(2) ECONOMETRIC MODEL FORECASTING – A STRONG AND WELL APPLICABLE TOOL IN VIETNAM (part 2)

Mon ancien article - My old papers - Bài viết cũ của tôi:

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

II. Some Exprerience and Issues Raised Regarding Further Building and Application of Macro Econometric Models in Vietnam

1. Role of Econometric Modelling

First of all, one should repeat the salient merits of the econometric modelling tool over other normal methods in economic research, namely:

(1) Ability to allow for a combination of quantitative and qualitative methods to test economic conclusions, i.e. to verify whether economic relationships are quantitatively in accordance with the relationships drawn from economic management theories and practices, thus enabling the drawing of correct economic conclusions.

Take inflation as an example. For the same phenomenon of inflation, different explanations can be provided for it in different countries and in fact, the evolution of inflation in different countries usually follows different economic cycles. So far, our economic analyses have been mainly based on logical arguments without accompanying quantitative analyses to verify them while economic relationships are usually very complicated and hence analyses that merely base on logical arguments may come up with wrong conclusions.

An application of econometric tools helps improve the reliability of economic analyses because, on the one hand, before an econometric analysis is conducted, an analysis based on normal logical arguments is always required; and on the other, this step will be followed by econometric analyses to test if the conclusion drawn from the qualitative analyses are in line with the quantitative ones. In this case, one must analyse inflation based on various economic theories to find out the most suitable explanation, then proceed with model building, estimation and analysis to come up with accurate conclusions.

(2) Provision of Permitted Simulation Results

a) Analysis of the role of different factors in the process of the economy’s production and re-production. Examples include the impacts of taxation policy, wage policy, or a combination of these two policies on the development of the entire economic system.

b) Evaluation of the impacts of macroeconomic policies on the economy:

- Evaluation of the effects of past policies. For example, if you want to know how the economy would look like in 1985 if a price-wage-monetary reform were not introduced? The answer to this question can only be available with an application of the simulation technique.

- Forecasts for the impacts of future economic policies: Impacts of one specific policy or measure, or those of many measures taken at the same time or consecutively over months, quarters and years (not simultaneously taken) can be forecasted. Only if positive impacts of a specific policy or measure are verified by the model, then actual introduction of them into the economy should be made.

- The econometric model can be combined with the growth model to study growth and formulate development plans; with input-output model and inter-sectoral balances to both produce short-term analyses and forecasts, and serve long-term planning.

c) Production of different development scenarios based on different input assumptions including those for economic policies.

(3) Some other strengths of the model

- Ability to ensure an accounting match among economic balances;

- Ability to take into consideration an unlimited number of interdependence mechanisms among different economic variables and  indicators;

- A description of relationships using explicit behavioural equations enables everyone to see them clearly, make comments, and adjust by themselves based on their own arguments in order to improve the quality of analyses and forecasts;

- Ability to make simultaneous, accurate and extremely quick computations of related formulas and derived results if necessary;

- Ability to easily insert necessary changes into the model when theories of economic development are improved or changed;

- Ability to compare the model with other models to select the one that is most suitable to the objective reality.

- Ability to use forecasts for many years provided that there are no changes in the theories of economic development.

2) Some Major Advantages of Present Modelling Work Compared to Previous One 

- Rapid progress in economic theories that enables increasingly enriched and practical modelling of economic mechanisms;

- Progress in econometrics, including increasingly sophisticated methods of estimation that enable the acquisition of increasingly reliable and realistic models; more complete and enriched methods of statistical hyphothesis testing to be properly applied in specific contexts...

- The algorithms available for solving big models have been increasingly plentiful and specialised..., and the ability to solve them has been increasingly strong, thus limiting the number of cases of model convergence failure.

- The quality and strength of computation means have already been very high, particularly the capacity of computer has been increasingly large, enabling increasingly accurate solution of increasingly large models with increasingly sophisticated algorithms.

- Rapid progress in modelling and forecasting knowledge enables model builders to soon make the model adaptive to the issues raised to and meet with less difficulties in establishing hypothesis, thereby rapidly reducing the costs of forecasting.

- Specialised softwares and computer tools for modelling have been increasingly available.

- Data have been increasingly reliable, and better reflecting the reality. The number of series have been more and the period of obseravations been longer...

- Increasing attention has been given to the role of analytical and forecasting work as well as that of modelling work.

3) Some Notes to Producers of Analyses and Forecasts Based on Econometric Models

From our experience with direct involvement in producing analyses and forecasts based on econometric models, and at the same time our experience in giving lectures, and being members of evaluation councils for scientific research projects that involve econometric model application, it is our recommentation that attention should be given to several issues as follows:

(1) One must capture the nature of econometrics, especially must clearly understand that a model is a set of interdependent relationships that allow for a combination of different values of one quantity number or indicator in accordance with one or several predetermined theories. At present, it is the case that estimations of relationships are made in a careless manner without any theoretical foundations, leading to the fact that relationships between different indicators are just  mechanical, not causal ones and therefore are of no value for policy simulation and development forecasting  .

(2) The role played by economic theories is extremely important. Model builders must firmly comprehend economic theories and contrast them with our country’s economic reality to identify which theory best explains for the nature of our country’s development process, and next to apply exactly that theory to establish a theoretical approach for the model and a development approach for the economy. This is similar to the case in which a doctor (the Government) treats for a patient (an economy). When a person is sick and goes to the doctor for examination, the latter would recall a list of the symtoms of all sorts of deseases (economic theories) and listen to the patient’s explanation of his/her health status (economic status) to come up with comparisons and subsequenly, conclusions on what kind of desease the patient has got. Once the desease has been identified, the doctor would then select certain treatment medicines (economic policies) according to theoretical instructions on that desease (economic theories) to treat that kind of desease. One must therefore comprehend economic theories, from classical and neo-classical to neo-Keynesian and post-Keynesian theories and their development branches to apply them in analysing the nature of our economy’s performance, and hence, designing relevant models.

(3) One should master the basic techniques of model building process. Minimum techniques required include:

- A clear determination of whether an economic variable is endogenous or exogenous;

- A clear disctiontion between behavourial equations and accounting equations.

- Guarantee for model solvability. In many cases, we unintentionally make our model unsolvable. For example, actual estimations show that consumption is always greater than income and imports are required to close the gap while one of the assumptions for (short-term) forecasts is trade surplus, making the problem unsovable.

- Selection of model must guarantee its compatability with the range of forecasting. Quarterly or annual models can be used for short-term forecasts. Quarterly models are quite valuable for economic management to resist the cyclicity of the economy. The forecasting range for quaterly models is between one quarter to two years and rarely as long as five years. Medium-term model is built based on annual data series and can be applied to produce forecasts for a period of four to seven years. To develop a strategy, long-term forecasting models requires to be established with very few equations so as to produce growth forecasts for a fairly long period of time, from 10 to 30 years.

Forecasting range is usually identified right after the model building because short-, medium- and long-term economic mechanisms usually differ. Long-term models pay little attention to situational or temporary phenomema but great attention to long-term trends, istead. On the contrary, short-term models do not pay attention to long-term issues... It is therefore almost impossible to build the models that can be used to produce both short-term and long-term forecasts.

- Ensure a general accountability (equilibrium), including the accountability among endogenous variables, a logical relationship along the direction from exogenous to endogenous variables, no relationship among exogenous variables themselves, and a logical relationship along the direction from  endogenous to exogenous variables...

- Ensure the homogeneity of the model. Most of the existing  models fail to avoid this simple mistake. Failure to ensure the homogeneity of the model involves the existence of a linear relationship between value and quantity variables, establishment of a relationship between log and absolute variables, confusion between growth rate, percentage share and absolute figures (for example GDP is dependent on inflation rate, and budget expenditure for development investment)...

- One should not be too ambitious in producing forecasts for too many things within one model. In fact, each model contains only few number of explanatory variables while the economy is an increasingly complicated system and the behaviour of economic agents is highly changeable because they can always access new information. Being too ambitious in using a model therefore will make the model “unfocused”, leading to inaccrurate forecast results.

For this reason, the scope of model should not be expanded and instead, only higher quality models should be built, focusing on the areas that this tool can produce good forecasts. It is quite noteworthy that when economic developments take place in a regular, stable and sustainable manner, modelling tool should be applied as it can produce better forecasts than those made by experts. In the opposite case, when economic developments are instable, highly unusual and volatile, expert forecasts would be a better choice. Nevertheless, since expert forecasts are usually unconnected, they should be connected to a model to produce aggregate forecasts that guarantee the systematicity.

- Relevant selection of representative indicators is required. Economic theories are usually quite complicated as they go from micro economic issues, hence many data series are usually required for modelling while in fact they can not be available at the macroscopic level. Take the tax model as an example, tax revenue should be computed based on taxation base while in reality it is impossible to identify this indicator. An alternative indicator, e.g. GDP, should therefore be used. In that case, testing whether GDP is really a relevant alternative indicator is required.

- Limitation of expectations from the model is required and one should convince model users to limit his/her own expectations from the model. In particular, a clear recognition of the great difficulty in producing accurate forecasts for all indicators is required; as for short-term forecasts, in many cases, results from trend-based forecasting using VAR or ARIMA models are more accurate than those produced by the econometric model. Forecasting for the development trend of certain major indicators is somewhat difficult while it is easiest to apply the econometric model in identifying overal policy mechanisms for the economy, and fundamental relationships that are most influential in the economy.

- Common mistakes in forecasting work include: When the economy is in a difficult situation, the forecasts and their accompanying measures usually expect that the situation would soon improve... It is not that easy in reality. Or when the economy grows at high rates, people tend to rarely predict that economic growth would slow down, or even that the economy would fall into crisis in some years to come (if measures for structural adjustments are not taken right now)...

(4) Basic techniques of the forecasting process, especially dummy forecasting, scenario-based forecasting, trend-based forecasting, standard forecasting, variant-based forecasting, central forecasting (also called background or baseline forecasting that serves as controlling forecasts to contrast with other alternative forecasts), pessimistic forecast alternative, optimistic forecast alternative,... must be mastered.

Long-term forecasting is in essence forecasting for growth trajectory along the line of converging to a potential growth rate of individual indicators. Such process of gradual convergence is called the stationary axis and the final point of convergence is called the stationary point; this also means producing forecasts for the process of going to the point of long-term equilibrium growth.

The pace of gradual convergence to the long-term equilibrium point is called the speed of adjustment. The period of time required to complete the process of adjustment is called adjustment period. When the economy is faced with overheated growth, it must be, or will be spontaneously adjusted to go back to the equilibrium point so as to be in line with the potential growth tendency. In the opposite case, when the economy is in stagnation, it will be gradually adjusted to achieve higher growth rates... Economic policies work to shorten the period of adjustment in order to raise the economic and social efficiency, enabling the economy to go into a more sustainable growth trajectory more quickly.

In addition, one should master the techniques of forecasting for exogenous variables including pure exogenous variables forecasting and policy exogenous variables forecasting, establishment of scenarios for exogenous variables, and error forecasting, i.e. extrapolation of them for the forecast period.

(5) Maintenance of the model for long-term use: In general, the improvement of an econometric model and production of forecasts require a lot of time and must go through many complicated and repeated (it may take many quarters or years in case of big models) stages. Hence, if one wishes to ensure that a model can be of long-term use and can produce reliable forecasts, periodical update of the model is required. Some of the things to be done include:

- Periodic update of the model’s database, notably the database for endogenous and exogenous variables;

- Re-estimation of the model when new updated pieces of information are available, and correction of no longer relevant equations;

- Update and correction of the model’s equations to keep up with the development (structural transformation) momentum of the economy and economic theories;

- Periodic re-establishment of hypotheses related to exogenous variables and forecasted error variables;

- If the model is already really good and if necessary, the scope of the model can be expanded by adding to the model certain new blocks in line with the development momentum of the market-oriented economy (international balance of payment block, monetary block, domestic market block...), expanding some of the old blocks, disagreggating certain industries and sectors, and endogenise some variables that are previously exogenous...

- Whenever the General Statistics Office and line ministries and agencies make any changes in the definition of certain indicators or recomputation of a certain data series, update of information for the model, and correction and re-estimation of the model are required. In particular,  it is noteworthy that the most up-to-date data are usually estimates and therefore need to be regularly updated.

- Following the model’s re-estimation, its qualities should be checked to be sure that the model remains acceptable. Minimum checks to be conducted include simulation of the past for error analysis, simulation of miltiplier analysis, dummy forecasting...

- In a number of cases, given time and financial resource limitations, maintenance of the entire model is not done and only certain steps of maintenance are taken, instead. In that case, the model would not be a good one and the forecasts it produces may be inaccurate.

III. Conclusions on the Role of Macro Econometric Model and Directions for Its Application in Our Country

1) Role of Macro Econometric Model

From the above-mentioned analysis of the experience with macro-econometric model establishment and application at the Department of General Economic Issues, one can confirm that despite limited and incomplete information and data available for modelling work in our country, given what is available, establishing quality macro-econometric models is still possible, and, at the same time, those macro-econometric models are still capable of producing fairly good analyses and forecasts that reflect the tendency of development, and foretell the threats of economic imbalances as well so as to help development policy makers to anticipate them and take initiatives in adopting necessary measures to address them before economic crises occur.

It is therefore possible to affirm that macro econometric model is a strong and well applicable tool in our country.

2) Directions for Macro Econometric Model Application in Our country

(1) At the initial stage, priority should be given to using the macro-econometric model for verifying specific issues, especially for identifying policies and mechanisms in each of the sectors of the economy, and determining the fundamental relationships that are most influential in the eocnomy.

For example, models are required to explain which factors are contributing to economic growth of our country at present? Which factors contribute to our country’s increasing attractiveness to FDI and ODA funds? What are the causes of increased inflation?... Next, the models are also needed to compare Vietnam with other countries to identify the country’s strengths and weaknesses in terms of policies and mechanisms so as to make the most of our strengths and to minimise our weaknesses...

Results of the model should be widely disseminated among the circle of economists  for evaluation, comparision and further improvement.

(2) Only when model users are already skillful in using separated econometric equations or macro-econometric models with very few equations, and, at the same time, are able to identify (agree on) the causes of individual phenomena in our economy, they should move to the second stage of modelling which is linking them into a system while determining general economic theories that provide general explanations for the operations of our country’s economic system, hence establish a forecasting model.

At present, very few staff and organisations are capable of operating this stage despite the availability of a fairly large contingent of staff with knowledge of the model.

(3) To effectively bring into full play the econometric model in our country’s economic analysis and forecasting work, assembling of forces is badly needed to promote collective strength since each individual is unable to establish a quality econometric model.

It is therefore time for the community of econometric modellers to join efforts in establishing the Vietnam Econometrics Association.

At the same time, organisations that need to use the results from econometric models should make further investments in this work and finance the activities of the Vietnam Econometrics Association.