A Macro-Economic Model for a Small, Open Economy with Limited Data

Total Page:16

File Type:pdf, Size:1020Kb

A Macro-Economic Model for a Small, Open Economy with Limited Data

A macro-economic model for a small, open economy with limited data

The case of Aruba1

E.E. Matos-Pereira2 and G.G. Croes3

Centrale Bank van Aruba

August 2011

ABSTRACT

This paper presents the work in progress a macro-economic model currently used by the Centrale Bank van Aruba to estimate and/or forecast the GDP based on a bottom-up approach from both the demand and production side. This implies that the GDP is estimated/forecasted via, on the one hand, the forecasts of the expenditure components of the GDP and, on the other hand, the projections of the gross value added across the different economic industries. This framework is a tool for examining the plausibility of results and detecting inconsistencies. Despite the problems of limited data availability in Aruba, this model enables forecasting on a quarterly and annual basis.

1 Prepared for the Bank of Jamaica Research Conference in September 2011. The views expressed here are solely those of the authors, and not necessarily those of the Centrale Bank van Aruba. 2 Deputy Manager, Research Department, Centrale Bank van Aruba, J.E. Irausquin Boulevard 8, Oranjestad, Aruba. E-mail: [email protected] 3 Economist, Research Department, Centrale Bank van Aruba. E-mail: [email protected] 2 1 Introduction This paper presents the GDP forecasting framework of the macro-economic model currently used at the Central Bank of Aruba (CBA). In this model, the GDP is estimated and/or forecast based on a bottom-up approach from both the demand and production side. This implies that the GDP is estimated/forecasted via, on the one hand, the forecasts of the expenditure components of the GDP and, on the other hand, the projections of the gross value added across the different economic industries. This framework is a tool for examining the plausibility of results and detecting inconsistencies. Despite the problems of limited data availability in Aruba, this model enables forecasting on a quarterly and annual basis two years into the future. Problems of limited data availability are likely to be common in many small-island economies, including Aruba. Because of the relatively short data set, performing some statistical exercises may be difficult. It is therefore imperative that the user of the model has comprehensive knowledge of the available data and uses economic judgment to determine if the results presented by the model make any economic sense. In some cases it may be better to use simple techniques such as moving averages, which do not require large time- series. In addition, with relatively short and/or limited data sets, sophisticated techniques probably will not work, so when possible it may be best to keep it simple. Besides the problem of short data set availability, only historical annual GDP figures in nominal terms are available, while the aim of the model is to produce forecasts in both nominal and real terms as well as both annual and quarterly forecasts. Therefore, price deflators for the overall GDP and its expenditure components are constructed, while a proportional technique is applied to calculate quarterly GDP series. However, price deflators for the economic sectors are neither available nor estimated in the model. The end-result is a model which has its short- comings, but which enables the user to gauge the developments in relevant economic sectors, despite the mentioned data limitations. Though preliminary work has been undertaken to integrate in the model modules enabling projections of, among others, money supply, net foreign assets, current account balance, fiscal balance and public debt, further work needs to be done in the near future to complete these efforts. As of yet, the latter variables are estimated exogenously. Furthermore, the GDP forecasting framework could be further refined to estimate price deflators for the economic sectors. The structure of this paper is as follows: section 2 of the paper discusses the methodology, followed by section 3 covering the estimation of price deflators. In section 4, the estimating process of the expenditure approach is introduced, while section 5 presents

1 the production method. In section 6, the balancing process of both approaches is discussed briefly, whereas section 7 deals with the quarterly calculation of GDP and its components. Next, the concluding remarks are presented.

2 Methodology The model applies a bottom-up methodology of forecasting GDP from both the demand side (i.e, the expenditure approach) and the production side (i.e., the production approach). The foundation of this model is therefore the national accounting identity which entails that the GDP estimated by the production approach is equal to the GDP calculated by the expenditure method. The national accounting identity is valid for the estimation of GDP in both nominal and real terms. Using the mentioned approaches, forecasts are made for the individual GDP components, i.e., consumption, investment, exports and imports (expenditure or demand-side methodology) and the gross value added of the individual economic sectors (production approach)4. The GDP according to the expenditure approach can be expressed as the following identity:

(1)

where Y is the nominal GDP, Cp is private consumption, Cg is government consumption, Ip is private investment, Ig is government investment, Xtour is tourism exports (i.e., tourism expenditures), Xoth is other exports (i.e., nontourism exports) and M is imports. Please note that all variables are in nominal terms. The production approach of estimating GDP is based on the following identity:

(2)

where GVAi is the nominal gross value added of economic sector i. The national accounts of Aruba distinguish sixteen economic activities (k=16), including one category for the adjustment for financial intermediation services indirectly measured (FISIM).

4 GDP is equal to the sum of the gross value added of all producer units plus taxes, less subsidies, on products. 2 From a theoretical point of view, the advantage of forecasting real GDP over nominal GDP is that this allows for more meaningful comparison of the structure, workings, and performance of the economy over time. Identity (1) can be transformed into real terms by dividing the variables on both sides of the equation by their respective price deflator:

(3)

where Py is the GDP deflator, Pcp is the private consumption price deflator, Pcg is the public consumption price deflator, Pip is the private investment price deflator, Pig is the public investment deflator, Pxtour is the price deflator for tourism exports, Pxoth is the nontourism exports price deflator, and Pm is the import price deflator. Identity (2) can also expressed in real terms:

(4)

where Ps is the price deflator of economic sector i. i

Data availability is a major problem for the construction of this model. National accounts data are available only from 1995 to 2009. Moreover, only nominal values of the expenditure components of GDP and nominal gross value added are estimated in the national accounts of Aruba, because the Consumer Price Index (CPI) is the only available price index. The CPI is not representative for the price changes in the individual GDP components with the exception of private consumption and should rather not be used as deflator for the other GDP components. However, one prerequisite of this model is the possibility of comparing GDP figures over time, requiring the calculation of the real values of all GDP components. Consequently, the first step is to calculate price deflators for public consumption, investments, imports, tourism exports, other exports and an overall GDP deflator. No sectoral price indices are calculated, but instead a second best solution is created by using the deflators of the GDP expenditure components to calculate the real gross value added of a number of economic sectors. In a second step, the behavioral equations are estimated for the expenditure components of the GDP (demand side). Next,

3 the behavioral equations are calculated for the gross value added of the domestic economic activities. Subsequently, an iterative process is used to ensure consistency between the various elements of the forecasts on an annual basis, including the GDP estimates of both estimation methods. Finally, quarterly estimates for GDP and its expenditure components are calculated.

4 3 Estimating price deflators The GDP deflator measures changes in the prices of domestically produced products for all types of final use. Two methods can be used for estimating and forecasting the GDP deflator, i.e., the aggregated approach and the disaggregated approach (IMF, 2007). The aggregated methodology involves examining the historical relationship between the CPI inflation rate and the inflation rate measured by the GDP deflator, and assuming a continuation of this relationship to forecast the GDP deflator. As there is no historical GDP deflator available for Aruba, this method is not an option. According to the disaggregated approach, the GDP deflator can be thought of as a weighted average of the prices of the components of GDP. This method is utilized to estimate the GDP deflator and other relevant price deflators for the past years. Moreover, the GDP deflator can be forecasted by individually forecasting the prices and volumes of the components of the GDP, summing the relevant components to obtain the nominal GDP and the real GDP, and then computing the GDP deflator as the ratio of nominal GDP to real GDP. The GDP deflator thus equals:

(5)

where y is the real GDP.

This model utilizes the disaggregated approach for only the demand side in order to forecast prices for the different expenditure components of the GDP. The constructed deflators of the GDP expenditure components are then used to calculate the real gross value added of a number of economic sectors. However, since this method is not applied to all economic sectors, it is not possible to derive a real GDP forecast from the production side. Therefore, the nominal GDP figures for the various sectors of the production side are added up and the total is then deflated using the overall GDP deflator derived from the disaggregated approach on the demand side. In the absence of historical demand side deflators and data needed to build these, this model produces proxy estimates for the price movements of the expenditure components

5 of GDP. The details of the calculations of the deflators for each individual GDP component are presented below.

Private consumption deflator The CPI of Aruba (indicator for prices of goods and services) is the private consumption deflator in this model. (6) where t indexes time.

Public consumption deflator Public consumption consists of wage-related expenses and purchases of goods and services by the government. Therefore, the price deflator is calculated as the weighted average of the public sector wage index (WI) and the CPI of Aruba. The public sector wage index is derived from the average wage costs per government employee on a cash-adjusted basis5.

(7)

where w1 is the share of wages in total government expenditures.

Investment deflator Due to data limitations, the model applies the same deflator to deflate the investment of both the private and the public sectors. Investment demand could be met by domestic supply or foreign supply of goods and services (i.e., imports). Consequently, the investment prices are calculated using a weighted average of the change in the domestic prices as measured by the CPI and the change in the non-oil import price ( . The weights are determined by the portion of each type of supply in total investments. Steeg (2009) provides an estimate for these shares for the year 1999. Unfortunately, more recent estimates for these shares are not available, which are therefore assumed constant.

(8)

5 The wage costs on a cash-adjusted basis are equal to the wage costs paid during a period plus the increase in the wage-related payment arrears. 6 where w2 is the share of foreign supply of goods and services in total investment expenditures. The historical change of the non-oil related import price is obtained using a methodology applied by Ridderstaat (2004) in which the price change for each of the source countries is weighted by their share in Aruba’s total merchandise imports. The price change for each country is measured as the CPI of that country or an export price index. The future non-oil import price changes are not projected in the model, but are obtained from the Department of Economic Affairs in Aruba, which uses a methodology similar to that of Ridderstaat (2004).

(9)

where is price of country j and wj denotes the share of country j in total merchandise imports.

Import deflator Total imports are divided into oil and non-oil imports. The import deflator is calculated using a weighted average of the change in non-oil import prices ( and the change in the oil import prices ( . The weights of the components are determined by assessing the share of each one in total imports.

(10)

Tourism exports deflator Tourism exports comprise accommodation services, food and beverage services, the consumption of other domestic goods and services, and the use of foreign products sold to visitors in Aruba. Tourism prices are, therefore, estimated as a weighted average of the changes in respectively, hotel prices ( , the restaurant price index ( (a sub component of the CPI), the CPI, and the non-oil import price. The weights are based on the breakdown of tourism exports into these expenditure categories (Steeg, 2009).

7 (11)

where w4+w5+w6+w7=1

8 Non tourism exports deflator The non-tourism exports consist in large part of different kinds of services, such as the maintenance fees paid by time share owners and management fees charged by resident companies to their foreign shareholders. Since no price index and no clear indicators of the price development for these exports exist, the price changes of non-tourism exports are assumed equal to the change in the Aruban CPI excluding the energy components, i.e., a measure of core inflation ( .

(12)

4 Estimating the behavioral equations of the GDP components: annual figures In this section, the selection process of the behavioral equation of each GDP component is discussed briefly. Each behavioral equation is fitted by Ordinary Least Squares using the maximum amount of data available for the variables in a given equation. This exercise results in behavioral equations for only the private consumption and non-tourism exports. For the other expenditure components, simple forecasting equations are used based on the presumption of a statistical relationship between the relevant expenditure and another economic aggregate.

Private consumption In theory, private consumption is a function of expected household disposable income, net wealth of households, after-tax real rate of interest and other variables. In this model, GDP minus tax revenues (expressed as Ydisp) is used as a measure for expected household disposable income, while real (narrowly defined) money supply (i.e., M 1) also is included in the private consumption equation. Please refer to Annex A for the selected equation.

Public consumption Public consumption is an exogenous variable in this model and is based on the government budget. When the budget is not available, public consumption is assumed constant in real terms.

Private Investment

9 Private investment is theoretically a function of real GDP, real interest rate, inflation, and other factors (availability of finance, expectations, etc.). Various regressions using variables such as real GDP, the real interest rate, and the CPI as independent variables do not render satisfactory results. Therefore, the forecast for real private investment is instead based on the expectation of the change in the gross value added of the construction sector (GVA Con) calculated in the production method.6 However, further research is necessary for the calculation of this component. The function for the private investment can be expressed as:

Ip = (GVA ), where GVACon refers to the gross value added of the construction sector. Con

Public Investment The forecast for public investment is based on the government budget for the first year to be forecasted. For the second year, public investment is calculated based on a 5 year-moving average of realized public investments and budgeted development spending by the Fondo Desaroyo Aruba (FDA).7

Imports According to conventional wisdom, imports are determined by real income (income elasticity of imports), the relative price of imported goods relative to domestically produced substitutes (may not be so relevant for Aruba as domestic substitutes are virtually non- existent) and the distribution of imports over the different trading partners, the credit growth of the banking sector, import tariffs and other trade barriers, and natural disasters that disrupt trade, and improvement in infrastructure that facilitate trade. Some of these indicators are easier to translate into a model than others. For this model, various regressions are done to calculate imports. However, regressions with independent variables such as tourism receipts, wage tax, loans by commercial banks and population do not render good results. Better results are obtained by a regression of imports on real household consumption, real consumer credit and loans issued by commercial banks, and a regression on real GDP as the dependent variable. However, the

6 For the estimation of the gross value added of the construction sector, please refer to Annex C. 7 Fondo Desaroyo Aruba (FDA) is a development fund funded by the Aruban and Dutch governments to finance several investment projects in Aruba. 10 data show that the ratio between consumption and investment on one hand and imports on the other is relatively stable over a longer period of time. Since the use of this ratio increases the model’s ease of use and is more practical, this ratio is used to project imports. The function for imports is as follows:

= ( , , , )

Exports Exports are a function of a country’s productive capacity (in Aruba’s case mainly the number of available hotel rooms), the ratio of export prices to the cost of domestic production, external demand for the domestic goods and services, and excess domestic demand that results in rising domestic prices. Due to the importance of the tourism sector for the Aruban economy, exports are broken down into tourism exports and non-tourism exports in this model. Tourism expenditures are largely impacted by developments in the United States, Aruba’s major tourism market. While a regression on the GDP of the U.S.A. gives a good fit, a Hodrick-Prescott filter is used to forecast the tourism receipts on a cash basis independently from the model. Because tourism exports and tourism receipts are in essence the same thing, the growth in tourism receipts estimated in this way is then applied to tourism exports in nominal terms, which is then deflated using the tourism deflator in the model to get tourism exports in real terms.

Xtour = (Tourrec), where Tourrec represents nominal tourism receipts.

Based on the theory, non-tourism exports are calculated via a behavorial equation with the independent variables real GDP of the U.S.A. and real exchange rate of the Aruban florin against the U.S. dollar. A dummy is incorporated in the regression to account for the significant decline in exports resulting from diminishing activities of the free-zone sector as of the year 2000. Please refer to Annex A for the specification of the equation.

11 5 Estimating the behavioral equations for the production approach The National Accounts of Aruba distinguish 16 economic activities, including one category for the adjustment for financial intermediation services indirectly measured (FISIM). Please note that the data on gross value added of these activities for the period of 1995-2006 are used in this model. In the meantime, data for 2007-2009 have become available, implying that the behavioral equations for the production approach should be revised in the near future. Note, also, that in the case of the production approach only annual forecast are estimated. The first step comprises the selection of potential variables which could possibly determine the development in the real gross value added of the respective economic activities (see Annex B for a list of potential variables). Preferably, the relationship between the variables and the gross value added of the respective industries should be based on economic theory. One major problem is that data on real gross value added and output deflators are not estimated by the Central Bureau of Statistics (CBS). As a second best solution, the deflators of the GDP expenditure components (consumption, investment, exports, and imports) are used to calculate real gross value added of several economic activities. The next step entails the application of linear regression analysis using the potential variables to determine the potential behavioral equations. After that, the behavioral equation that is consistent with economic theory, taking into account statistical properties is selected. One general assumption is that, given the forecasting framework covering two calendar years, the output of the domestic firms is determined by changes in aggregate demand (refer to Annex C for the selected equations). Next, the real gross value added of the respective economic activities is estimated/forecasted and the nominal gross value added for the different economic activities is then derived. The gross value added of four economic activities are not estimated using above-mentioned steps, because of the lack of useful indicators or explanatory variables and the relatively small size of these activities. Instead, the estimations of the gross value added of the industries of “agriculture, hunting, forestry, fishing, mining and quarrying” and “manufacturing” are based on a 5-year moving average growth rate. The forecast of the industry of “public administration; Compulsory social security; Education” is based on a 3-year moving average ratio of gross value added to personnel expenses on a cash-adjusted basis. The forecast of the “health and social work” industry is based on a 3-year moving average growth rate. The period over which the moving

12 average is calculated is determined based on which length has the smaller variation between the individual years. Moreover, the net taxes on products are estimated, based on the average tax-to-GDP ratios in the past of varying lengths depending on the specific type of indirect tax and the initial GDP forecast for that year of the expenditure approach. The last step of this method involves the calculation of the total nominal GDP. Real GDP is not calculated, because for four economic sectors only nominal gross value added are forecasted.

6 Balancing the demand and production approaches After estimating a GDP through both the demand and the production approaches, these two figures have to be reconciled. This process includes carefully reviewing each component of both approaches, and determining where adjustments can be made. This process ensures the consistency between the two approaches of estimating GDP. When all GDP components and gross value added of the sectors have been adjusted, but there remains a relatively small statistical discrepancy (usually less than 1 percentage point of GDP) between the two approaches, the statistical discrepancy will be proportionally distributed either over the economic sectors (if applied on the production side) or over the demand side GDP components (if applied on the demand side). Whether the adjustment is made to the production side or the demand side is determined through the judgment of the users of the model. Please refer to Annex D for the behavioral equations and the identities of both approaches in the model.

7 Quarterly calculations of GDP and its expenditure components The system of National Accounts in Aruba does not produce quarterly figures for the GDP or its components. In this model, the annual GDP estimates are the basis for the calculation of quarterly GDP. In brief, the first step in deriving quarterly forecasts is the selection of variables for which quarterly data are available and that can be used as proxies for the GDP components. Next, the annual estimates of the expenditure components of GDP are spread over the four quarters in accordance with the quarterly distribution of these proxies. Depending on the selected variables, either the nominal or the real value of the expenditure component is estimated for each quarter. The third step involves the use of price deflators to transform either the nominal or real value calculated in the previous step into the value (nominal or real) not yet estimated. As only annual price deflators are calculated in this model, it is assumed that the quarterly price deflators are equal to the annual deflators.

13 Private consumption For the quarterly forecasts of private consumption, the average distribution of the real tourism receipts over the quarters for the period 1999-2006 is used as a proxy for the distribution of real private consumption during those years. While far from ideal, tourism plays an important role in the Aruban economy, and the income and spending of people living there can thus be expected to have some correlation with consumption by households. In 2007 the government of Aruba introduced a turn-over tax (BBO), levied on the sale of nearly all goods and services. The BBO can serve as a very good indicator of consumption. Consequently, from the year 2007 onward, the quarterly distribution of the real BBO revenues is used to derive the quarterly distribution of private consumption. For the years to be forecasted, a moving average of this distribution is used to determine the distribution of the annual forecasts of real private consumption over the quarters.

Public consumption To determine the quarterly distribution of government consumption over the year to be forecasted, a 4-year moving average quarterly distribution is used. After calculating the nominal consumption for each quarter, the deflators are used to obtain data in real terms.

Private investment Private investment is another GDP component for which little or no quarterly data is available, meaning that alternative indicators needs to be found. Private investment in Aruba is largely related to construction activities. For that reason, it makes sense to look for an indicator related to construction to determine the quarterly distribution of private investment. An indicator that seems to be closely aligned with activities in the construction sector is the volume of imported cement, since cement has to be used fairly soon after being imported. Therefore, the average quarterly distribution of cement imports is used to determine the quarterly distribution of real private investment, despite the relatively large variation in the distribution over the years8. In an attempt to compensate for this variation, a long period of available data of cement imports, i.e. an 11-year moving average of the distribution, is used in the model. In the future an 11-year moving average can be used, or a

8 We also looked at the quarterly distribution of the value of building permits granted, but this distribution was very unstable (more than that of cement imports). 14 year of data can be added to determine the average distribution as new data points become available.

Public investment The quarterly distribution of public consumption is determined by applying an 11-year moving average quarterly distribution of the sum of development fund spending and government investment to the annual forecasts.

Imports Given that the balance of payments in Aruba is produced on a quarterly basis, quarterly imports data are available. To calculate the average distribution over the quarters, a 10-year moving average of the quarterly distribution of total imports f.o.b. registered in the balance of payments is applied to the annual forecast of real imports.

15 Tourism exports The distribution of tourism exports over the quarters is obtained by applying a 10-year moving average of the quarterly distribution of real tourism receipts as registered in the balance of payments to the annual projection.

Non-tourism exports The distribution over the quarters is based on a 3-year moving average of the quarterly distribution of the real disposable income of the United States.

8 Concluding remarks The model presented in this paper is a work in progress which provides a tool for policy- makers to gauge the overall state of the economy and the developments in specific economic sectors. The major advantage of the GDP forecasting framework in this model is that it generates forecasts using relatively little data and that it provides a mode for examining plausibility of results and detecting inconsistencies, while it also enables more comprehensive analysis of the sources of particular outcome of GDP. However, the model suffers from serious data limitations. The time series are generally quite short and a coherent set for annual data is available only for the period 1995-2009, which is the period of the availability of National Accounts statistics. Constant price values for GDP by sector are neither available not constructed in the model, and deflators, not estimated for this purpose, are used to convert current prices into constant price values and vice versa. Further research is needed to refine this model, including endogenizing variables that are currently exogenous. Preliminary work already undertaken to estimate modules enabling projections of, among others, money supply, net foreign assets, current account balance, fiscal balance and public debt, should be completed in the near future. Furthermore, the GDP forecasting framework could be further refined to estimate price deflators for the economic sectors. Finally, the forecasting performance of this model should be evaluated in the near term.

16 Bibliography Hahn, E., & Skudelny, F. (2008). Early estimates of Euro Area real GDP growth: A bottom up approach from the production side. ECB Working Paper Series No 975. IMF Institute. (2007). Financial Programming and Policies. IMF Institute. Ridderstaat, J. (2004). The imported and domestic inflation in Aruba. Working paper, Centrale Bank van Aruba. Roland, G., Greenidge, K., Belgrave, A., & Moore, W. (2003). A macroeconomic forecasting model of Aruba. Steeg, A. (2009). Accounting for Tourism; The Tourism Satellite Account (TSA) in Perspective. The Hague: Statistics Netherlands.

17 Annex A 1. Real private consumption

Dependent Variable: LOG(Cp/Pcp) Method: Least Squares Sample: 1996 2008 Included observations: 13

Coefficient Std. Error t-Statistic Prob.

C 1.506314 1.422866 1.058648 0.3147

LOG((Y-T)/Py) 0.488042 0.215173 2.268136 0.0467

LOG(M1/Pcp) 0.318829 0.052782 6.040467 0.0001

R-squared 0.931947 Mean dependent var 7.313869 Adjusted R-squared 0.918336 S.D. dependent var 0.099292 S.E. of regression 0.028375 Akaike info criterion -4.087459 Sum squared resid 0.008051 Schwarz criterion -3.957086 Log likelihood 29.56848 Hannan-Quinn criter. -4.114257 F-statistic 68.47162 Durbin-Watson stat 1.437523 Prob(F-statistic) 0.000001

2. Non-tourism exports

Dependent Variable: LOG(Xoth/Pxoth) Method: Least Squares Sample: 1996 2008 Included observations: 13

Coefficient Std. Error t-Statistic Prob.

C 1.626039 4.322421 0.376187 0.7155

LOG(y ) usa 0.466330 0.446618 1.044136 0.3237

LOG(exchrus$) 1.864790 0.739967 2.520098 0.0328 DUM_FZ -0.197824 0.042160 -4.692170 0.0011

R-squared 0.932722 Mean dependent var 6.686647 Adjusted R-squared 0.910296 S.D. dependent var 0.203997 S.E. of regression 0.061098 Akaike info criterion -2.505006 Sum squared resid 0.033597 Schwarz criterion -2.331175 Log likelihood 20.28254 Hannan-Quinn criter. -2.540736 F-statistic 41.59113 Durbin-Watson stat 1.858474

18 Prob(F-statistic) 0.000013

19 Annex B

1 Electricity, gas, and water supply 2 Manufacture of refined petroleum products List of potential variables List of potential variables 1. Volume of water consumption (in 1000 m3) 1. Quantity of oil refined (in 1000 barrels) 2. Volume of electricity consumption (in 1000 kWh) 2. Number of employees 3. Volume of gas consumption (in 1000 pounds) 3. WTI oil price 4. Utilities index 4. Real GDP U.S.A. 5. WTI oil price 6. Mid-year population

4 Wholesale and retail trade, repair of motor 3 Construction vehicles and household goods List of potential variables List of potential variables 1. Number of construction permits granted 1. Real tourism exports (in Afl. million) 2. Total cement imported (in 1000 Kg) 2. Partial Economic Activity Index: Trade index 3. Number of electrical installations approved 3. Mid-year population 4. Number of employees in construction sector 4. Consumer price index

5 Hotels 6 Restaurants List of potential variables List of potential variables 1. Stay-over visitors (x1000) 1. Stay-over visitors (x1000) 2. Tourist nights (x1000) 2. Tourist nights (x1000) 3. Average daily expenditure food & beverage 3. Average hotel occupancy rate (in US$) 4. Average daily rate hotels 4. Real tourism receipts (in Afl. million) 5. Number of hotel/timeshare rooms 5. Real tourism exports (in Afl. million) 6. Real GDP U.S.A. 6. Real GDP U.S.A.

7Transport, storage and communications 8 Financial intermediation List of potential variables List of potential variables 1. Real imports of goods, excluding oil sector (in Afl. 1. Bank credit to private sector (in Afl. million) million) 2. Real tourism receipts (in Afl. million) 2. Interest margin commercial banks 3. Real tourism exports (in Afl. million) 4. Real import of goods, general trade (in Kg) 5. Real GDP U.S.A.

9 Real estate activities 10 Other business activities List of potential variables List of potential variables 1. Mid-year population 1. Real tourism receipts (in Afl. million) 2. Nominal interest rate, CBA offered rate (7 days) 2. Real tourism exports (in Afl. million) 3. Real GDP U.S.A. 3. Real GDP U.S.A.

11 Other community, social and personal service activities 12 Adjustment for FISIM

20 List of potential variables List of potential variables 1. Real tourism receipts (in Afl. million) 1. Bank credit to private sector (in Afl. million) 2. Real tourism exports (in Afl. million) 3. Real GDP U.S.A. Annex C

1 Electricity, gas, and water supply

Dependent Variable: LOG(GVAELEC/PHOUS) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -20.71021 5.150511 -4.021000 0.0030 LOG(WTI) -0.284152 0.076952 -3.692586 0.0050 LOG(POP) 2.292636 0.469847 4.879541 0.0009

R-squared 0.731920 Mean dependent var 4.530819 Adjusted R-squared 0.672347 S.D. dependent var 0.107408 S.E. of regression 0.061482 Akaike info criterion -2.527840 Sum squared resid 0.034020 Schwarz criterion -2.406613 Log likelihood 18.16704 Hannan-Quinn criter. -2.572722 F-statistic 12.28603 Durbin-Watson stat 2.873548 Prob(F-statistic) 0.002674

2 Manufacture of refined petroleum products

Dependent Variable: LOG(GVAREF/Pm) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -10.57335 6.276678 -1.684546 0.1264 LOG(WTI) 0.501310 0.173928 2.882288 0.0181

LOG(yusa) 1.471923 0.736481 1.998588 0.0767

R-squared 0.895077 Mean dependent var 4.604217 Adjusted R-squared 0.871761 S.D. dependent var 0.395531 S.E. of regression 0.141641 Akaike info criterion -0.858720 Sum squared resid 0.180560 Schwarz criterion -0.737493 Log likelihood 8.152317 Hannan-Quinn criter. -0.903602 F-statistic 38.38875 Durbin-Watson stat 2.110913 Prob(F-statistic) 0.000039

21 3 Construction

Dependent Variable: LOG(GVACON/Pi) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -2.817944 1.792858 -1.571760 0.1471

LOG(EMPLCON) 0.986841 0.219003 4.506071 0.0011

R-squared 0.670018 Mean dependent var 5.259543 Adjusted R-squared 0.637020 S.D. dependent var 0.182061 S.E. of regression 0.109688 Akaike info criterion -1.431349 Sum squared resid 0.120314 Schwarz criterion -1.350531 Log likelihood 10.58809 Hannan-Quinn criter. -1.461271 F-statistic 20.30468 Durbin-Watson stat 1.500889 Prob(F-statistic) 0.001132

4 Wholesale and retail trade, repair of motor vehicles and household goods

Dependent Variable: LOG(GVATRAD/Py) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -24.17337 8.165126 -2.960563 0.0181

LOG(Pcp) -2.645591 0.496744 -5.325865 0.0007

LOG(Xtour/Pxtour) 0.534549 0.174698 3.059847 0.0156 LOG(POP) 2.327619 0.777007 2.995622 0.0172

R-squared 0.948054 Mean dependent var 5.776993 Adjusted R-squared 0.928574 S.D. dependent var 0.109743 S.E. of regression 0.029330 Akaike info criterion -3.959240 Sum squared resid 0.006882 Schwarz criterion -3.797604 Log likelihood 27.75544 Hannan-Quinn criter. -4.019083 F-statistic 48.66873 Durbin-Watson stat 3.035919 Prob(F-statistic) 0.000018

22 5 Hotels

Dependent Variable: LOG(GVAHOT/Pm) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -5.454436 2.046183 -2.665664 0.0237

LOG(yusa) 1.179225 0.222968 5.288770 0.0004

R-squared 0.736642 Mean dependent var 5.366664 Adjusted R-squared 0.710306 S.D. dependent var 0.148732 S.E. of regression 0.080052 Akaike info criterion -2.061260 Sum squared resid 0.064084 Schwarz criterion -1.980442 Log likelihood 14.36756 Hannan-Quinn criter. -2.091182 F-statistic 27.97109 Durbin-Watson stat 2.107108 Prob(F-statistic) 0.000353

6 Restaurants

Dependent Variable: LOG(GVAREST/Pcp) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -6.475053 1.167142 -5.547786 0.0004

LOG(yusa) 0.404203 0.165210 2.446603 0.0370

LOG(Xtour/Pxtour) 1.023165 0.231971 4.410752 0.0017

R-squared 0.907368 Mean dependent var 4.482376 Adjusted R-squared 0.886784 S.D. dependent var 0.120502 S.E. of regression 0.040546 Akaike info criterion -3.360436 Sum squared resid 0.014796 Schwarz criterion -3.239209 Log likelihood 23.16262 Hannan-Quinn criter. -3.405318 F-statistic 44.07955 Durbin-Watson stat 1.455060 Prob(F-statistic) 0.000022

23 7Transport, storage and communications

Dependent Variable: LOG(GVATR/Pcp) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -0.848553 1.088670 -0.779439 0.4538

LOG(yusa) 0.686743 0.118630 5.788955 0.0002

R-squared 0.770178 Mean dependent var 5.453310 Adjusted R-squared 0.747196 S.D. dependent var 0.084710 S.E. of regression 0.042592 Akaike info criterion -3.323298 Sum squared resid 0.018141 Schwarz criterion -3.242480 Log likelihood 21.93979 Hannan-Quinn criter. -3.353219 F-statistic 33.51200 Durbin-Watson stat 1.769267 Prob(F-statistic) 0.000176

8 Financial intermediation

Dependent Variable: LOG(GVAFIN) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C 0.883065 0.402676 2.192993 0.0531 LOG(CRED) 0.611851 0.054719 11.18165 0.0000

R-squared 0.925942 Mean dependent var 5.382595 Adjusted R-squared 0.918536 S.D. dependent var 0.179880 S.E. of regression 0.051341 Akaike info criterion -2.949636 Sum squared resid 0.026359 Schwarz criterion -2.868818 Log likelihood 19.69782 Hannan-Quinn criter. -2.979558 F-statistic 125.0293 Durbin-Watson stat 1.745469 Prob(F-statistic) 0.000001

24 9 Real estate activities

Dependent Variable: LOG(GVARE/Py) Method: Least Squares Sample (adjusted): 1996 2006 Included observations: 11 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -2.870101 0.852905 -3.365088 0.0099 LOG(1+r) -0.758720 0.609028 -1.245788 0.2481

LOG(yusa) 0.947025 0.093311 10.14910 0.0000

R-squared 0.966643 Mean dependent var 5.845171 Adjusted R-squared 0.958303 S.D. dependent var 0.100231 S.E. of regression 0.020467 Akaike info criterion -4.713012 Sum squared resid 0.003351 Schwarz criterion -4.604495 Log likelihood 28.92156 Hannan-Quinn criter. -4.781416 F-statistic 115.9136 Durbin-Watson stat 1.749788 Prob(F-statistic) 0.000001

10 Other business activities

Dependent Variable: LOG(GVABUS/Py) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -6.719526 1.557612 -4.313991 0.0015

LOG(Xtour/Pxtour) 1.675145 0.219860 7.619143 0.0000

R-squared 0.853052 Mean dependent var 5.147501 Adjusted R-squared 0.838357 S.D. dependent var 0.139831 S.E. of regression 0.056219 Akaike info criterion -2.768119 Sum squared resid 0.031606 Schwarz criterion -2.687301 Log likelihood 18.60872 Hannan-Quinn criter. -2.798041 F-statistic 58.05133 Durbin-Watson stat 1.579561 Prob(F-statistic) 0.000018

25 11 Other community, social and personal service activities

Dependent Variable: LOG(GVAOTH/Pm) Method: Least Squares Sample (adjusted): 1995 2006 Included observations: 12 after adjustments

Coefficient Std. Error t-Statistic Prob.

C -0.519710 0.587188 -0.885083 0.3991

LOG(yusa) 0.422047 0.083117 5.077748 0.0007

LOG(Xtour/Pxtour) 0.270782 0.116704 2.320248 0.0455

R-squared 0.920006 Mean dependent var 5.271451 Adjusted R-squared 0.902229 S.D. dependent var 0.065238 S.E. of regression 0.020399 Akaike info criterion -4.734373 Sum squared resid 0.003745 Schwarz criterion -4.613147 Log likelihood 31.40624 Hannan-Quinn criter. -4.779256 F-statistic 51.75407 Durbin-Watson stat 1.715100 Prob(F-statistic) 0.000012

12 Adjustment for FISIM Dependent Variable: FISIM Method: Least Squares Date: 08/31/09 Time: 10:13 Sample: 1995 2006 Included observations: 12

Coefficient Std. Error t-Statistic Prob.

C -67.29981 10.33905 -6.509282 0.0001 CRED -0.060350 0.006161 -9.794794 0.0000

R-squared 0.905605 Mean dependent var -165.0908 Adjusted R-squared 0.896166 S.D. dependent var 28.87842 S.E. of regression 9.305591 Akaike info criterion 7.450120 Sum squared resid 865.9402 Schwarz criterion 7.530937 Log likelihood -42.70072 Hannan-Quinn criter. 7.420198 F-statistic 95.93798 Durbin-Watson stat 1.940161 Prob(F-statistic) 0.000002

26 Table 2

Description of variables (in Afl. millions, unless otherwise stated) Variable Name Description Endogenous variables Cp Private consumption expenditure, in current prices

EXCHRUS$ Real exchange rate vis a vis U.S. dollar EXPORTS Total Exports, constant (1995) prices EXPORTS$ Exports of goods and services

Xoth Exports other than tourism, constant (1995) prices

Xoth Exports of goods and services other than tourism exports, in current prices EXPORTS_TOUR Tourism Export, constant (1995) prices

Xtour Tourism exports, in current prices

GVAAGR Gross value added from agriculture, hunting, forestry, fishing, mining and quarrying; in current prices

GVABUS Gross value added from other business activities, in current prices

GVACON Gross value added from construction, in current prices

GVAELEC Gross value added from electricity, gas, and water supply; in current prices

GVAFIN Gross value added from financial intermediation, in current prices GDPHe Gross value added from health and social work; Other community, social and personal service activities; in current prices

GVAHOT Gross value added from hotels and restaurants, in current prices

GDPMAN Gross value added from manufacturing (excluding oil refinery), in current prices GVA_REFINERY$ Value added from the oil refinery GDP_PUBADMIN$ Value added from public administration, compulsory social security and education GVA_REALEST$ Value added from real estate activities GVA_TRANSP$ Value added from transport, storage, and communication GVA_TRADE$ Value added from wholesale and retail trade, repair of motor vehicles and households goods GDPMP GDP at market prices, constant (1995) prices GDPMP$ GDP at market prices IMPORTS Real Imports, 1995 prices IMPORTS$ Imports of goods and services INT_RATE_R_CBOR7 Real interest rate, CBA offered rate (7 days) INV_PRI$ Private Investment M1_$ Money Supply (M1) P_CONS_G Government expenditure deflator P_CONS_PRI Consumer price deflator P_EXPORTS_OTH Price deflator for Exports other than tourism exports P_EXPORTS_TOUR Price deflator for tourism exports P_GDP GDP deflator P_IMP Imports deflator P_INV Investment deflator

27 TAXREV$ Total Tax revenue Exogenous variables CONS_G$ Government consumption expenditure CPI_USA Consumer price idex U.S.A. EXCHRATE Nominal exchange rate FISIM$ Value added adjustment for Financial Services Indirectly Measured (FISIM) yusa Real GDP U.S.A. (at 2000 prices, billion of US dollars) INV_G$ Government Investment

28

Recommended publications