(1) ECONOMETRIC MODEL FORECASTING – A STRONG AND WELL APPLICABLE TOOL IN VIETNAM (part 1)

Bài viết cũ của tôi:

ECONOMETRIC MODEL FORECASTING – A STRONG AND WELL APPLICABLE TOOL IN VIETNAM

Probably great attention has never been given to macroeconomic analysis and forecasting by the Party, State and economists of our country’s as it is now. This is explained by the fact that our country has been undergoing a drastic transition to a market-oriented economy and has actually found itself deeply integrated into the world economy. The operation of the market economy mechanism and the impacts of the world economy on ours have given rise to many quite new issues, some of which even appear for the first time in the history of our economy, such as the existing unprecedented huge inflow of foreign capital, or the unprecedented high weight of influence by world prices in our economy’s inflation rate, among others while the role of the Government’s direct instruments for economic regulation has been increasingly limited. It is exactly under such context that economic analysis and forecasting in a scientific manner based on which indirect instruments can be applied in economic regulation has emerged as a central task of the Government at the new stage of development.

This paper aims to review the experience with the establishment and application of econometric models in macroeconomic analysis and forecasting in recent years at the Department of General Economic Issues of the Ministry of Planning and Investment, hence to affirm that macro-econometric modelling is a strong and completely well applicable tool for macroeconomic analysis and forecasting in our country in service of national economic development planning and management.

The paper is composed of three parts:

1.    The first part provides a review of the application of econometric models in macroeconomic analysis and forecasting in recent years at the Department of General Economic Issues of the Ministry of Planning and Investment to come up with an initial evaluation of the effectiveness of this tool.

2.    Part II discusses a number of experiences and issued raised for the continued and more efficient establishment and application of macro econometric models in the context of Vietnam.

3. The last part presents conclusions on the role of the macro econometric model and, at the same time, general suggestions on how it should be applied in the coming time. In particular, on this occasion, we wish to express our desire to soon witness the formation of the Vietnam Econometrics Association, and see that further investments in econometric modelling work and financial support to the activities of this Association will be provided by the institutions interested in using the results from econometric modelling.

Following is a detailed discussion of the above-mentioned topics:

I.     A review of the Application of Econometric Models in Macroeconomic Analysis and Forecasting in Recent Years at the Department of General Economic Issues

From 2000 to date, two econometric models have been established at the Department of General Economic Issues (or the General Department in short) for short-term and medium-term macroeconomic forecasting in service of annual and five-year planning.

1. The Year 2000 Quaterly Macro Econometric Model

(1) Model Description

It can be said that the Quaterly Macro-econometric Model established at the Overall Department in the year 2000 was the first quaterly macro econometric model built in Vietnam for all previously constructed models were annual models due to the great scarcity of quarterly data during that time. This model was established to be used for economic analysis, policy simulation and short-term forecasting of our economy's developments in service of the management of annual socio-economic development plan implementation at the General Department.

Since the building of quarterly models was quite new in our country at that time, research results were mainly of experimental nature. Those involved in the model building shared the view that additional studies required to be conducted to ensure that results from the model were reliable enough. For this reason, our research project was entitled “Experimental Establishment of a Quarterly Econometric Model for Analysis and Forecasting of Short-term Developments the National Economy”, i.e. an empashis was placed on the aim to experiment the estblishment of the econometric model. It is expected that in the years to come, if the research project is extended, its participants would set a greater objective which is to concentrate on economic analysis, policy simulation and forecasting of short-term developments of the national economy based on improving and completing the model established within the framework of this research project, thus providing another scientific foundation in direct service of managing plan implementation at the Ministry of Planning and Investment.

Given the set objectives, the pieces of macroeconomic research modelled in the model cover major general macroeconomic indicators that are directly used for short-term analysis and forecasting at the Ministry of Planning and Investment. The model focused on analysing and producing forecasts for about 25 major macroeconomic indicators based on a maximum use of valuable reliable sources of information available.

The model, called Vietnam Quartely Econometric Model, or VQEM in short, describes  the economy from a financial and monetary perspective to evaluate the effectiveness of short-term economic adjustment policies and produce short-term forecasts.

In a market-oriented economy, the economy is usually divided into two sectors: financial and monetary sector and real sector. Since these two sectors are closely linked in an unified entity, an analysis of the interactions between them and a look at the impacts of changes in the financial and monetary sector on changes in the real sector is extremely needed. However, given that the aim of the model is to produce short-term analysis and forecasts, the main object to be analysed is the role, impacts and effectiveness of financial and monetary policies. This is also the common approach to the establishment of short-term forecasting models in many countries for the financial and monetary sector has been playing an increasingly significant role in economic growth and structural transformation. Moreover, since the ultimate goal of analysis and forecasting remains to be the possible growth of indicators of the real sector, the major outputs of the model are indicators of the real sector.

An analysis of the macroeconomic developments during the period 1997-2000 indicates that our economy was under a situation of deflation and excessive supply. If the scarcity of consumer goods and resources was a common phenomenon in the past, making the “requesting and giving” subsidy mechanism become a principle that dominated the entire mechanism of movements of our economy, and most models were established based on the supply side model, since 1998, a completely reverse situation was witnessed: excessive supply of goods, capital and money has been widespread, pulling down the price of consumer goods and the interest rate almost continuously every month. The central approach of VQEM is therefore a demand side approach in which the main factors affecting the real sector, particularly economic growth rate, would be financial and monetary ones, and, at the same time, major policies for accelerating economic growth rates during that period were nothing but consumer’s demand and investment stimulation.

Being a demand side model, the logic of VQEM is established along the line of Keynesian theory (demand theory). The model is composed of three blocks: real block, financial and monetary block, and external trade block. In the real block, according to Keynesian theory, production is determined by components of aggregate demand that include consumption, investment and net export (the difference between export and import). In particular, investment funds come from various sources each of which is influenced by various factors like real interest rate, population’s income, the economy’s monetarisation, and increased share of credit for the private sector... Population’s consumption is determined by their income and the Government's expenditure encouragement policy while government consumption serves as an exogenous variable so that the Government can take initiative in increasing rapidly their consumption.

The financial and monetary block provides indicators on budget revenue and expenditure, money supply, and inflation. Its general arguments are all based on Keynesian theory. Budget revenues depends on economic growth while budget expenditure depends on budget revenue and government’s investment expansion policy; budget revenue deficit is treated as an exogenous variable for it is an instrument of the State for exercising consumption stimulation through State budget performance.

The external trade block is also established based on the principle of the demand side model with addition of factors of the international competitiveness theory. On the one hand, exports depend on the demand of the trading partners of our economy while imports depend on the growth rate of the economy. On the other, both of these variables depend on price competitiveness of the economy through the real exchange rate, represented by the difference between domestic price and international price. The value of merchandise exports and imports will serve as the basis for computing that of services exports and imports to be put into the general equilibrium equation in the national accounts table.

In general, the model consists of 21 identity equations and 14 behavioural equations, with 35 endogenous variables and 11 exogenous variables. Exogenous variables are fairly sufficient for analysing, simulating and producing forecasts for the impacts of the demand stimulation policy on short-term economic growth.

In addition, to simplify the process of computation, all equations of the model are estimated by the ordinary least square method. Actually, all estimation results by this method generally meet the requirements of econometric techniques given that a number of re-estimations using the two-step least square method do not indicate any significant changes compared to the results of the coefficients estimated by the ordinary least square method.

(2) Major Forecasts Produced by the Model

Having run the model, the research team has put forward three forecast scenarios, namely the central scenario[1] (also called background scenario or baseline scenario), persimistic scenario, and optimistic scenario. Forecasts were produced for three years of 2001-2003. A comparison between the forecasts by the central scenario of the model and actual economic statistics published by the General Statistics Office (GSO) is shown below.

Table 1: Comparison of Forecasted and Actual GDP Growth Rates

 (in percentage)

 

Period of Time	Forecasted GDP Growth Rate	Actual GDP Growth Rate
First quarter 2001	7.16	7.15
Second quater 2001	7.00	7.00
Third quarter 2001	7.84	7.00
Fourth quarter 2001	7.07	7.20
First quarter 2002	7.3	6.6
Second quater 2002	7.9	6.6
Third quarter 2002	8.9	7.6
Fourth quarter 2002	9.1	7.5
First quarter 2003	8.4	7.0
Second quater 2003	7.1	7.1
Third quarter 2003	11.6	7.6
Fourth quarter 2003	8.5	7.7

                     

Table 2: Comparison of Selected Forecasts with Actual Figures

 

Indicator / Year	Planned	Forecasted	Actual
GDP growth rate (%)
2001	7.5	7.3	6.89
2002	7-7.3	8.0	7.08
2003	7-7.5	8.7	7.34
Total investment in the economy (in bill. VND)
2001	150	152	170.5
2002	175	171	200.0
2003	215	198	239.2
Total budget revenue (in bill. VND)
2001	96.3	98	103.9
2002	105	105	123.9
2003	123.7	112.2	152.3
Total budget expenditure (in bill. VND)
2001	116	117.5	129.8
2002	133.7	125	148.2
2003	158	135	181.1
Consumer goods price growth rate
2001	3	1.0	0.8
2002	3-4	2.5	4.0
2003	5	3.1	3.0
Exports (in bill. US$)
2001	16.1-16.3	16.4	15.0
2002	16.6	18.5	16.7
2003	18-18.2	20.3	20.1
Imports (in bill. US$))
2001	17.1	15.5	16.2
2002	17.5	16.6	19.7
2003	21-21.5	18.5	25.2

(3) Remarks

The compared figures from the above-mentioned two tables show that the quarterly and annual forecasts for the period 2001-2003 produced by the model reflect a fairly accurate trend of actual growth (upward and downward), especially the trend of gradual improvement of the economy over quarters.

On the other, forecasts are very close to planned targets, although there is a relatively big difference between both of these and actual figures. Moreover, a comparison between specific forecasts for about 25 major economic indicators and actual figures of the economy also indicates a considerable difference. 
                                                                                                                                                                                                                                                                                                                                                                                        The difference between forecasts and actual figures can be explained by many factors of which the four most important factors are:

- First, the model itself is not able to produce good forecasts despite its relatively strong ability to describe the past. In using the model to produce forecasts for 2001 in mid-third quarter of 2001, we discovered significant errors. For example, the forecasts produced by the model showed a GDP growth rate of 7.3% for 2001 while it was 6.8% according to estimates by the General Statistical Office, indicating a 0.5% difference. Similarly, consumer price was forecasted to decrease by 1.8% while estimated figure was minus 0.6%... One should have improved the model to produce better forecasts for 2001 before using it for forecasting. Nevertheless, given the limited timeframe for a ministerial level research project and that it was just an experimental period for quarterly model building, correction of the model were not done, leading to further exaggerated errors of forecasts for 2002 and 2003. As such, differences between forecasted and actual figures are fully understandable, and foreseeable before the completion of the research project.

- Secondly, the building of the baseline scenario for forecasting was conducted when the world economy was in a fairly satisfactory situation. Most forecasts indicated that the economies of the United States and Europe were on a solid recovery momentum, and Japan’s economy would see fairly rapid growth. It was not the case, however. The US and European economies, following a short period of recovery, saw a dramatic recession, particularly after the event of terror happened on September 11, 2001 in the US. The Japanese economy also fell into the circle of stagnance. The EU economy remained gloomy. In face of drawbacks and difficutlties in major economies in the world, reduction in the growth rates of East Asian economies was unavoidable, creating adverse impacts on our country’s trade and investment. This of course also led to errors in the results of forecasts.

- Thirdly, in producing forecasts for 2002 and 2003, those involved in the model building did expect relatively bold changes in the Government policies. In particular, the pace of demand stimulation was expected to be accelerated through tax reduction policy, budget expenditure restructuring, adoption of the monetary expansion policy, exchange rate adjustment, rapid growth of exports and reduction of imports to achieve net exports surplus right from 2001. As a result, those policies changes were expected to bring about not only a rise in consumption and expenditure of the State economic sector, but also, and mainly, a growth in those of other economic sectors. Moreover, the accelerated process of economic structural transformation and economic liberalisation would contribute to raising the competitiveness of goods and reducing production costs...

Unfortunately, these predictions did not come true. Instead, budget expenditure to GDP ratio continued to increase, budget deficit exceeded the forecasted level, the pace of money and credit expansion remained very slow; real interest rate was too high; adjusted exchange rate was lower than expectations... As a consequence, consumption and investment of the private sector grew very slowly while the activities of the foreign invested sector sped down (disbursement rate was quite low); and the only exception was the strong growth of the State economic sector’s consumption and investment. This also led to lower efficiency of the economy’s performance that was contrary to the expectation for improved effficiency of investment...

- Fourthly, during the period 2001-2005, many statistical figrues for the period 1990-2005 were reproduced by the General Statistics Office to reflect better the reality, creating a very big difference between the data series used for model building (the 1990-2000 series of data) and the one reproduced by the General Statistics Office that ultimately led to extremely great difficulties in comparing the forecasted figures with the actual ones. Typical examples were found in the case of figures on development investment and budget revenue and expenditure... Hence, not only forecasted figures, but also planned figures, were not accurate compared to adjusted actual figures.

- Finally, one should mention that due to limited timeframe for the research project, the forecasters did not use more sophisticated techniques, e.g. the technique to adjust errors at the final point of time of the model’s estimation period, or error forecasting technique...., to make their forecast results closer to reality.

2. Revised Quarterly Macro Econometric Model - VQEM-2003

(1) Model Description

In general, the structure of the VQEM-2003 is not much different from the original VQEM. However, given the availability of more abundant data, in the revised model - VQEM-2003, we established one more block of identity equations for international balance of playment which is completely new compared to the original VQEM. Therefore, VQEM-2003 consists of as many as 47 equations including 28 accounting equations and 19 behavioural equations while the original VQEM consists of only 35 equations (21 accounting equations and 14 behavioural equations). The revised model was established with the 1990-2001 actual data series and an estimated series for 2002, and is used to produce forecasts for the period 2003-2005.

(2) Major Forecasts Produced by the Model

a) Forecasts for three years of 2003-2005

Table 3: Comparison of Selected Forecasted Figures and Actual Figures 

Indicator / Year	Planned	Forecasted	Actual
GDP growth rate (%)
2003	7-7.5	7.55	7.34
2004	7.5-8	7.90	7.79
2005	8-8.5	8.20	8.44
Total investment in the economy (in bill. VND)
2003	215	217.6	239.2
2004	249-255	258	290.9
2005	295-300	300	343.1
Total budget revenue (in bill. VND)
2003	123.7	132	152.3
2004	148.3	149	190.9
2005	179.7	169	217.1
Total budget expenditure (in bill. VND)
2003	158	161	181.1
2004	186.7	184	214.2
2005	226.5	209	223.6
Consumer goods price growth rate
2003	5	4.0	3.0
2004	4-5	4.6	9.5
2005	< 8-8.5	4.2	8.4
Exports (in bill. USD)
2003	17.3	19.3	20.1
2004	21.7	22.0	26.5
2005	28.5	25.5	32.4
Imports (in bill.USD)
2003	19.3	22.9	25.2
2004	26.0	25.1	32.0
2005	34.0	28.1	36.8

As such, the forecasts produced by the VQEM-2003 model shows that our economy continued to see improvements during three years from 2003 to 2005, reflecting the positive impacts of the domestic demand stimulation policy. Nevertheless, the rate of economic growth would slow down since mid-2005, particularly as a consequence of the war in Iraq.

b) Forecasts for Possibility to Complete the 2001-2005 Five-Year Plan

Table 4: Comparison of Forecasted Figures with Planned and Actual Figures of 2001-2005 Five-Year Plan

Indicator	Unit	Planned	Forecasted	Actual
GDP growth rate	%	7.5	7.5	7.5
of which:
- Industry	%	10.8	9.2	10.2
- Agriculture	%	4.3	4.3	3.8
- Services	%	6.2	7.3	7.0
GDP Composition by 2005
- Industry	%	38-39	39	40.0
- Agriculture	%	20-21	21	20.9
- Services	%	41-42	40	38.1
Exports growth rate	%	14-16	11.2	17.5
Growth rate of consumption fund	%	5.5	7.2	7.0
Domestic savings to GDP ratio	%	28-30	32.5	35.4
Disbursed foreign funds	Bill. US$	9-10	12.8	14.3
Budget revenue to GDP ratio	%	20-21	22.1	24.4
Development investment growth rate	%	11-12	13.2	17
Investment to GDP ratio	%	31-32	35.2	38.9
Composition of invested funds
- State sector	%	58-60	57.4	52.1
- Population and private enterprise sector	%	24-25	25.0	28.6
- Foreign invested sector	%	16-17	17.6	16.6

(3) Remarks

- The above-mentioned comparison table indicates a fairly closeness between forecasted and planned figures, and, at the same time, a certain difference between them and the actual figures. However, the forecasted figures are closer to the planned ones.

- Compared to the results from the original VQEM, those produced by the VQEM-2003 are much closer to reality.

- Yet, many indicators remain far from the reality. The main explanation for this is similar to that in the case of the original VQEM. In addition, another very important factor leading to this is the approach to economic development taken by the forecasting model design and application group was different from that of the Government, reflected in different policies (In fact, the forecasting group set wrong assumptions on development policies to be introduced by the Government.). As such, their forecast results were greatly influenced by those incorrect policy assumptions.

3) Annual Macro Econometric Model (VMEM)

(1) Model Description

An analysis of macroeconomic relationships shows that the development process of the economy of Vietnam following the 1986 reform is one of gradual transformation from a basically centrally planned economy to a market-oriented economy. As of 2004, Vietnam’s economy had basically operated in accordance with the principles of the market economy, in which market factors including both supply and demand are truly determinants to growth.

In the medium and long run, the supply factors that serve as inputs into the process of growth (capital, labour, technology level...) continue to be determinants to economic growth. However, in addition to traditional factors explaining for growth according to the modern neo-classical theory of growth, considerable roles are also played by demand factors according to the theory of structure. Therefore, the annual econometric model that produces economic analyses, policy simulation, and short-and medium-term forecasts would be basically a mixed model that combines the neo-classical model and certain factors of the modern Keynesian theory, i.e. supply plays a central role in the model while certain elements of demand are added to the model. The VMEM-2004 model is composed of: (i) The real block that identifies the macroeconomic indicators of the real sector; (ii) The monetary, financial and price block that identifies the indicators of the financial sector; (iii) and the external trade block that identifies indicators on exports, imports, and external economic relations.

(2) Possibility for Development during the Period 2006-2010 Following the Baseline Alternative

Since the model was established in 2004, forecasts for the intermediate year of 2005 is required for the model to be able to produce forcasts for the five-year period 2006 to 2010. The model should have been revised in 2005 so that by the end of this year forecasts for the period 2006-2010 can be made with updated information for 2005. Nevertheless, this was not the case. An evaluation of the results of the model run in August and September 2004 is provided below.

Actual figures of the 2006-2010 five-year plan are yet available at present, following are some pieces of information drawn from the model in 2004 to compare with actual implementation of the five-year plan during the first two years 2006 and 2007 as well as the estimates for the third year 2008 of the plan period.

a) Population and Employment: Results from the model indicate that in the coming five-year (2006-2010) plan, our population would grow at an annual rate of 1.23% (actual growth rates in 2006 and 2007 were 1.26% and 1.3% respectively, and a growth rate of 1.25% is expected for 2008); by 2010 the country will have a population of 88.3 million inhabitants (actual figures for 2006 and 2007 were 84.2 million and 85.3 million, and a population of 86.3 million people is expected for 2008).

Employment growth rate of the economy was forecasted to be high at about 3% per annum (the three years 2006-2008 actually saw an average growth rate of 2.6%), in which employment growth rate is highest in the services sector, around 8.7% per annum, followed by that in industry sector, about 4.7% while that in agricultural sector is the lowest - 1.25% per annum. By 2010, agricultural employment would make up for only 53.4% of the total workforce (actual figure for 2007 was 54.7%)[2]...

b) Production: According to the model forecasts, the period 2006-2010 would see an average annual GDP growth rate of 8-8.1%, or similar to the rate predicted in the high scenario of the 2006-2010 five-year plan (7.5-8%). In particular, the forecasted rates of GDP growth tend to accelerate over years: 7.7% for 2006 (actual figure was 8.17%), 7,8% for 2007 (8.5% in reality), and respective figures for 2008, 2009 and 2010 are 8.1%, 8.3%, and 8.5%.

c) State budget: Total budget revenue for the five-year period 2006-2010 was forecasted to reach VND 1,697 thousand billions, representing an average annual growth of 21.1% (actual figrues for the three-year period 2006-2008 was VND 874 thousand billion). The budget revenue to GDP ratio is forecasted at 20.8%.

Total budget expenditure for the five-year period 2006-2010 would reach VND 2,142 thousand billion according to the model results, indicating an average growth rate of 21.9% per annum (actual figrues for the three-year period 2006-2008 was VND 1,087 thousand billions). The budget expenditure to GDP ratio is forecasted at 26.3%.

d) Prices: During the period 2006-2010, inflation rate is forecasted to continue its upward trend that has been established in recent years. The forecasts produced by the model indicate that the inflation rate in 2005 would be 7.5% (actual figure was 8.4%), the rate would go up to 8.5% for 2006 (it was actually 6.6%), and 9.7% for 2007 (actual figure is estimated at 8.5%), 11% for 2008, 12.2% for 2009 and 13.5% for 2010. On average, the inflation rate during the five-year period 2006-2010 would be approximately 11% per annum. As such, the inflation rate in the coming five-year (2006-2010) plan would probably be out of control, particularly from 2008 onwards if the demand stimulation-based growth expansion policy continues.

Forecasts up to 2010 show that the economy of Vietnam would not only get out of the state of prolonged negative inflation that lasted for many previous years, but also see overheated growth, causing increases in inflation rates. On the other hand, if a precautious montetary policy is not adopted, and efforts are just made to meet the economy’s demands for credits and investments without giving due attention to improving the efficiency of investment capital, then the rate of monetary expansion would continue to be quite high, leading to rapid growth of inlation.

e) Foreign trade: Export turnover is forecasted to reach approximately US$ 58 billion in 2010 (equal to the planned target for 2008). The figure for imports in the same year is nearly US$ 77 billion (the planned target for 2008 is US$ 69 billion). Rapid growth of imports is explained  by continued VND over-evaluation (resulting from high domestic inflation rates) and furthermore, a relatively high growth rates of both consumption and exports. Due to considerably higher growth rate of imports against exports, foreign trade deficit would dramatically increase from some US$ 3.4 billion in 2005 to US$ 3.9 billion in 2006, US$ 5.2 billion in 2007, US$ 7.7 billion in 2008, and suddenly to as much as US$ 12 billion 2009 and nearly US$ 19 billion in 2010. The foreign trade deficit to GDP ratio would go to 13% in 2010 accordingly.

As such, similar to the case of inflation, the problems arisen for which concentrated efforts are required to address would only occur at the end of the five-year plan period, i.e. when imbalances in the economy exceed acceptable limitations, causing great changes that spontaneously restructure the economy. This implies that attention should be given to price control and foreign trade balance of payment, otherwise the developments in the last years of the plan period would deteriorate very quickly.

(3) Remarks

- In general, the model’s forecasted results so far have reflected a better economic situation compared to the planned targets, especially in terms of economic growth rate and employment generation. The model also produces fairly good forecasts for inflation growth given that the expanded growth policy continues to be in effect until 2010. The next consequence would be rapid growth of total budget revenue and expenditure (at current prices) while their ratios to GDP decrease; and, at the same time, export growth would slow down and import surplus would strongly increase...

- Nevertheless, since the model’s forecasted results mainly depend on the trends of indicators, it failed to foresee the Government’s timely introduction of policy measures to control inflation growth right in mid-2007; as a result, in the coming time (2008-2010), the rate of inflation would go down, leading to smaller total budget revenue and expenditure and larger budget revenue and expenditure to GDP ratios... The model also failed to foresee huge impacts of the Vietnam – US bilateral Trade Agreement (BTA) on Vietnam’s economy, Vietnam’s accession to the World Trade Organisation (WTO), and huge increases in crude oil price..., leading to its failure to anticipate strong growth of exports (although growth of import surplus has also been forecasted). Therefore, for the whole five-year period, predictions by the plan may be more reliable than forecasts produced by the model.

---

[1] The most possible forecast, i.e. taking into consideration the policies to be introduce by the Government.

[2] Adjusted data on labour were different from those used for model building; the latter forecasts that the share of agricultural labour in 2005 would be 60.30% while actual figure is only 57.3%... Therefore, one cannot make a comparison between forecasted figures and actual figures.