ORANI-SL – A Computable General Equilibrium Model to Assess the Climate Change Impacts on Agriculture in Sri Lanka Developing a CGE database for ORANI-SL

This paper presents a detailed technical description of features added to the standard ORANI model to construct a CGE for Sri Lanka, namely, ORANI- SL. The model distinguishes between rain-fed and irrigated agricultural land and introduces water as a production factor. In addition, the new production structure offers the option of substitution between rain-fed and irrigated agricultural lands. The paper also describes the model's database and the steps of the data transformation process from the latest available input-output table of Sri Lanka to the ORANI-SL database, which is necessary for model simulation. The new model could be used to analyse the economy wide impacts of climate change in Sri Lanka, study the economic implications of increased irrigation capacity or sustainable water usage in Sri Lanka, and evaluate the efficacy of various other adaptive strategies to cope with the impacts of climate change. In addition this model can be easily adapted to another country.


Introduction
Agriculture is globally one of the major forms of human activity and livelihood, employing available land and water resources to feed the population. Undeniably, research in the related disciplines of soil chemistry, plant biochemistry, and environment studies, as well as the development of technology has enabled the agricultural sector to boost its production during the past century. Yet, the uneven distribution of these benefits has not provided a proper solution for hunger and poverty in some parts of the world (Barrett, 2010). Currently, the agricultural sector faces a monumental challenge to feed the growing global population, projected to reach nine billion by 2050 (Godfray et al., 2010).
The agricultural sector and its production are intrinsically interlinked with climate and the environment. Topping all existing global challenges to food production and agroindustry growth, is the impact of climate change. The agricultural sector will experience a greater economic impact from climate change compared to any other industries (Cline, 2007;Mendelsohn, 2008). In addition, rain-fed agriculture is still the main livelihood of the majority of the rural populations of developing countries. Predicted rainy seasons with sufficient rainfall and optimum temperatures will continue to enhance production ensuring food security. However, unpredictability of rainy seasons and the predicted rainfall shortfalls will reduce yield, leading to food insecurity. As it is both an island in the Indian Ocean and a developing nation, Sri Lanka is highly vulnerable to changes in climate and the economic impacts of climate change. It is evident from existing literature that the mean annual temperature and extreme weather events on the island are on the increase, and the rainfall patterns are changing, making them unpredictable (Esham & Garforth, 2013;Panabokke & Punyawardena, 2010).
Many studies on climate change impacts have clearly elucidated that the changes in climate in Sri Lanka have directly and indirectly affected the agricultural sector, imposing significant consequences for the economy and national food security (Ahmed & Suphachalasai, 2014;De Costa, 2008De Silva, 2009;Eeswaran, 2018;Esham & Garforth, 2013;Premalal, 2009). However, the implementation of specific planned developments in the agricultural sector and its productivity are regarded as crucial in achieving sustainable development goals in developing countries, and Sri Lanka is no exemption.
Expansion of irrigated land areas will ensure the availability of the right volume of water at the right time for agricultural production and reduce the risk of production losses due to rain-water unpredictability. Moreover, the water demand is projected to increase in the future, particularly due to climate change (Berrittella, Hoekstra, Rehdanz, Roson, & Tol, 2007). Aside from the impact of climate change, the scarcity of land and water resources offers very limited options for expanding the arable land areas and increasing food production. Thus, quantitative assessments on the effects of alternative investment options and adaptation strategies, such as promoting irrigated agriculture, have become vital to counter the agricultural and economic challenges and impacts of climate change.
According to the Food and Agriculture Organization of the United Nations, agriculture accounts for 70 percent of global water withdrawal and 90 percent of water consumption (Kohli, Frenken, & Spottorno, 2010). Moreover, irrigated agriculture, which covers 20 percent of the world's agricultural land, contributes about 40 percent of world crop production (Kohli et al., 2010). Under such circumstances, it is vital to control and optimise water distribution and consumption in the agricultural sector as an adaptation strategy against climate change. In contrast, ineffectual water-usage policies providing water as an under-priced commodity or free resource have hindered the efficient distribution and usage of water resources. As national budgets and annual financial expenditure returns do not usually indicate water as a disaggregated component, as there are no recorded economic transactions related to water, this has limited the inclusion of water resources in economic models.
Despite these challenges, the Computable General Equilibrium (CGE) model, an accepted economic modelling approach, has been used by many scholars for climate change impact and adaptation assessments. CGE models can assess the impact of climate change on the agricultural sector by considering its interactions with other economic sectors. CGE models can determine the impact of climate change on the whole economy and evaluate the impact of different adaptation strategies. This paper presents a development of the ORANI-G model known as ORANI-SL, which introduces water as a factor of agricultural production and distinguishes between rainfed and irrigated agricultural land. It facilitates the assessment of adaptation strategies and alternative investment options, as to how these meet the challenges of climate change impacts on the agricultural sector in Sri Lanka and thereby the whole economy of the nation.
The paper is structured as follows: the first part of the paper reviews the literature on similar economic models. The second section explains in detail the revised version of the ORANI-SL model. The third section describes the database development, and the final section presents a conclusion.

Literature Review
Scholars have used economic models to investigate numerous water-related issues, such as its availability and usage in the agricultural sector and water policy. Partial equilibrium and CGE models have been applied to the analysis of similar water-related issues. Partial equilibrium methods can analyse water-related issues in a particular sector, assuming that it has no impact on other economic sectors, while CGE models can analyse impacts on the whole economy.
Nevertheless, incorporating water factors into a CGE model is challenging due to the following factors. The agricultural sector consumes a major portion of water for production, and it is natural to consider water as a factor within the production function.
However, as water is generally a non-market commodity it is difficult to calibrate the parameters which determine the marginal productivity of water for different sectors to include in a CGE model. Many input-output tables (I-O tables) are available for water collection, treatment, and supply, including the I-O table of Sri Lanka which records the water distribution transactions to the other industries. However, these industries do not produce water and prices do not reflect the real production costs. Moreover, it is difficult to calculate the exact amount of rain-water and ground-water used for agricultural production, which also complicates incorporating water factors into a CGE model.
To overcome these challenges, scholars have used different approaches to incorporate the water component into CGE models. Berrittella et al. (2007), Sahlén (2008) and Tirado, Gomez, and Lozano (2006) introduced the concept of a water industry into the model, This method transforms rough water into effective water as an exogenous intermediate input for production and consumption. Another approach has been to embed the value of water within the value of land, assuming the source or supply of water to be a factor that determines soil characteristics, and thereby land value (Calzadilla, Rehdanz, & Tol, 2010). One group of researchers used agronomic studies to obtain water productivity values to include in CGE models (Roson & Sartori, 2015). Calzadilla, Rehdanz, Roson, Sartori, and Tol (2017) classified water-related simulation experiments into two categories. The CGE model has been used to assess the economy wide impacts due to water resource changes driven by climate change (Calzadilla, Zhu, Rehdanz, Tol, & Ringler, 2013;Roson & Sartori, 2015;Taheripour, Hertel, & Liu, 2013), as well as to evaluate the impact of economic policies on water demand, consumption, etc. (Berrittella et al., 2007;Calzadilla, Rehdanz, & Tol, 2011b;Gomez, Tirado, & Rey-Maquieira, 2004).
As the ORANI-SL model differentiates between rain-fed and irrigated agriculture by introducing water as a factor of production embedded in land rental, this study mainly reviews models that considered water as a factor of the land component. Such studies have been applied both on a regional and global scale. Berck, Robinson, and Goldman (1991), followed by Robinson and Gehlhar (1995), Mukherjee (1996), Seung, Harris, MacDiarmid, and Shaw (1998), and Decaluwe, Patry, and Savard (1999) all embedded the water component into land rental to analyse the regional water-related issues in the '90s. Berck et al. (1991) analysed the impact of reduced amounts of water as an agricultural input on aggregate Gross Domestic Production (GDP) for the southern portion of the San Joaquin Valley in California, using a multi-level function production technology for the agricultural sector with low and high elasticity variants. The agricultural sectoral capital varies with land use in the low elasticity variant and is fixed in the high elasticity variant. Robinson and Gehlhar (1995) and Mukherjee (1996) used a similar production function to analyse the efficiency of land and water use economic policies. At the top level of the nested structure of the sectoral production function, the sectoral output is a linear function of real value added and intermediate inputs. The intermediate inputs are required as fixed input-output coefficients and real value added is a constant elasticity of substitution (CES) function of primary inputs (labour, capital, and land/water aggregate), where the land/water aggregate is a linear aggregation. Seung et al. (1998) used a regional CGE model to assess the impact of water reallocation from agricultural to recreational use in rural Nevada and California. The production function has a quadruple-nested structure.
At the bottom level, water and acreage are combined in fixed proportions to produce land. Capital and land is then combined with a CES function to produce the capital/land aggregate, and another CES function is used to produce value added by combining labour and capital/land aggregate. At the top-level, the output is a linear function of intermediates and value added. Decaluwe et al. (1999) analysed different water pricing schemes in Morocco. The agricultural production function consists of a four-level nested structure. At the bottom level, land and capital are combined with a CES function to produce land and capital composite, and another CES function is used to combine water and fertilizer to produce intermediate inputs. Diao and Roe (2000), Gomez et al. (2004), D. C. Peterson, Dwyer, Appels, and Fry  (2009) all used regional CGE models to analyse water-related issues in the 2000s. (Table 1).

ORANI-SL Model
ORANI-SL is a single country CGE model. The ORANI-SL model core follows the framework of the ORANl-G single country generic computable general equilibrium model (P. B. Dixon, 1982;J. Horridge, 2000). ORANI is a comparative static CGE model of the Australian economy, developed by the Centre of Policy Studies (CoPS) and Impact Project at Monash University in Australia (Mark Horridge, 2003). It was developed to analyse policy successes and failures in Australia. ORANI-G is an adapted version of ORANI, which has served as a foundation for the construction of many new models (Lkhanaajav, 2016).
Since the study's primary focus is on the agricultural sector, the model was modified by disaggregating agricultural land into irrigated land and rain-fed land. Achieving this primary fundamental modification involved the following three main steps: The first step required calculating the model's data requirements -the traditional I-O As the major modification involved the production structure, an overview of the model's new adaptive production structure follows (Figure 1) In the ORANI-SL model, it is also assumed that rain-fed lands can be converted into irrigation lands by applying water. Elasticity of substitution values can easily be obtained from the GTAP-W model provided for the South Asian region (Calzadilla, Rehdanz, & Tol, 2011a). However, it is more appropriate to calculate the elasticity of substitution values specific to Sri Lanka. Therefore, the elasticity of substitution between irrigated land and rain-fed land was calculated using the price elasticity value of water estimated by Dharmaratna and Parasnis (2010) for Sri Lanka. Calculations were done using the following equations: The two-input production function is represented in the following equation; Where, Y is the output, z is the water input and x all other inputs. Considering p as the composite price of all the other inputs and r as the price of water, the cost function can be expressed as; According to production efficiency, the marginal rate of technical substitution (RTS) equals the price ratio of the two inputs; The above CES function is defined for > −1, ≠ 0, positive levels of inputs, continuous, differentiable, monotonic and strictly quasi-concave, and exhibits constant returns to scale (Gohin & Hertel, 2003). Accordingly; If we consider two different water prices as r and r (1 + ), the above equation can be expressed as, Considering equations 6 and 7; 1 +1 = 2 +1 (1 + ) Price elasticity of demand can be represented as; Equation 9 can be expressed as; Therefore, Combining equations 8 and 11, elasticity of substitution can be given as the function of the price elasticity of demand for water; The estimated value for price elasticity demand for water as calculated by Dharmaratna and Parasnis (2010) is -0.15. Using equation 12, the elasticity of substitution between irrigated land and rain-fed land was calculated as 0.05. The elasticity of substitution used in the GTAP-W model for the South Asia region is 0.06. The ORANI -SL model, however, will incorporate the calculated value for Sri Lanka based on the elasticity of substitution between irrigated and rain-fed lands.
The model was then validated to ensure its accuracy. A real homogeneity test was conducted to establish whether the model satisfied the constant returns to scale condition. All real variables in the exogenous list were shocked by X% and examined as to whether all were changed by X%, while the nominal variables remained unchanged in the solution. The ORANI-SL model satisfied the above homogeneity test.  Table 2 shows the disposition of industry i's output for final consumption, capital formation, and exports. The summation of values in each row of Q1 and Q2 provides the total usage of each industry's goods and services. Quadrant three (Q3) in Table 2 shows the entries usually referred to as value additions. These include compensation for employees, taxes on production, consumption of fixed capital, and operating surpluses. Since this study uses the ORANI-NM (no margins) model, margins were treated as services in the USE matrix (e.g., transport margin is viewed as transport service).

Database Development Input-Output Data Tables
Therefore, the margin values of the transport and trade were also added to the base value of transport and trade. Similarly, margin values were also incorporated into the supply matrix, in order to leave the margin matrices as null.    The row totals in Tables 3,4, and 5 represent the number of total goods and services (Commodities "C") supplied by each industry at unit cost, while the column total denotes the total demand or the total amount of commodities absorbed by each industry ("I") for its current production. The simplified version of Sri Lanka's SUTs, exhibits three commodities and three industries, namely, "Agriculture," "Manufacturing," and "Services." Thus, the matrix is symmetric, having an equal number of commodities and industries. Similarly, this study employs a symmetric I-O table benchmarked for the   year 2010, the most up-to-date I-O table published by the DCS, consisting of 127 industries and 127 commodities. A highly aggregated version of this I-O table based on real values can be found in Annex 2.
In the ORANI-G model, total demand must be equal to total domestic supply.
Therefore, the column totals in the total USE table (Table 3) must be equal to the column totals of the MAKE matrix (Table 6), while the row totals of the domestic USE table (Table 4) must be equal to the row totals of the MAKE matrix (Table 6). Hence, the MAKE matrix denotes each commodity's domestic production level (row headings) by each industry (column headings).
In addition to the I-O table, different elasticity data, such as elasticities for substitution between domestic and imported sources (Armington elasticities) of commodities, elasticities of substitution between primary factors, household expenditure parameters, and export demand elasticities, are required to implement the model. Readily available data for elasticities were obtained from the GTAP database.

I-O Table data transformation into the ORANI-SL format
The next requirement is to transform the data in Sri Lanka's I-O table into the ORANI-SL format (Figure 3) to implement the model. As the ORANI-G model is accessible for modification, the ORANI-SL has extended the standard ORANI-G structure to subdivide the agricultural land factor into rain-fed agricultural land irrigated agricultural land.
Any economic system should allocate its scarce resources among economic agents such as producers, investors, households, non-nationals, and governments for their competing uses. Economics examines how to make choices and its efficiency in allocating these resources among economic agents in an economy. Households make their choices to maximize their utility under budget constraints, while the producers make choices to maximize their profits (minimize their cost) under the production technology constraint. This optimizing of economic agent behaviour will reveal the market price, which is determined by market equilibrium. Hence, this behaviour is capable of being captured in an economy wide general equilibrium framework.
The basic structure of the ORANI-SL model and the data required to construct a country-specific ORANI -SL model is presented in the figure below ( Figure 2). The model is split into a series of demanders listed below: 1. Domestic producers divided by I industries; 2. Investors divided by I industries; 3. A single representative household; 4. An aggregate foreign purchaser of exports; 5. Government demands; 6. Changes in inventories.  Any industry is capable of producing any commodity in the economy. Therefore, the MAKE matrix at the bottom of Figure 2 shows the values of the output of each commodity type produced by each industry. The summation of values in each column in the MAKE matrix must be equal to the total cost of the absorption matrix. By convention, the row totals associated with each commodity in the MAKE matrix must be equal to the total domestic production value of the respective commodity plus the direct and indirect usage of (domestic) margin commodities. The last matrix in Figure   2 represents the import tariff revenue (V0TAR).
Margin values (transport and trade) obtained from Sri Lanka's supply table were added   to the I-O table base values. The import duty matrix was created based on the importrelated data mentioned in Sri Lanka's supply table. In addition, the supply table also provides data on indirect taxes and subsidies relevant to each commodity. The data transformation steps conducted to transform I-O table data into the ORANI-SL structure are described below (Figure 2).

Investment Matrix
In the standard Sri Lankan I-O industry matrix to implement the ORANI-SL database represented in figure 2.
As there was no additional information found relating to the investment matrix, the fixed capital formation column values contained in the Sri Lankan I-O table were apportioned among the 127 industries based on the input demand and the assumption that larger usage inputs by industries have a greater investment weight.  Total  GFCF  A  X  TRA  IA  B  TRB  IB  C  TRC  IC  D  TRD  ID  E  TRE  IE  F  TRF  IF  Total  TCA TCB TCC TCD TCE TCF   Where; TCA is the total of column A TRA is the total of row A Therefore, the V2BAS value for cell X = (X/TRA) x IA.
For example, in the Sri Lankan I-O in the GEMPACK program. Therefore, all the ORANI-SL model data were saved in HAR files and can be viewed using a special program called ViewHAR.

Conclusion
This paper presents a new version of the ORANI-G model known as the ORANI-SL, which is a regional CGE model developed for Sri Lanka. The new model's production structure uses water as a factor of production in the agricultural sector and enables substitution between rain-fed and irrigated lands. To the best of our knowledge, this is the first CGE model that differentiates rain-fed and irrigated agricultural land in Sri Lanka. However, similar studies can be found in other countries. This paper provides a detailed description of the new production structure and the implementation of the new model's database.
The new model facilitates the assessment of the economy wide impacts of climate change and alternative adaptation strategies against climate change in Sri Lanka. Thus, policymakers can prioritize adaptation strategies to mitigate the impacts of climate change based on its economic impacts. Further, it enables the modelling of green and blue water usage in the agricultural sector in Sri Lanka. The author proposes to use this model to analyse the economic impacts of climate change in Sri Lanka, evaluate how enhanced irrigation capacity in the agricultural sector could sustain the national economy, and analyse the economy wide impacts of other adaptation strategies to mitigate climate change.
However, the ORANI-SL only considers water usage in the agricultural sector and neglects usage in households and industries. Therefore, future development of the ORANI-SL model could consider this limitation and incorporate home and industry water usage into the model.