical role, and has strong correlation with average temperature of countries He used cross-section data for income per capita, changing temperature in long run, and some explanatory variables of 100 countries which stand for 94.7% world population and 95.4% world GDP The dependent variable is the average GDP per capita from 2002 to 2004, which accounts for both inside and outside economic activities There are three reasons for John (2009) took averaged monthly temperature of stations in cities to construct the year average temperature data, which is same as Choiniere and Horowitz (2006): first, he considers that all growth of area, infrastructure will increase temperature in cities more than other places of countries; second, it seems to be the most exogenous;

finally, it is easy to calculate This is the first difference between John (2009), Nordhaus (2006), and Dell et al (2008) While Nordhaus (2006) took geographic average temperature, John (2009) said that unit was one-degree cell of latitude/longitude than countries; consequently, economic data of some cells were excluded On the other hand, John (2009) considers that using the population-weighted average temperature as Dell et al (2008) has less endogenous, although it

is unknown and slight implications Besides, John (2009) also used other explanatory variables which have same exogenous with temperature on country level, such as saving rate, percentage of population living around 1 OOkm at a coast, population growth, energy resource, and form of government He demonstrated the effect of temperature on GDP by depending on the Slow-Swan model and the Cobb-Douglas production function To estimate across regression, he used the log-log form, and the log-cubic form to capture the hump_ shaped relationship of very cold climate and economic activities Briefly, John (2009) analyzed directly the effect of temperature changing in long period on income with combine all sectors The consequence that he has found out is that increasing 1 °C will decrease around 12% world income without mortality data In the contrast, temperature flattens world income explicitly with mortality since 1 °C average temperature raising leads to 3.8-4.2% GDP reducing This figure is considered be bigger than result found by bottom-up approach He also analyzed separately 38 African countries and 14 Soviet republic countries because of their size in the effect of temperature on income, however, his report had just demonstrated the pervasiveness of the income-temperature correlation due to less observation Moreover, John (2009) put dummy variable for wealthy resources countries such as the OPEC, he has come to the conclusion that there are not difference between these countries and others in the log-log regression Although all his estimations are significant, there are some limitations in this paper Measuring the income-temperature relationship when temperature gets over 2°C may be impossible since it requires expanding sample The prediction does not include non-market losses or economic growth because it

does not capture exogenous factors such as technological change Finally, it cannot reflex exactly in measuring temperature change by the cross-sectional model

in aspect the output of country in short run The aggregate economic effects of climate fluctuation on country in term of monetary value, is measured by country's GDP per capita which is representative for the country's total economic benefit (or cost) Hence, it is necessary to conduct the correlation between climate fluctuation and production then suggest policies which is not only essential for adapting changing of climate, but also relying on those changing to develop the world economy Moreover, there is different characteristic between each country, so keeping some country-specific variables about social-economy prospective is important The control variables differ from each one in each country as inflation, population density, urban population, unemployment, the growth rate of GDP per capita However, human activities through production, especially such an industry, also impact environment; then create greenhouse gas in long run Thus, in this paper's framework I pass the influence of production on greenhouse gas (which is showed by the grey dotted-line) for further research

1 I

GDP per capita

Figure 2.1: The way climate fluctuation affects the country's output

This chapter has described briefly studies about the relationship between climate change and the country output both in short time and long time by many different approaches The effect of climate change with temperature and precipitation - two typical elements on the output is significant However, poor countries have to be vulnerable changing of climate more than rich countries while rich countries get benefit on GDP from rising temperature, it contradicts poor countries Besides, unobserved effects at country level also must be considered in this relationship to make robustness the result

CHAPTER3:METHODOLOGY

3.1 Econometric model

To analysis by econometric, I am going to conduct a panel analysis for testing the relationship between climate fluctuation and the output The reason for choosing panel data is that it is useful in spreading the number of observations to avoid bias

in analysis progress due to data of developing countries is very spare to collect, especially far past years In the econometric model, GDP per capita is a function of temperature, precipitation and control variables such as inflation, population density, urban population, GDP per capita growth rate, and unemployment I take log of the climate variables and GDP per capita to demonstrate the elasticity of variables Economic growth theories have mentioned the role of both endogenous factors such as policy, technology; and exogenous factors as population Although those factors are important, in this study I have just focused on exogenous factors because developing countries are considered that there are not much differences about endogenous factors

According to previous empirical studies, for instance, Mendelsohn et al (2000), Nordhaus (2006), and John (2009) the relationship of GDP with temperature and precipitation is not linear, but in a quadratic equation It implies that an increasing temperature and precipitation originally benefits the output of country when GDP gets upward trend, however there is a peak When rising of temperature and precipitation reaches certain point at peak, then it makes GDP decrease In this study, I apply a polynomial regression base on three previous models to test the hump-shape relationship between dependent variable- GDP per capita and two independent variables- temperature and precipitation of developing countries

The general function form is:

Ln(GDP)it = ~ 0 + ~~ x Ln(Temperature )it+ ~ 2 x Ln(Precipitation)it

+ ~3 x (Temperatureiit + ~4 x (Precipitation)2it + ~s x Zit+ ai + Ei,t (3) Consider country i at year t, thus Z is control variables such as inflation, population density, urban population, GDP per capita growth, unemployment of country i at

year t While, a is a fixed effect for the country, and represented for unobserved elements that impact GDP per capita within country level such as saving, population, Ei,t is an error term

The control variables are assumed that they are linear in trend The regression is also assumed that all countries get same slope, however, the intercept may be different for each country due to different country's structure about population growth, saving, etc (Dijkgraaf et al, 2005, and Aslanidis, 2009)

The analysis will follow "top-down" approach (Mendelsohn et al, 2000) with three steps The "top-down" approach is simple to consider changing of the relationship of climate change and output under different variables, although the numbers of observations might be reduced because of missing data

(i) First, I test the pure impact of climate change with output Thus, a function

which GDP per capita dependents on temperature and precipitation, is applied

(ii) Second, I test the relationship of climate change with output under impact of

some other elements which are mentioned in the Cobb-Douglas production function So, I conduct function (i) with adding control variables

(iii) Final, as there is a variety of unobserved inherent influences between

countries, those influences have different and important effects on the country's output Therefore, I add fixed effects into function (ii) to examine the overall relationship of climate change and output

I am going to apply upper three steps into five models The first three models are used to answer the first research question about the relationship between temperature and precipitation with the output of developing countries The last two models are tested to answer the second research question about the vary impacts of temperature and precipitation across "hot country" and "agriculture country" The analysis process is summarized as follow:

• Modell:

Ln(GDP)it =Po+ PIx Ln(Temperatureh + P2 x Ln(Precipitation)it

+ P3 x (Temperature)2it + P4 x (Precipitation)2it + Ei,t

• Model2:

Ln(GDP)it =Po+ P1 x Ln(Temperature)it + P2 x Ln(Precipitation)it

+ P3 x (Temperature)2it + P4 x (Precipitation)2it + Ps x Zit (inflation, population density, urban population, GDP per capita growth, unemployment) + Ei,t

• Model3:

Ln(GDP)it = Po+ PI x Ln(Temperature )it +P2 x Ln(Precipitation)it

+ P3 x (Temperature)2it + P4 x (Precipitation)2it + Ps x Zit (inflation, population density, urban population, GDP per capita growth, unemployment)

+ ai (country fixed effects)+ Ei,t

• Model4:

Ln(GDP)it = ~ 0 + ~ 1 x Hot country dummy+ ~2 x Ln(Temperature )it

+~ 3 x Ln(Precipitation)it + ~4 x (Temperature )2it + ~5 x (Precipitation)2it

+ ~ 6 x Ln(Temperature)it x Hot country dummy + ~ 7 x Ln(Precipitation)it x Hot country dummy + ~8 x (Temperature)2it x Hot country dummy + ~9 x (Precipitation)2it x Hot country dummy + ~10 x Zit (inflation, population density, urban population, GDP per capita growth, unemployment)

+ ai (country fixed effects)+ si,t

• ModelS:

Ln(GDP)it = ~o+ ~1 x Agriculture dummy+ ~2 x Ln(Temperature )it

+~3 x Ln(Precipitation)h + ~4 x (Temperature )2it + ~5 x (Precipitation)2it

+ ~ 6 x Ln(Temperature)it x Agriculture dummy + ~ 7 x Ln(Precipitation)it x Agriculture dummy

it x Agriculture dummy + ~9 x (Precipitation)2it x Agriculture dummy + ~1o x Zit (inflation, population density, urban population, GDP per capita growth, unemployment)

+ ai (country fixed effects)+ Ei,t

The climate data are the temperature and precipitation data for all countries that are collected from the United Nation Development Programme called UNDP The UNDP Climate Change Country Profiles is one of the researches of UNDP about Climate Systems and Policy It includes the data of 62 developing countries under both observed and model data, and the report of diagrams demonstrating and map for each country (http://www geog.ox.ac uk/research/climate/proj ects/undp-cp/) The data from UNDP is already country level climate data with taking average per year Two independent variables of the study chosen observed were the data which temperature was measured by Cell degree (°C), and precipitation was measured by millimeter of rain fall (mm) The advantages of these data are that they are formatted in text format which is easy to read, collect, and manipulated

The data for world development indicators of countries are collected from the database of the World Bank (http://data.worldbank.org/) Basing on empirical studies, especially according to Maddison et al (2005), I have identified five chief control variables such as Inflation, GDP per capita growth, population density, urban population, and unemployment However, they have a lot of missing data of some periods of the developing countries I will state the effect of control variables

on country's output in the next paragraph

First, the GDP per capita growth not only describes the country's economic expanding, but it also is used to measure accumulation means in production So, it is