A simple Ricardo-Malthusian model of population, forest and biodiversity 

Get Complete Project Material File(s) Now! »

Tests for overdispersion

Before any estimation, it seems important to test for overdispersion, which implies checking the mean-variance equality assumption of the Poisson distribution, as huge differences are observed between the mean and the variance of the series on threat-ened animal and plant species (Table 2.1). Dean and Lawless (1989) and Hilbe (2011) propose a Z-score test which seems straightforward. Applying the latter to the different model specifications, we find results suggesting overdispersion in the counts on threatened species (Table 2.2). Thus, modelling counts of threatened species, overdispersion should be considered as in NB model.

Determinants of biodiversity loss

Considering the counts of threatened species to be NB distributed, the econometric literature indicates that NB estimates are asymptotically normal, efficient and unbi-ased. However, this unbiasedness is violated in presence of regressor endogeneity, a  previously mentioned regarding GDP per capita. We tackle this issue by exploiting the CFA (Wooldridge, 2014, 2015) discussed above. To instrument for GDP per capita, we rely on political institutions and government expenditures, namely the index of control for corruption and the share of goods and services expenditure (% of total government expenditures). The latter seem to be good instruments since economic theory acknowledges government expenditures and good political insti-tutions as driving macroeconomic performances. Also, they show high correlation with lnGDP per capita.
Since our dataset has a very small T characterized with low time-variability in the response variable for a relatively high number of individuals N , we rely on pooling the data. In addition to over-dispersion tests and first stage regressions, we report the results of estimating the parameters of different NB model specifications in Table 2.2. Observing the results for both animal and plant species, one notices that compared to a linear fit a quadratic specification in lnGDP per capita fits better the data. This corroborates the discussion in Section 3 which suggests a non-linear modelling for the economic growth and threatened species nexus. By comparing information criteria, Model 4 shows larger predictive power and therefore will be considered to discuss the peaceful cohabitation hypothesis.
• Animal species: Our results broadly indicate that the expansion of human habitat’s characteristics is not neutral in biodiversity loss, as income per capita and human population dynamics significantly affect the number of threatened animal species. More precisely, our parametric estimations reveal a non-linear relationship between income per capita and threatened animal species implying that economic activities increasingly threaten biodiversity in low-income countries, while it decreas-ingly does in high-income countries. Such a result, largely known in the existing literature as the presence of an EKC relationship, seems to hold as the parameter of the quadratic term remains statistically significant throughout specifications 2, 3 and 4. This suggests an inverted U-shaped relationship for species, which in-deed supports a peaceful cohabitation in high-income countries, as suggested by the patterns observed LPI’s over 1970-2005.
Besides tests of the peaceful cohabitation hypothesis, controlling for species richness and climate zone by using the total number of animal species identified and distance from equator indicates that more threatened animal species lie in species-rich regions while less are found in countries far from the equator. These results imply that in tropical zones, where biodiversity mostly lies, relatively high species are threatened by extinction.
As Polasky et al. (2002) and by Alam and Quyen (2007) argue that trade openness and agricultural production lead to deforestation and species loss in the South, we include agricultural production, exports, imports and forest cover in the regression. It results that forest size is positively linked to the number of threatened animal species.
Forests largely serving as natural habitat for species, it is not surprising to observe that the larger forest size there is in a country, the more threatened species it shelters. Concerning agricultural land, its GDP share and exports, our results do not globally support conclusions by Alam and Quyen (2007). Agricultural production (share in GDP) and exports are not to blame for threatening animal species, at least when considering the whole sample. Additionally, FDI and exploitation of natural resources do not globally drive biodiversity loss. Countries of our sample being at different stages of development, separating countries according to income level and considering geographical subsamples will probably help more clearly apprehend the role of trade openness and agriculture in endangering species.
A final interesting result in the case of animal species is the role of human population growth. Population density is found to have a positively effect on the number of species at threat, meaning that the higher human population is, the more threatened animal species there are. Such a result underlines the existence of a possible competitive exclusion over habitat between human population and animal species, all other things being equal.

Regional heterogeneities

In order to investigate whether heterogeneities exist over regions, the same analysis is performed using data classified by continent (Table 2.3). Thereby, we further use the control function approach in solving for potential endogeneity with respect to income per capita.
• Africa: Being mostly in tropical zones and largely covered by forest, animal and plant species-rich African countries are also those sheltering relatively high numbers of species threatened by extinction. Besides these environmental factors, FDI, exports and population growth appear to be the main factors threatening biodiversity in Africa, supporting the conflicting cohabitation argument. Exports, mostly of primary goods and raw material, threaten species in Africa.
• America: A non-linear relationship is observed between the number of threat-ened animal and plant species and income. Besides climate zones, our results show that larger forest covered American countries also shelter more threatened animal and plant species. Increasing agricultural land and exports enhance the threat to biological species. However, no conflict is observed between population dynamics and others biological species in America.
• Asia: The results suggest an inverted U-shaped curve between the number of threatened species and income per capita, similar to those observed in Figure 2.2 for the whole sample. In addition, species-rich and large forest covered Asian countries shelter relatively high threatened animal and plant species. Human population dynamics are not significantly harmful to biodiversity in Asia.
• Europe: Regarding animal species, the results also indicate that species-rich and large forest covered European countries shelter relative high threatened species. The share of agricultural land positively drives biodiversity loss, while population density has no impact on endangered species. Finally, relatively high plant species are at threat in European countries which larger share of industrial value-added in GDP.
This regional analysis has revealed the divergent role played by FDI, exports and population dynamics in driving species loss. While FDI, human population dynamics and exports promote biodiversity loss in Africa, they are found to be insignificant to species loss in Europe and Asia. In complement, we disentangle countries according to income levels using the sample median of GDP per capita (in log) to distinguish high and low-income countries.
• High-income countries: An inverted U-shaped relationship appears between GDP per capita and threatened species, supporting a possible harmonious cohabita-tion between biological species and human economic activities. However, the initial condition for biodiversity (species richness) matters as well, since the more species-rich high-income countries are, the more threatened animal species they shelter. Among our control variables, it is to observe that exports, industrial production and population dynamics do not harm biological species in developed countries.
• Low-income countries: Focusing on threatened plant species, our results signal a upward trend which implies that more plant species are threatened by extinction with increasing income per capita. This contradicts the peaceful cohab-itation hypothesis. Moreover, in contrast to developed countries, FDI, industrial production, exportation and human population growth positively drive species loss, enhancing conflicts between human and natural habitats in developing countries.


A simple Ricardo-Malthusian model of popu-lation, forest and biodiversity

Abstract: This paper assesses the interactions between human and nature, arguing that population growth and forest resources use cause natural habitat conversion, which resolves into biodiversity loss. Relying on profit and utility maximization behaviours, we describe the joint evolution of population, forest and species stock by a dynamic system characterized by a locally stable steady state. Compared to existing studies, we enlighten the possibility of total extinction of biological species. Furthermore, our analysis supports an impossible peaceful cohabitation, as in the presence of human population growth, forest resources and species stock diverge from their carrying capacity. Finally, scenarios analyses associated with high fertility and preference for the resource-based good globally indicate rapid population growth followed by a sudden drop.

The basic structure of the model

As in population-resources models, this paper considers a two production sectors economy: A manufacture and a forest resource harvest sector. The manufactured good is produced by a representative firm using only labour, LM , while the resource-harvest sector employs labour, LH , and forest resources, F . Labour is freely mobile across sectors, implying wage equality between sectors (wH = wM = w). The struc-ture of the model described hereafter closely follow resource-population discussions in Brander and Taylor (1998), Dalton and Coats (2000), Nagase and Uehara (2011), among others.

Firms’ behaviour

Manufactures: They are considered as numeraire using a Ricardian production func-tion YM,t = LM,t, where LM stands for the quantity of labour used in sector M . Assuming the price of the good to equal one, the optimal behaviour of the repre-sentative firm is: M ax ΠM,t with ΠM,t = YM,t − wtLM,t ≡ LM,t − wtLM,t (3.1).
Harvest sector : Forest resources use is governed by the supply of good, H, using the 43 = wtγ/pH,t

The basic structure of the model

well-known Schaefer (1957) production function, YH,t ≡ H(Ft) = qEtFt, where Et is the harvest effort (labour) and q a positive parameter to be seen a scaling parameter or level of technological knowledge. Since there are no property rights over land, the firm i hires a quantity of labour, LH,t ≡ Et, to maximize the following function: M ax ΠH,t with ΠH,t = pH YH,t − wH,tLH,t ≡ pH qLH,tFt − wH,tLH,t (3.2).
First order condition of profit maximization yields: pH qFt = wH,t which implies: pH,t = wH,t (3.3). (3) expresses the supply price of the harvest good, pH,t, as positively dependent on the wage rate and negatively on forest resources harvested in the production process.

Preference and budget constraints

At each period t, a new generation of agents is born and lives 2 periods, childhood and adulthood. Adult individuals in t (born in t − 1) are endowed with one unit of time which they supply inelastically to labour force participation to earn wt. By definition, children consume a fraction of their parents’ time endowment and do not make any economic decision. Thus, adult individuals (Nt) choose the optimal mixture of M and H to maximize their utility function. Such formulations of in-dividuals’ behaviour are intensely described in De La Croix and Michel (2002) and Galor (2011).
The utility function of the representative agent is defined over consumption of the resources and harvest goods Ht and Mt, respectively cH,t ≡ CH,t/Nt and cM,t ≡ CM,t/Nt. The problem of the representative individual is: M ax U (cH,t, cM,t) with U (cH,t, cM,t) = (cH,t)γ (cM,t)1−γ where γ ∈ (0, 1) (3.4) ht ,mt.

Table of contents :

1 Chapitre introductif 
1.1 Motivation
1.2 Aperçu et contribution
Introductory Chapter
2 Is there a peaceful cohabitation between human and natural habitats? 
2.1 Introduction
2.2 Related literature
2.3 Data and descriptive statistics
2.3.1 The data
2.3.2 Data on threatened species
2.4 Econometric specification
2.4.1 The count model
2.4.2 Endogeneity
2.5 Estimation results
2.5.1 Tests for overdispersion
2.5.2 Determinants of biodiversity loss
2.5.3 Regional heterogeneities
2.6 Discussion: Beyond the peaceful cohabitation
2.7 Concluding remarks
3 A simple Ricardo-Malthusian model of population, forest and biodiversity 
3.1 Introduction
3.2 A brief literature review
3.3 The basic structure of the model
3.3.1 Firms’ behaviour
3.3.2 Preference and budget constraints
3.3.3 Competitive equilibrium and market clearing
3.4 Dynamics of population, forest and species
3.4.1 Population dynamics
3.4.2 Forest dynamics
3.4.3 Dynamics of species stock
3.5 Steady state and linear stability analysis
3.5.1 Steady state
3.5.2 Linear stability analysis
3.5.3 Population, forest cover and species stock interactions
3.6 Scenarios analysis
3.6.1 Applying the population-forest-biodiversity model to Easter Island
3.6.2 Population-forest-biodiversity in a developing resource-intensive economy
3.7 Discussion and concluding remarks
3.7.1 Brander and Taylor, HANDY and the Population-Forest-Biodiversity model
3.7.2 Concluding remarks
4 Do Species-poor forests fool conservation policies? 
4.1 Introduction
4.2 Related literature
4.3 The data
4.3.1 Descriptive statistics
4.3.2 An Insight into the data on forest and PAs cover
4.4 The PAs-income, forest and species richness
4.5 Econometric model
4.6 Estimation results
4.6.1 Some tests: Evidence of spatial dependence in PAs
4.6.2 Estimating spatial lag models of PAs
4.7 Robustness and heterogeneity analysis
4.7.1 Robustness check
4.7.2 Regional analysis
4.8 Concluding Remarks
5 Institutions and geography: A « two sides of the same coin » story of primary energy use 
5.1 Introduction
5.2 Related literature
5.3 Data and descriptive statistics
5.4 Econometric model
5.5 Results and discussion
5.5.1 Modelling primary energy use: Preliminary tests
5.5.2 Results of estimating spatial FE models for primary energy use100
5.6 Robustness, role of geographical location and functional forms
5.6.1 Robustness check
5.6.2 Does location matter to primary energy use?
5.6.3 The primary energy use, income and urbanization nexus
5.7 Concluding remarks
Additional Materials
6 Conclusion générale 
6.1 Contexte et motivation
6.2 Principaux résultats
6.3 Implication de politiques environnementales
6.4 Limites et extensions possibles
Concluding Chapter


Related Posts