The 2014 Mongolian Social Accounting Matrix

The construction of the Mongolian Social Accounting Matrix (SAM) for 2014 is described. The SAM included fifty-six sectors, seventy commodities, two types of production factors (capital and labor), three types of institutions (households, government and the rest of the world) along with capital accounts, three types of taxes (direct taxes, import duties and indirect taxes on commodities) and investment accounts (public investment, private investment and changes in inventories).


I. Introduction
The Mongolian Social Accounting Matrix (SAM) for 2014 was constructed with primary data sources that comprised Supply and Use Tables (SUTs), government budget data from the National Statistical Office (NSO), 1 and Balance of Payments (BoP) information from the International Monetary Fund (IMF). 2 The SAM is a square matrix with 210 columns and rows.
Its accounts consist of fifty-six sectors and seventy commodities, two production factors (capital and labor), three types of institutions (households, government and rest of the world) along with their capital accounts, three types of taxes (direct taxes, import duties, and indirect taxes on commodities) and investment accounts (public investment, private investment and changes in inventories). 3 Income and expenditure of all accounts of the SAM were equivalent (that is, the SAM was balanced). 4

II. Contributions to the SAM literature
The literature on SAMs covers areas such as construction, estimation, and application of SAMs in policy analysis. Some literature has also considered other methods based on the SAM framework such as multiplier analyses and empirical studies that employed SAMs. This paper concerns the methodology for building SAMs for policy analyses. Regarding this area, the works of J. I. Round should be noted. For example, Pyatt and Round (1977) reported their experience in constructing SAMs for three economies and highlighted problems encountered, solutions adopted, and lessons learned. Round (2003) reviewed progress in constructing SAMs so far and problems ahead. 5 In addition, a large number of papers describe the construction of SAMs for various economies. For example, Cicowiez and Lofgren (2018) provided a detailed technical description of their construction of a Mongolian SAM for 2015 which was the closest reference for our paper. 6 There were similarities between this SAM and our SAM in terms of methodology, and these are mentioned in Section 3.2; there are also notable differences, which are discussed in Appendix 2.

The "Proto" SAM
The first version of the SAM without adjustments or augmentations was called the "proto" SAM. Table 1 shows the "proto" macro SAM as a share (22.2 trillion Mongolian Tugrik or MNT) of nominal Gross Domestic Product (GDP) in 2014. 7 As Table 1 shows, private consumption contributes more than half of the GDP (57%) while current government spending is 13% of the GDP. Gross fixed-capital formation (GFCF) and inventory changes (VSTK) accounted jointly for 35% of the GDP. The values of both export and import were more than half of the GDP (52% and 57% respectively). The economy was highly capitalintensive (i.e., about 70% of value added was payments to capital owners). The share of value added in the GDP was 90.1%, and the remaining 9.9% came from indirect taxes on commodities (7.7%), import duties (1.6%), and net taxes on production (0.5%).  6 We neither used their 2015 SAM nor did we base our SAM on it because, before the authors constructed the 2015 SAM, we had already built a previous version of our SAM for 2014. On the other hand, we applied what we learned from their approach in revising our 2014 SAM. We chose 2014 as the base year because, when we started building our SAM, SUTs were only available for 2014. 7 This is a version of the SAM in which all industries and commodities are aggregated into one account.

Adjustments in the SAM
In the "proto" SAM, some sectors (trade and livestock, e.g.) were reported as highly capital-intensive. In reality, however, they were likely to have been labor-intensive, which is consistent with the structure of the Mongolian labor force. According to Labor Force Survey (LFS) data for 2013, 2014, and 2015, roughly 27% and 15% of workers were in the livestock and trade sectors respectively. 8 In that sense, the structure of value added in the "proto" SAM may be unrealistic.
In addition, to be compatible with Computable General Equilibrium (CGE) models, we needed to distribute domestic production of each commodity into domestic and export categories and exports at purchaser price into (trade and transport) margins and exports at basic price.
To make the SAM more realistic and suitable for CGE models, we made a few adjustments, which are explained below. In doing so, we followed the approach of Cicowiez and Lofgren (2018).

Separating "Mixed Income" from Gross Operating Surplus
In the 2014 SUTs, the value added in each sector was composed of employee compensation, consumption of fixed capital, and gross operating surplus. The first was considered labor payments while the latter two were considered capital payments. The structure of value added, meaning the factor intensity in all sectors, is given in Table 2 (Columns "labshr0" and "capshr0"). Because this structure may have been unrealistic, inaccurate analyses and wrong simulation results may have occurred. Capital intensity in some sectors may have been overestimated because gross operating surplus included "mixed income," which was likely to be a combination of the income of self-employed people, owners of small enterprises, and employers. Although production activities used labor, people in these employment categories tended to consider their income to be profit 8 The LFS is an annual survey organized by the NSO which is the main database for generating employment data at the country level. Raw data of this survey is accessible online to be used by researchers and other people.
Electronic copy available at: https://ssrn.com/abstract=3600954 from their activities and hence reported it as overall operating surplus rather than as labor income. A part of their income, however, which we called "mixed income," should be considered labor income.
To address this problem, we first calculated the number of salaried and non-salaried workers in each industry (shown in Table 2 under the columns "salaried" and "non-salaried") using LFS data from 2013, 2014, and 2015. In doing so, we merged LFS data to create a larger database and calculated the aforementioned numbers as a three-year average. The survey participants reported their sectors of activity, which were consistent with the sectors in the SUTs, and their employment status. We used the employment status of workers to define whether they were salaried or non-salaried.
There were six types of employment status in the LFS: • paid employee; • employer; • self-employed; • member of a producer cooperative; • employed in animal husbandry; • unpaid family worker.
Paid employees and unpaid family workers were considered "salaried" while employers, self-employed people, members of producer cooperatives and people employed in animal husbandry were considered "non-salaried." Because we also knew each worker's sector of activity from LFS data, we were able to obtain the number of workers of each type in each sector. The average wage for each industry was then calculated using the original employee-compensation values in the SUTs and the number of salaried workers (computed from the LFS). By multiplying the number of non-salaried workers by the average wage, we found "mixed income" values for each sector. Finally, these values were added to the original compensation of employees while being subtracted from the original capital payments. After this adjustment, labor and capital intensity in each sector were recalculated (also shown in Table 2 in Columns "labshr1" and "capshr1"). 9 As a result, capital intensity in some sectors decreased significantly and some sectors turned out to be more labor-intensive. Sectors relatively more affected by this adjustment were livestock, agriculture, apparel, leather, printing, repair and installation, trade, and land and water transport because all of these sectors included more "non-salaried" workers. Generally, the magnitude of the change depended upon the average wage and the number of non-salaried people.

Adjustments in Re-Exports
The exports of some commodities in the SUTs (crude oil, metal ores, other minerals; general-purpose machinery; special-purpose machinery; office, accounting and computing machinery; communication equipment; and transport equipment) were greater than domestic production. This is known as a re-export problem in CGE models and could exist in real life because stock accumulated in previous years could be exported even in the absence of new domestic production. CGE models do not take this feature into account, however, and we were forced to adjust exports exceeding domestic production to make the SAM consistent with CGE models.
To address this problem, we reduced the imports and exports of the aforementioned commodities simultaneously by the value exceeded. This adjustment was not possible for some commodities, however (crude oil, metal ores, and other minerals) because imports were too small or zero. Note, however, that exports were at purchaser price, meaning that they include trade and transport margins. On the other hand, domestic production was at basic price. Thus, we needed to calculate exports at basic price, excluding margins. To do so, we used a small model that distributed margins into domestic transactions and exports (see below). Once export margins were subtracted from purchaser-price exports, the reexport was solved for two of the three commodities (crude oil and metal ores). 10 Exports at basic price still exceeded domestic production at basic price for one commodity, however: other minerals. For this commodity, we reduced exports at purchaser price by the exceeded value and simultaneously increased inventory changes by the same value. Inventory changes for this commodity were initially negative and then that negative number shrank-in other words, we eliminated some changes in inventories because we assumed that exports at purchaser price included the value of exports from accumulated stock. CGE models, conversely, specify exports and stock variations separately and do not take this feature into account; they assume that commodities are exported solely from current production and not from accumulated stock.
According to the SUTs, forestry products were not exported. A value for export tax was included, however, for that commodity. To eliminate this inconsistency, we included this value in the export tax of livestock products while transferring the same amount from indirect taxes on livestock products to indirect taxes on forestry products.

Distributing margins into domestic transactions and exports
In the SUTs, trade and transport margins appear by commodity. There is, however, no information on how the margins were distributed among different transactions (domestic sales, imports, and exports). To distribute the margins among these transactions, we employed the transaction-cost model of Cicowiez and Lofgren (2018). In doing so, we first aggregated trade and transport margins to find the total margin for each commodity. The transaction-cost model then split the total margin into different transactions. 11 From the results we separated trade and transport margins for each commodity using the original structures. Imports and domestic sales were considered jointly as domestic transactions.

IV.
New Accounts in the SAM

Exported Coal
We separated the existing coal sector/commodity in the SUTs into domestic coal ("coal") and exported coal ("excoal"). The exported coal sector may represent coking-coal reserves such as Tavan Tolgoi and Nariin Sukhait. These are the largest coking-coal reserves in Mongolia and most exported coal is extracted from these reserves. On the other hand, the domestic-coal sector may represent such thermal coal reserves as Baganuur and Shivee-Ovoo which jointly supply most domestically-consumed coal.
11 For a detailed explanation of how the transaction-cost model works, see Appendix 4.
The cost structure of these two sectors was assumed to be identical because we had no data to permit disaggregation. Output of the "excoal" sector was equal to the value of the original coal sector's exports. On the other hand, output of "coal" sector was equal to the original coal sector's domestic output. Domestic demand for the original coal was then considered demand for the "coal" commodity.

Railway
We created a new sector/commodity for railway. 12 The railway sector was extracted from the original transport sector and was assumed to have the same cost structure as the whole railway system in the economy. 13 Railway service was produced exclusively by the railway sector and represents that part of the service used only for coal export as a transport margin.
Thus, we had three margin commodities: trade, railway, and other transport.
The original value of the transport margin associated with coal export was about 135 billion MNT. We assumed that 10% of this value (13.5 billion MNT) could be allocated to the railway, implying that the total demand for the railway together with the total output of the railway sector was 13.5 billion MNT.

Separation of Coke-Oven Products and Refined Petroleum Products (Fuel)
In the SUTs, coke-oven products and refined petroleum products were aggregated into one account. Coke-oven products include peat, coke, semicoke, and lignite which are domestically produced and exported. On the other hand, according to the Mineral Resource and Petroleum Authority of Mongolia, refined petroleum products were exclusively imported (i.e., there was no domestic production). Hence, we distinguished these commodities from one another by dividing the original commodity in the SUTs into two: "fuel" and "coke." Refined petroleum products ("fuel") contributed 92.3% of domestic demand for the original commodity while the remaining 7.7% was produced domestically. We used this structure to distribute indirect taxes on the commodity, domestic trade, and transport margins between the new commodities. The structure of sectors producing "coke" were left unchanged and the original values of outputs were considered to derive exclusively from the output of "coke."

Capital Accounts
To construct a more realistic financing mechanism of public and private investment and include financial flows between institutions in the SAM, we added a capital account for each institution. This resulted in three new accounts: "cap-h", "cap-gov" and "cap-row." Here, we once again followed the approach of Cicowiez and Lofgren (2018). The significance of these accounts is that they allowed us to include details on the sources of private investment, government borrowing and foreign direct investment. Such details are useful for policy analyses in CGE models.
Each institution's savings (calculated as the difference between income and expenditures) went to the corresponding capital account (Table 3): • Household/firm savings -SH; • Government savings -SG; • Net foreign savings -SROW.
In addition, the following financial flows were included in the SAM (Table 3): • Net domestic borrowing of the government -NDFG; • Net foreign borrowing of the government -NFFG; • Net foreign borrowing of the private sector (households) -NFFH; • Foreign direct investment -FDI. The SAM included accounts for private and public investment expenditures (gross fixed capital formation or GFCF) and changes in inventories: Electronic copy available at: https://ssrn.com/abstract=3600954 • Private fixed capital investment expenditure -GFCF_PRI; • Public fixed capital investment expenditure -GFCF_PUB; • Changes in inventories -VSTK.
The first two were calculated in Equations 1, 2, and 3 below while VSTK was provided in the SUTs. Equation 1 implies that private fixed capital formation (GFCF_PRI) was financed by non-FDI (N_FDI) sources and foreign direct investment (FDI). Non-FDI sources were household savings (SH) and net foreign borrowing of households (NFFH) in excess of household's net lending to the government (NDFG) and inventory changes (VSTK) as in Equation 2. Public fixed capital formation (GFCF_PUB) was financed by government savings, net domestic and foreign borrowing of the government as in Equation 3.
The value of NFFG was taken from IMF Article IV 2017 (1.6 trillion MNT

V. Aggregation of the SAM
After the adjustments and augmentations, the SAM was referred to as "revised." The revised SAM was then aggregated in terms of sectors and commodities for the purposes of facilitating interpretation. Specifically, seventy commodities and fifty-six sectors were aggregated into twenty-four commodities and twenty-four sectors. The mapping between original and aggregated sets of sectors/commodities used in this paper was similar to that of Electronic copy available at: https://ssrn.com/abstract=3600954 the NSO. Sectors/commodities more important for the economy, such as agricultural and mining sectors/commodities were largely left disaggregated while relatively small sectors/commodities, especially manufacturing, were aggregated. For instance, forestry and fishery were aggregated into "agriculture" while wood, paper, metal industries, machinery, equipment, and vehicles were aggregated as "other manufacturing sectors/commodities." The new sectors/commodities (exported coal, fuel and railway) were not aggregated. 14 The aggregated SAM is a square matrix with eighty-six columns and rows. Accounts in the aggregated SAM are shown in Table 4.  respectively. In comparison with the "proto" SAM, export and import shares were slightly decreased as the result of the adjustment of re-exports. Capital intensity decreased due to the "mixed income" adjustment. As a result, the economy was equally capital-intensive and in labor-intensive in the revised SAM.
14 For the mapping of sectors/commodities in the aggregated and the detailed SAM, see Appendices 5 and 6.

Production Structure
Livestock and trade sectors contributed most to labor income while metal ores sector contributed most to capital income. The crude oil, food, metal ores, and coke and chemicals sectors were highly capital-intensive while the public-administration, education, health, railway and livestock sectors were most labor-intensive (Table 6).  Table 7 shows trade structure. Metal ores represented more than half of total exports.

Trade Structure
Fuel, coke-oven products, chemicals, and other manufacturing products contributed most of the imports. Crude oil and metal ores were almost exclusively exported. Most manufacturing commodities were imported. In particular, fuel was not produced domestically and was exclusively imported. On the other hand, some commodities, including trade, railway, and public administration were not traded internationally.  Table 8 shows the demand structure for each commodity. Most of food, textiles, and accommodation and information services were consumed by households whereas public administration, education, and health were mostly consumed by the government. Electricity and mining commodities were mainly used as intermediate inputs for production. Trade and railway were a 100% margin commodity while 19% of other transport services were used as a margin. Construction services were mainly used for investment.  Labor and capital income of households included factor payments from abroad net of factor payments to foreigners. Transfers from/to ROW were taken from the BoP while consumption spending of households was taken from the SUTs. Household savings was calculated as the difference between income and expenditure.

Government Activities
The government received the majority of its revenue from households (firms) as direct taxes (47.3%) and transfers (16.7%). Commodity taxes made up 27.4% of its revenue. Other sources of income were relatively small. Almost half the government budget was spent on purchasing goods and services while 36.7% went to households as government transfers.
Government savings, used to finance capital expenditures, were 14.4% of the total budget. Import duties, export taxes, net taxes on products, net taxes on production. and public consumption were taken from the SUTs and transfers from/to ROW were taken from the BoP.
Government savings reflected actual budget data from the NSO, calculated as government revenue minus current expenditures. Direct tax revenue was calculated as a residual by subtracting other types of taxes from total tax revenue in the actual budget data. Similarly, transfers from households was calculated as a residual to replicate total government revenue in the actual budget data. Transfers to households was also calculated as a residual to balance the government account.  Exports and imports took their values from the SUTs while factor payments, factor income, and transfers from/to Mongolian domestic institutions were taken from the BoP. Finally, savings of the ROW was computed as the difference between total income and total expenditures.

Investment/Savings Structure
More than half of total investment was financed by private savings. Rest of the world and the government contributed nearly 33% and 12% of total investment budget (source) respectively. The majority of the investment budget was dedicated to financing gross fixed capital formation (that is, private and public investment at 38.5% and 42.6% respectively). A relatively small fraction (18.9%) was spent on inventory changes.

VII. Conclusions
This paper has presented the building of the Mongolian SAM for 2014 and some analysis done on the basis of the SAM. The SAM was constructed based on the Mongolian National Accounting System database and integrated all transactions in the economy for 2014 to exhibit a general and complete picture of the economy at one glance. This SAM has become the main database for Mongolian CGE models.
Appendix 1. Balance of the SAM Generally, we made sure that each account in our SAM was balanced at every step of the construction. Technically, we checked the balance of the SAM using either GAMS software or Microsoft Excel after any change had been made.
We took some values in the SAM from the primary data sources mentioned earlier. On the other hand, we calculated other values in the SAM on the basis of several assumptions. To balance the labor account, for example, household labor income was calculated as compensation of employees in the SUTs plus net labor income from abroad. The same applied to the capital account. For household and ROW accounts, the savings of each institution was calculated to balance the corresponding account while government transfers to households balanced the government account. Industry and commodity accounts were already balanced in the "proto" SAM because their values were taken from the balanced SUTs. The same was true for tax accounts.
Because the "proto" SAM was balanced, all adjustments and augmentations were made to maintain that balance. Household income from each factor absorbed the difference in factor accounts caused by the adjustment of mixed income, for example, which changed the factorpayment structure from industries. Total factor payment from individual industries were unchanged by the adjustment, however. In addition, reducing export and import by the same amount in the adjustment of re-exports ensured that the balance of the ROW account was not disturbed. When splitting total output into domestic supply and export and total margin into domestic and export margin, the transaction-cost model ensured that the disaggregated values equaled the original total values. It also ensured that exports at basic price, along with export margin and tax, equaled exports at purchaser price. When we split an existing sector or commodity, we kept the original aggregate values so that other parts of the SAM were unaffected. Moreover, we imposed assumptions on some components of the SAM to maintain the overall balance. When splitting the coal and transportation sectors, for example, we based our split on the income side of the sectors and then applied the resulting structure to the expenditure side.

and the SAM Discussed in this Paper
As a result of the following differences, some values in the two SAMs differ: •  The TCM took the values of the following from the SUTs as exogenous: • Domestic production at basic price; • Exports at purchaser price; • Imports at basic price; • Aggregate transaction costs by commodity.
The TCM generated values for the following endogenous variables: • Exports at basic price; • Domestic supply at basic price; • Share of transaction costs in total raw trade values by commodity. The model assumed that all transaction costs of a commodity were the same constant share of their corresponding raw trade values (that is, there was only one transaction cost share per commodity). • Eventually, transaction costs related to exports and imports and domestic transaction costs were computed using the shares and raw trade values. Before running the model, we only had domestic production at basic price. On the other hand, with the TCM, we knew the composition of domestic production for each commodity (meaning that we knew how much was exported and how much was sold domestically).
Without the TCM, we knew only how much was paid by foreigners for exports. In contrast, the TCM provided the distribution of these values between producers and transaction costs. Meat, fish, fruit, vegetables, oils, fats, dairy products, egg products, grain mill products, starches and starch products, other food products, beverages, and tobacco products Textiles Yarn, thread, woven and tufted textiles, textile articles other than apparel, knitted or crocheted fabrics, clothing, leather, leather products and footwear

Manufacturing
Products of wood, cork, straw and plaiting materials; Pulp, paper and paper products; printed matter and related articles; Furniture, other transportable goods; Basic metals; Fabricated metal products, except machinery and equipment; General-purpose machinery; Special-purpose machinery; Office, accounting and computing machinery; Electrical machinery and apparatus; Radio, television and communication equipment and apparatus; Medical appliances, precision and optical instruments, watches and clocks; Transport equipment. Fuel Refined petroleum products

Coke and chemicals
Coke-oven products, nuclear fuel, basic chemicals, other chemical products, man-made fibers, rubber and plastics products, glass and glass products, and other non-metal products Construction Construction services Trade Wholesale and retail trade services Accommodation Accommodation, food and beverage services Transportation Passenger transport services, freight transport services other than railway, rental services of transport vehicles with operators and supporting transport services Railway Railway transport services for coal export Financial activities Financial and related services