Empirical Models and Policy Making

Interaction and Institutions

Mary Morgan, Frank den Butter

• Publisher: Routledge
• Year: 2003
• eBook ISBN: 9781134573127
• Edition: 1
• Topic: Negocios y empresa
• Subtopic: Negocios en general

This collection, written by highly-placed practitioners and academic economists, provides a picture of how economic modellers and policy makers interact. The book provides international case studies of particular interactions between models and policy making, and argues that the flow of information is two-way.

Part I: The Tinbergen tradition

Chapter 1: The relevance of economic modelling for policy decisions*

G.Zalm

The development of macroeconometric models is influenced by policy makers and vice versa. In this chapter the situation in this field in the Netherlands is sketched. The CPB National Bureau for Economic Policy Analysis plays a special role. While in other countries various institutes and ministries evaluate economic policy and prepare forecasts, in the Netherlands these functions are concentrated within the CPB. This approach holds some (scale) advantages, which should be used to make progress in the evaluation of new problems with which policy makers are confronted.

JEL classifications: A11, C5

1. Introduction

The information flow between economic policy makers and economic model builders is not a one-way flow. The requirements and questions of policy makers play an important role in the development of economic models. It is not only the information that flows, but also people who flow. I have crossed the bridge between empirical modelling and policy-making several times, so I can touch upon my own experience with respect to the interaction between macroeconometric modelling and policy-making.

First, I will sketch briefly the major economic problems of the twentieth century and their influence on macroeconomic modelling. It is interesting to notice that economic problems, such as the oil crises, the deterioration of public finance and long-term unemployment have led to fundamental changes in macroeconometric models. Second, I describe the use of macroeconometric models for policy evaluation. In the Netherlands, policy makers have relied heavily upon macroeconometric models. Third, I will talk about the influence of forecasts in the process of policy-making and fiscal policy. Finally, I give some suggestions for future research.

2. The importance of modelling for policy evaluation

2.1. Using models for policy evaluation

The most well-known disruptive economic problem in the twentieth century was the Great Depression. In the 1930s Western economies experienced massive unemployment and greatly reduced incomes. In the year 1935, in the Netherlands more than 10 per cent of the working population was unemployed, and real GDP was 25 per cent below its 1930 level. From the start of the Depression a theory, a model or a structure was needed, to offer possibilities that could reduce the economic hardship so many people faced. Confronted with the question of developing policy instruments to reduce these economic problems, Tinbergen (1936) developed the first macroeconometric model of the Dutch economy in 1936. Later he also prepared macroeconometric models for the US and UK economies. The efforts of Klein (1950) and Stone and Stone (1939) followed Tinbergen’s model. After the Second World War, these Keynesian models increased in scale. The advances made in computer technology and better economic statistics improved the scope for developing models in this tradition.

However, in the 1970s, the inadequacy of these Keynesian models to deal with the large structural changes linked to the oil crises shook the trust of policy makers in these models. The increase in oil prices lead to double-digit inflation and rising unemployment and that clearly shook the confidence of the belief in the traditional Keynesian trade-off between inflation and unemployment. In a critical evaluation of the existing type of modelling, Lucas (1976) argued that conventional macroeconometric large-scale models were fatally flawed and were not useful for the policy debate. Economists and policy makers turned their heads to the supply side. It was realized that structures that were developed were far from ideal and blocked a continuation of a high growth path. Gradually, also by model builders, more emphasis was put on endogenizing the supply side of the economy. For example, it was demonstrated that a rise in real wages exceeding the rate of technical progress would increase unemployment. In the Netherlands this model extension has been very important for the acceptance of the policy of wage restraint (e.g. Den Butter, 1991).

In the 1980s, the structural problems had been worsened by the deterioration of public finance. In Western Europe this had been caused by the inability to trim the welfare state. In the Netherlands the deficit rose from almost zero in the beginning of the 1970s to almost 9 per cent of GDP in 1983. In the United States the combination of a policy of cutting taxes and increasing public spending was responsible for the rise in the budget deficit in the beginning of the 1980s. Moreover, this combination of a loose fiscal policy with a tight monetary policy to fight inflation led to a rise in the interest rate and the dollar. To deal with these problems, new macroeconometric models with a fully-fledged description of monetary sectors with an endogenous portfolio behaviour of the private and banking sectors were developed. These so-called monetary blocks have not always been successful. Although Ministers of Finance in a lot of countries were quite successful in bringing budget deficits down, they got little support from model builders. Nevertheless, almost all economists will agree that the future of the Dutch economy is better off with the present budget deficit than with the deficit of 1983.

Currently, the economy is in a better shape than in the 1980s. However, we still face some serious economic problems. One of the most severe problems is the duration of long-term unemployment, especially amongst low-skilled workers. As is typically the case in Europe today, the Dutch unemployment rate in this area is more than double the overall rate. The functioning of the labour market in its connection with the social security system is held responsible for this large share. Therefore a model with a detailed description of the tax system and the various labour-market institutions was required and the applied general equilibrium framework seemed appropriate.

Despite its incapability to predict large economic shocks or to deduce the structural flaws of the economy, the use of macroeconometric models in policy design is quite common and largely undisputed. Especially in the Netherlands, we have a very long tradition of relying upon technical expertise from macroeconometric models as a guide in macroeconomic policy formation. In 1936 the publication of the first macroeconometric model by Tinbergen (1936) was accompanied by the simulation of several policies and the effect on employment and the current account. In the post-war period different generations of this model have played a role in policy evaluation.

By experience, the idea that there should be one model for all problems has been abandoned. Instead, a large variety of economic models has been developed to cope with the broad range of different policy questions. This is a very natural development that Tinbergen certainly would have appreciated. A physicist by origin, in his early days nuclear scientists were looking for one model for the atomic nucleus. Nowadays, there are many models describing different aspects of the atomic nucleus. Economic science shows a similar development. At the CPB National Bureau for Economic Policy Analysis different models are used for different pur poses. Donders and Graafland (1998, Chapter 2 in this volume) give a historical overview of (macroeconometric) model development at the CPB.

2.2. Fiscal policy and policy evaluation

In the use of macroeconometric models for policy evaluation, simulations to assess the effect of fiscal policy have been very important. Keynesian models have been focusing on the level of government expenditures and taxation. The so-called conjunctural-structural models have been used to investigate the impact of different tax rates on wages and, nowadays, detailed tax proposals are discussed with the help of a model.

At several points in time, macroeconometric models have been used intensively to investigate the effect of a particular policy under consideration. There are many examples to give. In 1992, the report of the committee for green taxes, the so-called Wolfson committee, was guided by a thorough investigation of the employment effects of a shift between labour and green taxes and the impact on the competitive structure of industries. Similar efforts have been made with the Quest-model of the European Commission and the Green-model of the OECD.

Sometimes an outcome of a model creates its own policy rhetoric. In 1992, the CPB published an applied general equilibrium model, called MIMIC (Gelauff and Graafland, 1994). One of the doubtful outcomes was that a progressive tax system encourages employment. It even led to suggestions for a ‘Robin-Hood’ policy of raising the tax rate of the last bracket and lowering the tax rate of the first bracket. However, in these applied general equilibrium models, it is very hard to model the consequences of a more progressive tax system on training, work intensity and the desire for promotion and, therefore, the adverse effects of a progressive tax system can be underestimated. The model does not seem adequate for analysing such questions. Fortunately, policy makers base their judgements on more than the outcome of a single model.

2.3. The Dutch case of models and policy analysis

The institutional setting in the Netherlands differs from that in other countries. A central role has been laid down by law for the CPB, which is an independent organization within the Ministry of Economic Affairs. The CPB has the obligation to prepare every year a Spring forecast, the so-called Central Economic Plan. Since its founding, the CPB has expanded its tasks and nowadays it also puts a lot of time into policy analysis.

In the Netherlands, models, especially those from the CPB, have played an important role in policy analysis. As I have worked both with the Ministry of Economic Affairs and the CPB, I will shortly comment on the special relationship between the two as far as model-based policy analysis is concerned. At some Ministries, and at the Ministry of Economic Affairs in particular, own models were developed in reaction to the view that adjustments in CPB models to new circumstances were, in certain cases, slower than policy makers would like. This has not much to do with laziness or the monopoly position but with the high standards of the CPB. Models developed at Ministries, I know from experience, can be quickly fixed if and when it serves the policy objectives of the ministry. So, at the Ministry of Economic Affairs a number of model exercises have been conducted, and as a director of the department doing those exercises I have always supported it. However, I have never seen results published from the research at the Ministry of Economic Affairs which gave ammunition to the Ministry of Social Affairs or the Ministry of Finance in case of conflicting interest.

Besides, if the Ministry of Social Affairs or Finance were to develop their own models, the Ministry of Economic Affairs would rather rely on the CPB than on the models of the other ministries. This does not mean that the CPB work is beyond criticism and the government policies should only be based on it. The CPB has been criticized by independent sceptics, but also by economists at the various ministries.

The chapter of Donders and Graafland (1998, Chapter 2 in this volume) shows that important developments have occurred in the work of the CPB while, at the same time, there is a clear consensus of the limitations of models. Therefore I whole-heartedly agree with the new stress on international comparative and qualitative institutional studies. The role of the CPB as the economic conscience for government policy should not be overstated. Nevertheless, I just hope that the balance between international and scientific reputation on the one hand, and policy relevance on the other hand, will be kept, and that Tinbergen’s idea about mission will continue to lead us. There is no use in reputation if it is not useful for improving policies.

2.4. Forecasting and policy-making

In the process of policy-making the forecasts of a macroeconometric model are still of relevance, although, as I explain later, its relevance in our country has decreased as the CPB always wanted. In the Netherlands, the yearly Budget Memorandum in September is accompanied by the yearly forecast of the CPB—the so-called Macro Economic Outlook—and the Memorandum is based on these insights. Technical information from other agencies such as the Ministry of Finance or the Ministry of Social Affairs are taken into account in order to come to a forecast. In addition, the Nyenrode Forum for Economic Research publishes their forecast, but until now the differences were rather small. The same holds for the secret shadow forecasts of the Central Bank.

In other countries a number of different organizations or independent research institutes publish annual forecasts. For example, there are seven different competing economic forecasts in the United Kingdom (e.g. Whitley, 1994). It should be noticed that the key features of these models differ substantially. In Germany, five institutes are engaged in forecasting, leading to a cry for a consensus forecast.

There has been a controversy about the usefulness of forecasting in the policy debate. Some economists argue that they do not have any faith in the forecasts of macroeconometric models (e.g. Gordon, 1984). In his Tinbergen lecture, however, Klein claims that in most cases during recent decades the forecast performance of macroeconometric models has been improved considerably and compared with their alternatives, such as naive models, time-series analysis, or judgemental forecasts, they do reasonably well (e.g. Klein, 1988). Nevertheless, forecasts should be treated with caution. Point estimates of important variables such as percentage change in GDP and unemployment are always surrounded by rather large confidence or prediction intervals.

Don (1993) argues that there are several sources of unpredictability due to uncertainty in policy and non-policy exogenous variables and mis-specification in the model. Especially for a small open economy as the Netherlands, the short-term forecasts are highly conditional on international developments and to a lesser degree on domestic policy measures (Van den Berg, 1986). One of the possibilities for dealing with uncertainty about the exogenous environment is to present forecasts in different scenarios, for example with respect to the guilder/dollar exchange rate or the growth of world trade.

To illustrate the changing role of forecasting, a few words on Dutch fiscal policy. In the 1960s, fiscal policy was seen as an instrument for stabilizing the economy. In practice, this so-called structural fiscal policy consisted of two parts. First, the acceptable budget deficit was calculated for a cyclically neutral base year. Second, given this net amount of borrowing, the yearly additional budgetary resources were defined as the trend growth of tax revenues. The calculation of these additional budgetary resources required forecasts about trend GDP and trend tax elasticities.

After the recession of the mid-1970s, there was a tendency to overestimate trend GDP and public finances deteriorated. A major change in fiscal policy was unavoidable. Fiscal policy was concentrated on following a yearly rule to lower the budget deficit. Accordingly, economic forecasts became even more important. Every change in economic forecasts influenced the ex ante deficit and led to a fiscal reaction in order to re-establish the deficit target. It is clear that such an approach is unsatisfactory and could sometimes work pro-cyclically.

Therefore, in 1994, the present coalition announced a new type of fiscal policy: the so-called trend-based fiscal policy, where the budget deficit may change in order to absorb non-structural deviations in revenue. To create room for cutting the deficit and tax cuts, the central government has set ceilings for real expenditures from 1995 to 1998. Economic forecasts derived from a cautious scenario were used for calculating the budgetary resources for the government period. Hereby, the need for repeated adjustments of the budget due to changing economic conditions is reduced. So, fiscal policy nowadays takes into account the downward risks of economic forecasts much more. The life of a Minister of Finance has become easier, as have the lives of his colleagues.

3. Macroeconometric modelling in the twenty-first century

If we look back at more than half a century since Tinbergen published his first macroeconometric model, the widespread applications and the usefulness of models for policy-making seem an achievement to me. A good example for the Netherlands is that political parties also rely heavily on the scenarios and the effects of changed policies as calculated by the CPB. This leads to an unusual discipline in political programmes as well as to the comparability of the different programmes. It is impossible to get away with grand promises that are impossible to realize.

For model builders, there are still many avenues to explore. First, the modelling of market behaviour is still rudimentary. More emphasis should be put on imperfections in financial and goods markets. Hereby, one could perhaps rely more on the appealing results of the industrial organization literature. In order to assess behavioural responses to government policies, the focus should be changed from a macroeconomic orientation towards a microeconomic orientation. Second, many current policies and policy proposals are aimed at improving the basic technological and ecological infrastructure of the economy. Therefore, it seems challenging to model the driving forces behind the process of economic development, such as infrastructure and technology, and I encourage the current efforts. For these issues the insights from the endogenous growth literature can be useful. Third, the effects of the ageing population in the years to come should be analysed in more detail in long-term models. Fourth, given the overwhelming problem of long-term unemployment, the efforts should be continued on modelling the effects of the welfare state and the institutions on the labour market. Finally, I could mention the problems of debts and deficits. But if I was a director of the CPB I would not concentrate on that as I am quite sure this will be taken care of effectively with or without models to support it. Th...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• List of illustrations
• Notes on contributors
• Preface
• Acknowledgements
• Part I: The Tinbergen tradition
• Part II: The variety of experience
• Part III: Model products and model usage
• Part IV: The interaction process and institutional arrangements
• Part V: Empirical models and policy-making