Fiscal policy in the Netherlands in real-time : 55 years of planning and realization

University of Amsterdam

Fiscal Policy in the Netherlands in Real-time

– 55 years of planning and realization –

Mark Walschot ID 5702429

Master’s thesis November 11th, 2009

Supervisor: Dr. M. Giuliodori Second examiner: Prof. Dr. L.H. Hoogduin

Abstract

This thesis studies the planning and realization of fiscal policy in the Netherlands, using real-time data from the annual budget, the “Miljoenennota” (1955-2009). Key results are that (1) there is a systematic negative error between the real-time realized fiscal change and its ex-ante planning. Methodologically, a fiscal framework is best evaluated through the integrated analysis of fiscal changes, outcomes, and revisions. Furthermore, (2) the fiscal plan explains a relatively small part of the realization of fiscal policy changes, casting doubt on the credibility of fiscal plans. Lastly, (3) general economic conditions and the state of public finances are relevant determinants of both plans and errors in fiscal policy. Current fiscal plans compensate for previous errors, and errors in realized fiscal change are persistent. Left-wing partisanship and elections are significant political determinants. Subsequent switches in the institutional setting have led the Dutch fiscal framework to improve significantly. The current year is a substantial exception, as the economic crisis caused the fiscal error to be greater than ever before.

Declaration

I confirm that this is my own work and that the use of all material from other sources has been properly and fully acknowledged.

Table of contents

Abstract...iii

Table of contents ... v

List of tables ...vii

List of figures ... ix

1. Introduction ... 1

1.1. Planning and realization of fiscal policy... 1

1.2. Societal relevance ... 1

1.3. Scientific relevance and methodology... 3

1.4. Research questions... 4

1.5. Summary of results... 5

1.6. Outline of the thesis... 6

2. Literature Review ... 7

2.1. Review of previous empirical work ... 7

2.2. Underlying theory ... 10

2.3. Hypotheses... 12

2.4. The fiscal framework in the Netherlands ... 13

3. Errors in Fiscal Policy ... 17

3.1. Data Set... 17

3.2. Methodology Question 1: Errors in Fiscal Policy... 18

3.3. Results Question 1: Errors in Fiscal Policy ... 20

3.4. Discussion Question 1: Errors in Fiscal Policy ... 29

4. Variability in Fiscal Policy... 31

4.1. Methodology Question 2: Variability in Fiscal Policy... 31

4.2. Results Question 2: Variability in Fiscal Policy... 31

4.3. Discussion Question 2: Variability in Fiscal Policy ... 33

5. Explaining Fiscal Policy... 35

5.1. Methodology Question 3: Explaining Fiscal Policy... 35

5.2. Results Question 3: Explaining Fiscal Policy ... 38

5.2.1 Planning Stage ... 38

6. Discussion ... 47

7. Conclusions ... 49

Bibliography ... 53

Annex ... 57

A1. Abbreviations and definitions (A-Z)... 57

A2. The development of data incorporated in the “Miljoenennota” ... 59

A3. The data set and how to read it ... 60

A4. Abbreviations used in primary data set... 62

A5. Headline fiscal figures: correlation and aggregation... 64

List of tables

Table 1: Summary statistics for planned and realized annual percentage changes (plus errors)... 22

Table 2: Summary statistics for planned and realized outcomes (plus resulting level errors) ... 25

Table 3: Hypothesis tests on first-release revisions of the balance, different sample-periods... 26

Table 4: Interrelations of errors in changes, errors in outcomes, and revisions... 28

Table 5: (Co-) variances of realized-, planned-, and errors in change ... 32

Table 6: Determinants of the planned change of the annual balance ... 39

Table 7: Determinants of the error in the change of the annual balance ... 42

Tables in Annex: Table A.1: Information- and definition-changes in the “Miljoenennota” over the years ... 59

Table A.2: Example of a part of the data set. ... 60

Table A.3: Correlation coefficients of (first-release estimated) level errors... 65

Table A.4: Summary statistics for planned and realized annual percentage changes (1955-1992) ... 66

Table A.5: Summary statistics for planned and realized annual percentage changes (1993-2010) ... 66

Table A.6: Summary statistics for planned and realized outcomes plus level errors (1955-1992) ... 67

List of figures

Figure 1: Illustration of different estimations of the government’s balance ... 19

Figure 2: First-release errors between the realized and planned annual change of the balance ... 20

Figure 3: First-release errors between the realized and planned annual level of the balance... 25

Figure 4: First-release revisions of the annual balance (as conducted in t, regarding t-1) ... 27

Figures in Annex: Figure A.1: Errors between planning and first-release realization of fiscal policy for BB, FB, EB... 64

To my parents

Who unlike most governments Have not yet breached their EU targets

Fiscal Policy in the Netherlands in Real-time

1.

Introduction

Annually, every third Tuesday in September, the fiscal plans of the Dutch government are officially and ceremoniously presented to the public in the “Miljoenennota”. The queen is driven through the political capital of the Netherlands in a golden carriage and the city of The Hague can barely cope with the vast numbers of press and public that try to catch a glimpse of this special occasion. On a total population of 16 million people, a notable 600.000 visit the fiscal planning website that day1. You do not need to be a politically engaged economist to infer what the OECD2 (2002) officially states: “the budget is the single most important policy document of the government, where policy objectives are reconciled and implemented in concrete terms”. There is no escaping the news this day: the new budget plan is out.

1.1. Planning and realization of fiscal policy

This thesis not only focuses on the budget plan. It also looks at the eventual realization of this plan, and analyses the error between the two. Because no matter how relevant the budget plan may appear for Dutch society, in case the presented plans eventually are not implemented in reality, the public ends up with nothing but a non-credible and inconsistent government. The planning and realization of fiscal policy are both part of the budgeting process. This process was described in great detail already over fifty years ago by Bakker (1952). This indicates how the cycle is both absolutely elementary, and historically intertwined with public finances in the Netherlands. A study often referred to in modern research (Von Hagen & Harden, 1995) divides the budget process in the exact same four phases as Bakker did in 1952: (1) a budget plan is made by the government, (2) the parliament approves of the plan, (3) the government implements the plan (realization of the plan) and (4) ex-post control takes place. Following Beetsma et al. (2009), in this thesis the first two steps of the budgeting process are taken together as the planning phase. This planning is then compared to the realization of fiscal policy, in order to obtain the error between the two. Implicitly, ex-post control of the fiscal realization is also measured, since lagged fiscal figures are investigated as well. A major extension of this thesis is that it elaborates on the transition from real-time to ex-post data through subsequent revisions. This method proves highly relevant as revisions of fiscal data are substantial.

1.2. Societal relevance

In general, three dimensions of budgetary policy can be distinguished: fiscal discipline, stabilization of the economy, and allocation of resources (European Commission (EC), 2004). In the presence of an error between the planning and realization of fiscal policy, all three dimensions are negatively affected.

Fiscal discipline, the primary focus of this thesis, is highly relevant as persistent government deficits and high levels of public debt call into question a country’s public finances. This is because too high

1

Source: http://www.sogeti.nl/Home/successen/Website_prinsjesdag.jsp. The number may be positively biased because information stems from the agency that built the website, and also international press visits the website.

2

deficits and debts cast serious doubt on the long-term sustainability of national public policy in the light of ever-growing debt service obligations – which may ultimately crowd out more pressing societal needs considering e.g. infrastructure, education, or health care. This may not be urgent for the current generation, but it undoubtedly will be for the ones that follow. Thus, to achieve fiscal policy that is sustainable in the long-run, fiscal discipline of the government is key (Von Hagen & Harden, 1995). In the current economic crisis this fiscal discipline is ever more relevant (Beetsma et al., 2009): Dutch government revenues are lower due to lower corporate taxes and falling non-tax gas receipts, while expenditures are higher because of expected substantial increases in unemployment benefits. On balance, this results in a deficit that for 2009 is forecast to rise to 4,8% of GDP, breaching the 3%-of-GDP EMU reference value3 (CPB, 2009). Also, the government’s debt ratio is expected to have increased to over the 60%-of-GDP threshold before 2010 (over 80% GDP by 2014). No wonder the government announced consolidation efforts amounting to 0,3% and 0,8% of GDP in 2009 and 2010, respectively (EC, 2009; IMF, 2009). These consolidations are a necessary but painful effort, and merely a plan will not guarantee their execution. This insight guides this research: the budget plan is important for an economy to function properly, but it proves to be useless without its actual realization.

Briefly considering the second dimension of budgetary policy – the stabilisation of the economy – this study is important because the budget plan is an official information flow from the government to the private sector. The budget plan provides the private sector with a signal (e.g. about future taxation), on which the private sector bases its future behaviour. When, as in most current studies, ex-post data are used, this signal is not included and the estimated effects of fiscal policy may be biased (Leeper et al., 2009). In case planned fiscal policy is realized, private sector behaviour is aligned with government action. If however there is an error between planning and realization of fiscal policy, the private sector’s assumptions about government behaviour are wrong, so it cannot rationally or consistently act on the government’s plan. This potentially leads to a destabilization of the economy. It is therefore relevant to know whether an error between planning and implementation of fiscal policy exists and if so, how it could be minimized to make sure that private sector behaviour is well-founded on the government’s fiscal actions. Moreover, since monetary policy in the European Union is now uniformly determined, credible national fiscal policy has become relatively more important (EC, 2004).

To indicate the size of the Dutch public sector, nowadays roughly fifty percent of gross domestic product is spent by the general government (CPB, 2009). With respect to fiscal allocation (the third dimension of budgetary policy), the credibility and consistency of these expenditures are critical, as a government’s budget plan is in fact the financial representation of its underlying goals and policies. For example, a change in health care expenditures as stated in the budget planning, in fact represents a planned health care policy change. Thus, a systematic error between the planning and realization of fiscal policy would indicate that general public policy is systematically off target. In other words: the government systematically does not achieve the goals it strives for (World Bank, 2005).

1.3. Scientific relevance and methodology

Judging from the above societal relevance, it would be rather surprising for this study to be the first to deal with the issues presented. Section 2 of the thesis elaborates in detail on theory and previous empirical work. What differentiates this study scientifically is first that it explicitly divides fiscal policy in various stages, analogous with Beetsma & Giuliodori (2008). By distinguishing between planning and realization of fiscal policy, the crucial divide between the intention of and the actual implementation by the policymaker is addressed, and the policymaker’s behaviour becomes apparent. Second, the study’s primary focus is on the government’s degree of adherence to its previously made planning. A policymaker has to cope with the information set available at the actual decision point, so it is neither fair nor realistic to afterwards judge behaviour on the basis of ex-post data (including previously unknown information). Consequently, this study uses real-time data and puts fiscal changes at the heart of the analysis, instead of focusing only on revised ex-post data and fiscal outcomes. This is to mimic actual practice and behaviour of the policymaker as close as possible (Golinelli & Momigliano, 2009; Koenig et al., 2003). At the same time, the study also recognizes that fiscal outcomes are obviously relevant for the Dutch public, so it examines these too. Moreover, as revisions in economic time series can be substantial, these may play an important role in the real-time estimation of fiscal parameters. As time passes from first-release to lagged estimates, revisions shrink and real-time data turns into ex-post data4. A particularly original contribution of this study is that it analyses the transition from real-time fiscal changes to ex-post fiscal outcomes in great detail. It does so through the examination of subsequent revisions. This way, interrelations between fiscal changes, outcomes, and revisions are fully exposed, and their development over time is laid bare. Third, the data used in this research are unique. A new data set was constructed on the basis of the official annual fiscal policy plans (“Miljoenennota”-vintages) that were issued by the Dutch government over the post war period. The data set thus provides a significant extension compared to previous research on planning and realization of fiscal policy, which covered shorter time periods starting mid-1990s (e.g. Von Hagen, 2008). Fourth, this study focuses on the fiscal framework of one country in particular. The Netherlands has a long tradition of independent economic forecasting. The Central Planning Bureau (CPB) carries out this task since 1945 (from 1961, it even publicly provides independent estimates in its annual “Macro Economische Verkenning” (MEV)). These estimates are aligned with the fiscal projections made by the government, and can function as an extra control. Also, the Netherlands has strong fiscal rules, and its budget formulation process is “on par with the best found in any OECD member country” (Annett, 2006; Blöndal, 2002; Bos, 2007; Von Hagen, 2008). The Netherlands may thus act as a benchmark. Fifth and last, concentrating on one country brings about the advantage of specificity. Although the external relevance of the research is potentially partly abandoned, this specificity is highly relevant. When e.g. panel data from different EU member states is used, homogeneity is assumed with respect to both the planning and realization of fiscal policy (and its determinants) in these countries. However alike EU countries may seem, this homogeneity-assumption may be flawed in practice. The automatic stabilizers are different for each EU country, depending e.g. on size of the

4 _{Given that there are no methodological switches. As revisions may prove particularly large in case of shocks, the}

public sector or progressiveness of the tax system (European Commission, 2005). Also with respect to discretionary action, fiscal policy may be contradictory. The different preferred policies of France and Germany (“Should a fiscal stimulus be provided or not?”) to counter the current crisis provides a case in point5. In the future, studies with a similar methodology may be conducted for each EU country individually. This would result in an overview of country-specific best-practices, providing policymakers all over Europe with useful information about most effective planning and realization of fiscal policy.

1.4. Research questions

The basic framework of this study is captured by the following conceptual equation on fiscal policy:

error

planning

n

realizatio

≡

+

Both the realization and the planning of fiscal policy can be directly observed from the available data. What is unknown is the error. Therefore, the first question concerns the calculation of this error with respect to different fiscal and economic variables in the Netherlands:

1. Is there a systematic error between the planning and realization of fiscal policy?

This first question addresses fiscal discipline. Using real-time data, it judges whether the government has adhered to its planning by comparing the planned fiscal change (from year t-1 to t) to the realized change in fiscal policy. The investigated fiscal parameters are the balance and its components (expenditures (TE) and revenues (TR)) as a part of national income. Examined economic variables are nominal and real national income. Two different errors are analyzed. The error in fiscal changes is the main focus of this thesis. This error explicitly depicts the degree of adherence to planned policies (e.g. how a change in expenditure from this year to the next is planned, compared to how it is realized). Next to fiscal changes, also outcomes are considered. So second, the level errors focus on differences between the planned and realized fiscal results (e.g. of expenditures over GDP) of year t. Both errors can be estimated differently over time (from real-time to ex-post data). Thus lastly, by studying the subsequent revisions of these fiscal parameters, their interrelatedness and development over time are exposed. Section 3 provides the methodology, results and discussion on this question.

After question 1 is answered, all parts of equation (1) are known. Of interest now, are the relative impacts of plans and errors on the variability of the realization of fiscal policy:

2. To what extent is the variability of the realization of fiscal policy explained by the ex-ante planning, and to what extent is it explained by the ex-post implementation errors?

Question 2 specifies the results of question 1, by providing an overview of respective proportions of plans and errors with respect to the variability of the realization of fiscal and economic variables. It

clarifies to what extent the variability of realized fiscal changes in aforementioned variables is due to planned changes, and which part is explained by the errors in these changes. The analysis provides an overview of how some variables turn out as planned (following equation (1) the error would then approach zero), while the realization of other variables can be decomposed into two clear parts: the planned change and the error. In the latter case, the error systematically makes up for a great part of the realized fiscal change. This casts doubt on the value of a fiscal policy plan in the first place, and could indicate the possible strategic use of the planning by the policymaker. The analysis also exhibits the development of estimates over time (i.e. from first-release estimates to third-lag estimates).

Question 1 examines the errors in fiscal policy. Question 2 investigates to what extent the error and the plan make up for the realization of fiscal policy. Subsequently, question 3 of this thesis deals with the determinants of fiscal policy change. To discover which variables explain fiscal changes planned ex-ante, and which variables explain ex-post implementation errors, question 3 asks:

3. What are the factors determining the ex-ante fiscal plan and the ex-post implementation error, considering fiscal policy in the Netherlands?

The goal is of question 3 is to distinguish as precise as possible what parameters contribute in what way to the ex-ante planning and the ex-post errors in fiscal policy in the Netherlands. Therefore, economic, political and institutional determinants are regressed on (i) the planned fiscal policy change and (ii) the error in this change (i.e. the degree of adherence). First, the determinants of the ex-ante planned change are explored. Second, the determinants of the ex-post error in this change in fiscal policy are examined. Analogous to equation (1), these two regression-analyses implicitly uncover main determinants of realized fiscal changes in the Netherlands (thereby contributing to insights on fiscal reaction functions). The underlying theory on determinants of planning and errors in fiscal policy is given in section two, but to give a brief example; plans and errors (and thus realizations) of fiscal policy can be explained by economic- (e.g. growth), political- (e.g. elections), or institutional factors (e.g. fiscal rules). As economic parameters are also included, it is accepted that the government does not have full control over fiscal policy. Given this fact, in the assessment of Dutch fiscal performance it is possible to distinguish between true “bad luck” (by which intended fiscal action has become genuinely impossible to carry out in practice) and (intended) discretionary action of the government. Ultimately, determinants of the ex-ante planned fiscal change and its ex-post errors instantly provide insights on how to improve the fiscal framework. This question is dealt with in section 5.

1.5. Summary of results

The primary assessment of fiscal policy is on the real-time adherence to planned policy change. A systematic negative error is found for the change in the balance (-0,71% national income), caused by upward biases in expenditures (+1,04%) and revenues (+0,36%). These biases subside somewhat when moving from first-release to ex-post data, indicating that revisions are important. A contribution of the thesis is the decomposition of the interrelations between changes, outcomes and revisions. This

reveals that the answer to question 1 depends partly on the research-perspective. When studying ex-post outcomes of fiscal policy (including intermediate revisions) the negative error in the balance disappears, while positive biases in expenditures and revenues remain biased upward. Furthermore, national income is biased upward to a strong degree. Positive revisions of national income appear a driver for fiscal revisions. Cause and nature of revisions have to be studied more closely.

Regarding question 2, the variability in realized fiscal changes is explained largely by co-variances with ex-post implementation errors. This casts doubt on credibility of a fiscal plan and indicates that its use may be strategic. The scope of control of the government regarding its fiscal plan is unknown.

Results on question 3 show the general economic scenario has significant effects on both plans and errors in fiscal policy. The regression on the plan shows a higher debt leads to more ambitious plans, while the previous balance does not have effect. Greater previous errors lead to more ambitious plans (compensatory behaviour by government), while positive revision of the previous balance leads to a less ambitious plan. Gas revenues are important in Dutch public finances. Results indicate that new governments present more ambitious plans (of which less is realized). Fragmentation and left-wing partisanship lead to smaller planned changes. Institutions play a minor role regarding fiscal planning. The overall fit of regressions on the fiscal plan is low. This signals the fiscal plan lacks transparency: it is not explained by predictable indicators but perhaps more by rather imperceptible (political) factors. The regression on the error shows that more ambitious plans lead to greater errors. A better initial balance and a positive revision of the previous balance lead to greater errors. Greater previous errors lead to greater current errors, indicating fiscal policy persistence. A better-than-expected growth scenario leads to a smaller error. There is evidence for an electoral cycle, and left-wing partisanship weakly indicates a greater error. Fragmentation-effects on the error are small. Regarding different institutions in Dutch fiscal history, the error decreased over time, strongly suggesting improvement of fiscal institutions. Useful extensions of this study would be: to investigate determinants of components and subdivisions of the budget in greater detail; to examine revisions and their effects on fiscal outcomes; to put more emphasis on real-time data in fiscal research, and to conduct similar studies for other countries. Particularly in the current economic downturn, fiscal discipline is highly relevant. Comprehensive knowledge on diverse fiscal frameworks could potentially lead to mutual improvement, e.g. through partial adoption of each nation’s distinguishing useful fiscal institutions.

1.6. Outline of the thesis

Section 2 describes previous empirical research and underlying theory on fiscal policy planning and realization. It also provides hypotheses and elaborates on the Dutch fiscal framework. The data set on which the research is based is described at the start of section 3. Sections 3, 4 and 5 present methods and results per research question. Every section ends with a specific discussion of the findings. A short general discussion is given in section 6. Section 7 presents the main conclusions of the thesis.

2. Literature Review

This section describes literature on the planning and realization of fiscal policy. First, empirical work is reviewed. In the process, the originality of this study becomes clear. Second, underlying theory behind the empirical findings is given. Third, the hypotheses regarding the research questions of this thesis are presented. Part four defines various distinctive features of the fiscal framework in the Netherlands.

2.1. Review of previous empirical work

Beetsma et al. (2009) investigate fiscal policy in real-time for a panel of 14 EU countries (1998-2007). The study uses data from Stability and Convergence Programs6 (SCPs), which national governments are obliged to send to the European Commission in light of Europe’s Stability and Growth Pact (SGP). Analogous to an earlier contribution of Beetsma & Giuliodori (2008), the ex-post realized fiscal policy is compared to its ex-ante plan. The authors find both expenditures and revenues turn out higher than planned. As the error in expenditures is larger than the error in revenues, this in sum leads to a realized budgetary adjustment that falls systematically short of planned adjustment. A systematic error thus exists between planning and realization of fiscal policy in the EU. In addition, the variability in the realization of fiscal policy in the EU is explained more by implementation-errors, than it is by the planning of fiscal policy. Concerning the determinants of plans, errors, and thereby outcomes of fiscal policy, the study indicates that economic variables contribute significantly to planning and realization of fiscal policy in the EU. A better initial budget leads to less ambitious plans and vice versa. At the same time, an initial budget that is worse leads to an implementation that is better7. Higher-than-projected GDP growth leads to a better ex-post realization of the ex-ante plan. Political variables play a significant yet relatively minor role in planning and realization of fiscal policy. Although effects are small, elections and left-wing partisanship lead to higher-than-planned expenditure and lower-than-planned revenue. Institutions are an important determinant of fiscal policy. Fiscal rules addressing e.g. the accountability within or transparency of the budgeting process, play a key role in planning and realization of fiscal policy. Also, the strength of the medium-term budgeting framework is significant, as it partially redeems political time-inconsistency problems. The authors conclude that implementation errors with respect to fiscal policy in the EU are systematic, implying that governments (consciously) anticipate the realization-“error” in their fiscal policies. This error is then not attributable to unforeseen economic shocks, but can be assigned to discretionary action following a biased (potentially strategic) fiscal plan. Better institutions lead to a more realistic fiscal planning geared towards fiscal discipline through lower expenditures and higher revenues. Institutions also lead to better adherence to this plan during the realization of fiscal policy.

Extensions of this research with respect to Beetsma et al. (2009) are that (1) a much longer time-period is under consideration so that more than one business-cycle is investigated, (2) as also several

6

These SCPs contain fiscal and macroeconomic projections for the following year and 1-3 years ahead, based on the national annual budget and congruent with roughly attaining budget balance in the medium run.

7

lagged estimates (approaching ex-post data) are incorporated in the analysis, the plan made in year t-1 can be compared to estimations of the fiscal realization beyond still fairly premature first-release data. This way, the definitive error between fiscal plans and realizations becomes visible. Furthermore, (3) the study exposes the transition from real-time to ex-post data by examining the interrelations between the errors in changes, errors in outcomes, and data revisions. Lastly, (4) higher specificity of results is realized. Although Beetsma et al. (2009) control for potential heterogeneity of EU members by including country-fixed effects, the country-specific data set used in this study does not suffer from any unjust cross-country generalization. In panel studies assuming homogeneity among panel-members this is a permanent threat: the use of broader versus narrower samples represents the common trade-off between generalization and specificity of results (Alesina & Perotti, 1996; Heckman, 2001; Maddala, 1999). Although this study overcomes the possibly unjustified assumption of cross-country homogeneity, it may suffer from another generalization bias. This is because homogeneity in the relationships in fiscal policy (and determinants) over time is assumed. Tests on stability of results and parameter constancy are conducted to rule out this potential threat.

Concerning methodology and focus, the above study is almost fully aligned with this research. Only slightly less analogous is a study by Von Hagen (2008). Also based on annual SCP data (1998-2004) of 15 EU member states, the study focuses on projections for GDP growth and government budget balance, revenue and expenditure. It compares these projections to ex-post data instead of to first-release data (thereby ignoring the real-time character of fiscal policy making), and finds that the fiscal plan is often biased and inefficient with respect to eventual outcomes. The author thus also finds a systematic error between planning and realization of fiscal policy in the EU. Furthermore, this study too finds that budgetary institutions have a great impact on fiscal planning, errors, and realizations. Particularly, a contract-based mode of fiscal governance (as opposed to delegation to a finance minister), and stringency, enforcement and commitment to numerical fiscal rules significantly reduce the realization error. An important remark is that in this respect, the contract-based fiscal framework of the Netherlands is considered among the strongest in the EU. This implies a tendency to produce cautious fiscal forecasts and display relatively high fiscal discipline. Considering political determinants, the author finds that in election years the planned budget balance is systematically better than its ultimate realization, indicating an overly optimistic forecast. This thesis departs from the work of Von Hagen (2008) by that it focuses on one country in particular, it explicitly distinguishes between the planning and realization phase of fiscal policy, it uses real-time data and revisions from a unique data set, it extends the time period under investigation, and lastly by that it considers planned annual changes and their errors – instead of examining only the level differences of various parameters.

Cimadomo (2008) investigates real-time data of 19 OECD countries (1994-2006). Because the author expects the political bias to be too large in these documents, not the government’s fiscal plan is used as the primary data source. Instead, the allegedly more realistic, filtered OECD’s Economic Outlook estimates are studied. Even so, the author finds the ex-ante planning of fiscal policy to be counter-cyclical while its ex-post realization turns out to be pro-counter-cyclical, resulting in a systematic error between

the two. The main explanatory variable is the output gap. Although elections are found to have a weakly significant effect with respect to overspending, political and institutional determinants are not considered in detail. Also here, the Netherlands displays exceptionally accurate future projections of fiscal and economic parameters. Lastly, the authors conclude that only in the planning phase, fiscal policy along the business cycle is asymmetric. When, for example, public debt is above a certain threshold, governments are more concerned with public finances than usual, and thus plan a more fierce consolidation. In analogy with the above reported error between planning and realization however, the author does not find this asymmetry when testing the ex-post realization of fiscal policy.

Pina & Venes (2007) study 15 EU countries and use data from the Excessive Deficit Procedure (EDP) notifications (1994-2006), containing fiscal forecasts in light of the EMU. The authors find negative and positive systematic errors between planning and realization of fiscal policy. They largely contribute the errors to economic and politico-institutional determinants. Higher commitment in fiscal governance and stronger expenditure rules lead to more cautious forecasts and a smaller error. Elections are found to have an increasing effect on the error, while partisanship is not. All effects became stronger since the implementation of the SGP. The Netherlands displays relatively cautious fiscal forecasts.

Golinelli and Momigliano (2006) examine the EU in real-time (1987-2005). They focus mainly on the determinants of fiscal policy, and do not specifically test for an error between planning and realization. Considering explanatory variables such as the state of public finances (primary balance, debt) and cyclical conditions (output gap), the authors find fiscal policy is stabilizing and counter-cyclical, both when studying fiscal plans and ultimate realizations. This alignment suggests an error is in fact absent, although differences between the two phases are not explicitly shown. It is stressed that results may differ greatly depending on the type of data used (real-time or ex-post). Elections are found to have a significant loosening impact on fiscal policy, conditional on a favourable economic environment.

Marinheiro (2008) uses real-time panel data (1999-2006) of 14 EU countries, and does again show a deviation between the ex-ante plan and ex-post realization of fiscal policy. Contrary to others testing fiscal policy in view of the business cycle (e.g. Annett, 2006; Cimadomo, 2008; Golinelli & Momigliano, 2006), this author remarkably finds pro-cyclicality in discretionary fiscal planning, while finding counter-cyclicality in the ex-post realization. A review of 26 empirical studies on fiscal policy and counter-cyclicality of the business cycle in EMU countries, reveals that the nature of this interrelationship is truly unsettled (Golinelli & Momigliano, 2009). The authors show that, depending on research choices (e.g. applied fiscal model; use of ex-post or real-time data), empirical results may vary greatly. Partly because of this “non-consensus”, the response of fiscal policy to the business cycle is not the main focus of this thesis (even though growth is controlled for in conducted regressions). Mentioning a few studies that do treat the subject remains relevant, since these studies exemplify the existence of an error between ex-ante plans and ex-post realizations of fiscal policy, which is indeed the central aspect of this thesis. Marinheiro (2008) does not find a significant influence of either political fragmentation or partisanship, while elections and the number of spending ministers (SMs) both lead to higher budget deficits.

Schaltegger & Feld (2009) study a panel of Swiss cantons8 (1981-2001) and do not specifically focus on the separate phases of planning and realization of fiscal policy. Also, ex-post data is used. What makes this study relevant nevertheless, is that it analyses political and economic determinants of fiscal discipline. The authors find the initial economic situation of a government matters for fiscal discipline: consolidation is more likely to succeed if the deficit in previous years is high and GDP growth is weak. This implies that governments do indeed act when the need for fiscal reform is highest. Considering political variables, the authors find that more political fragmentation leads to higher fiscal discipline, leading to better long-run fiscal sustainability. Nevertheless they also find that large parliaments exhibit less fiscal discipline. This provides some evidence for a common pool problem (see paragraph 2.2).

Similar to the level of analysis of Schaltegger & Feld (2009), Goeminne (2008) examines fiscal policy of Flemish municipalities (1992-2002). A systematic error between planned and realized government revenue is found. The author explicitly studies the effect of political fragmentation on planned revenue: higher fragmentation leads to more cautious forecasts, potentially leading to lower overall debts in the long-run. Hallerberg et al. (2007) implicitly take the deficit bias for granted, and look particularly at the role of institutions (fiscal rules; forms of governance) regarding fiscal discipline in the EU (1985-2004). They find differentiation in “institutional best practices”: the budget process of one-party governments would benefit most from strong delegation to finance minister, while countries with a coalitional and ideologically more dispersed government (such as the Netherlands) would benefit more from stringent contracting. Also, evidence is found for an electoral cycle. Noteworthy is Lewis (2009), who finds that when comparing real-time and ex-post data for central and Eastern Europe (1995-2008), no difference in fiscal policy is found regarding cyclicality of the business cycle. Both methodologies indicate a-cyclicality. Concerning politics, only fragmentation is found to effect fiscal policy (in a loosening way).

Concluding, there is not one study encompassing the time period under investigation in this research. Also, this thesis elaborates not on aggregated panel data but on a unique data set covering a single country. Furthermore, the division of fiscal policy in its distinct phases is still relatively uncommon. Combining the use of real-time data and ex-post data, and exposing the transition from the one into the other through revisions, is distinctive. Lastly, the simultaneous investigation of the effects of economic, political, and institutional variables on the planning, error, and realization of fiscal policy as applied here is extensive. Although possibly too far-reaching, this broad view may prove inspiring to the empirical framework in country-specific fiscal policy assessment.

2.2. Underlying theory

This part summarizes underlying theory with respect to the above empirical findings. The ex-post realization of fiscal policy is often not aligned with its ex-ante planning. What is the explanation for this error? A few reasons are mentioned here. Firstly, the realization of fiscal policy may depend largely on

8

Accordingly, this study thus does not investigate country effects, casting doubt on its comparability with the current research. However, all 26 Swiss cantons are entirely free to determine their fiscal policies. As such, the impact of economic

implementation-lags, as years may pass before planned fiscal measures take effect (e.g. Cimadomo, 2008). Yet, governments are expected to take this generally accepted notion into account and incorporate potential implementation-lags into their budget plans. Therefore, a critical assessment of fiscal realization versus planning is still justified. Second, the error may arise because of common pool problems in public finance: through general taxation, the costs of a certain public policy are dispersed over the entire population. However, benefits of this policy are concentrated on a specific group or interest. This entails that, relative to the dispersed costs, the concentrated benefits are more strongly advocated by the spending ministers representing these specific groups in the population. As a result, only the finance minister senses an incentive to constrain overall public expenditure – while spending ministers representing the concentrated interests do not (Annett, 2006; Von Hagen, 1995; Wehner, 2008). Common pool problems may manifest itself both in the planning and realization of fiscal policy (e.g. Beetsma et al., 2009; Von Hagen & Harden, 1995). Ex-ante, spending ministers pressurize the finance minister for higher allowances, either in parliamentary discussions prior to the formulation of the budget, or in the transition from this initial plan to its approval in the budget law (e.g. Koopmans & Wellink, 1978). Also ex-post, during realization, governments may give in to political pressures to loosen their fiscal stance (e.g. stemming from economic circumstances: extra stimulus is ‘needed’ in bad times, dampening the economy is ‘not necessary’ in good times) (Golinelli & Momigliano, 2009). A third explanation for the error is political myopia. A country’s debt (and the deficit contributing to this debt) is permanent, while politicians governing the public finances only last their elected terms. This brings about time-inconsistency problems. Politicians are inclined to postpone painful reform for their successors, while stimulating the economy during their own term (e.g. overspend to gain popularity). This is especially relevant in the light of elections (e.g. Annett, 2006).

Theory on determinants of fiscal plans and errors states that these may be economic, political or institutional. The most prominent and innocent reason why an error between planning and realization of fiscal policy would occur in the first place, is that the future is uncertain. The realization of a plan is never under a government’s full control. Deviation may not only stem from discretionary action, but also from truly unexpected (economic) shocks, such as e.g. the current economic crisis. These genuinely unexpected shocks call for short-term flexibility of fiscal policy (consider the stimulus-packages used to counter the crisis). This thesis abides this, by incorporating economic growth as a main determinant of fiscal plans and errors. However, it still holds that, in case errors are found to be systematic, a government cannot hide behind this “general policy uncertainty” (Wyplosz, 2005).

What parameters are expected to influence fiscal outcomes? Several factors influencing the planning, realization, and errors of fiscal policy are given here. Economic determinants such as expected output and government debt may influence the realization of fiscal policy compared to its planning. When a macroeconomic variable such as national income growth is projected overly optimistic, this irrationally provides the promise for higher future revenues and lower necessary expenditures. Both seemingly improve the budget balance. The same analogy holds for lower debts, and subsequent lower required expenditures on interest payments. In case overly optimistic projections are strategic (thus systematic)

the projection will, as a result, turn out not to be realized after the year has passed. The government may then attribute the resulting “unexpected” deficit to “unforeseen economic circumstances”. When accountability is lacking, this leaves the “benevolent-but-unlucky government” and its factually non-credible budgeting policy untouched (Alesina & Perotti 1996; Jonung & Larch, 2006). Not only does this lead to flawed expectations of the public, it also endangers overall long-run sustainability of fiscal policy. This strategic use of fiscal planning could also be applied to meet external budgeting criteria, such as the EMU-guidelines of the Stability and Growth Pact on deficits and debts – institutions that may thus also influence the error. Another reason for fiscal realizations to be off target may be that in the budget process “creative accounting” occurs (ignored in this study). Institutional determinants logically originate from all the traits of (and flaws in) the budget process. Any reform, numerical rule, or guideline affecting fiscal policy that can credibly get the political temporality out of the more perpetual character of public finances, is expected to improve the realization of fiscal policy with respect to its initial plan. Examples are independent forecasting bodies, measures improving political accountability, or the strengthening of possibilities for legal enforcement (Hallerberg et al., 2007; Wyplosz, 2005). A few political determinants are: electoral uncertainty causing over-stimulation, fragmented coalitions failing to contain the budget, proportional electoral systems leading every party to “want a piece of the pie”, a high number of SMs pressurizing the finance minister, and short average government duration, by which accountability is minimized. Also left-wing partisanship is often associated with overspending. A final example is that governments may annually install a multi-year planning of expenditure-cuts, and then revise or postpone it again in the following years – merely promising fiscal reform but never truly executing it. Many political factors thus stem from common pool and myopia-problems (Alesina & Perotti, 1996; Annett, 2006; Von Hagen, 1995; Koopmans & Wellink, 1978).

2.3. Hypotheses

First and foremost, the reviewed empirical studies are panel-studies covering several countries or regions over a relatively short period. Also, theory on the topic is quite generic. Therefore, the tentative generalization of the above results to this country-specific research covering a longer time span could turn out to be unfounded. Nonetheless, the similarity between the studies is deemed sufficiently large to credibly formulate some hypotheses. Most studies examining the deviation between planning and realization explicitly, do find an error. Therefore, the main hypothesis to be phrased from the above exploration with respect to the first research question is that a systematic error between the ex-ante planning and ex-post realization of fiscal policy can indeed be expected. Whether the error manifests itself in the overall balance or in its components (expenditures and revenues) is to be examined. To the best of my knowledge, the development of errors in change and outcome over time (from first-release to third-lag estimates), and their interrelation with subsequent revisions has not been studied before. Regarding the second question, the partition of the variability of the realized change in fiscal policy into its initially planned change and an implementation-error is only examined in detail by Beetsma et al. (2009). It seems too resolute to extrapolate the findings of one EU-wide study to this country-specific research, in order to postulate a firm hypothesis on question 2. Numerous answers to

the third research question of the thesis are possible, since many studies focus on determinants of fiscal policy. The broad view of this study obliges to incorporate economic, political and institutional factors determining fiscal policy. Important economic variables are the state of the economy (measured by national income growth) and the shape of public finances (measured by balance and debt). Elections, political fragmentation and partisanship are the most relevant political determinants. The institutional setting considers fiscal rules and various modes of governance. A brief remark is that the Netherlands is considered to be a country with a transparent and strong fiscal framework. It has a long history of independent economic and fiscal forecasting, and its coalition governments engage in contract-based budgeting processes. These institutional, economic, and political traits may contribute to the sound planning and realization of fiscal policy (leading to a smaller error), exemplifying that this country might prove the exception to the fiscal policy rule9.

2.4. The fiscal framework in the Netherlands

According to the IMF and the OECD, the fiscal framework in the Netherlands is rather unique and its design and implementation are considered highly recommendable (Bos, 2007; Debets, 2007). Annett (2006), Blöndal (2002), Cimadomo (2008), Pina & Venes (2007) and Von Hagen (2008) also consider its budgeting process as one of the best among OECD and EU member states. Why is this the case? Over time, the Netherlands show to be a frontrunner considering process, efficiency, transparency and accountability of its fiscal framework. The early recognition of the significance of all phases of the budget cycle in Bakker (1952), described in the introduction, is an example of this10. Also, the Central Planning Bureau (CPB), the Dutch independent economic forecasting body, came into force already very early in 1945. This agency not only provides macroeconomic forecasts on e.g. expected GDP growth, it also calculates whether governmental policies and their costs are correctly presented. Stronger still, the Dutch parliamentary opposition parties traditionally hand in an “alternative budget plan” to counter the government’s plans, and even these documents pass through the CPB for fiscal and economic verification. The CPB is thus fully integrated in the Dutch budgeting process (Bos, 2007; Debets, 2007). More recently, again two innovative fiscal policy reforms took place. The accounting system operation of 1985 evolved against the overspending and high deficits of the 1980s and sought to decentralize the government and stimulate public commitment (in action until 1993). Policy oriented budgeting, a 1999 reform, focuses on transparency and accountability11. Being output-based instead of guided at (monetary) inputs, actual results of the government’s plans and actions are examined. Within the government, spending ministers are individually held responsible for their programmes and use of resources, possibly leading to smaller common pool problems (Debets, 2007; Nispen, 2006). For a historical perspective on the development of the Dutch fiscal framework, see Box 1.

9

Next to science, there is off course the general frugality the Dutch are renowned for internationally. Perhaps the “going

Dutch” mentality, according to which each person participating in a shared activity pays for herself, is entrenched even

throughout governmental spheres, thereby having a positive influence on fiscal discipline…

10

Bakker (1952) even mentions a fifth phase in the budget cycle; after ex-post internal and external control of the annual account have been completed, the Dutch legislative force ultimately ascertains and finalizes the account.

Box 1: Broad history of the budgeting process in the Netherlands (Summarized from Postma, 2006)

After World War II, the Netherlands was engaged primarily in rebuilding its economy. Budgeting policy was very strict and fell under the austere supervision of a dominant minister of finance, Mr. Lieftinck. An extraordinary man1, he was known for an extensive control of the budget. At the same time, in September 1945, the CPB started to be involved in the budgeting process. With Mr. Tinbergen (later to be the first to win the Bank of Sweden Prize in Economic Sciences) at its head, this independent macroeconomic forecasting agency thoroughly checked and verified all policy plans in the Netherlands. In the 1950s, the Keynesian view of an active government with respect to the economy became common practice. Anti-cyclical policy was key and the role of the finance minister was less pronounced. This policy with the business cycle at its core, proved to be difficult to carry out in practice, and lost its relevance with emergent internationalization. In the 1960s, a more

forward-looking view was adopted in the “structural budgeting policy”, based on high economic growth. The decade was

characterized by an enlargement of the Dutch welfare state, with matching public expenditures2 _{on social welfare and health} care. In response to this, finance minister Mr. Zijlstra adopted a top-down approach to the budget: first overall expenditure was determined and second every spending minister could request his part. The growing complexity of public finance forced the former centralized approach to be abolished; the budget system was decentralized while accountability and transparency were strengthened to ensure budget discipline. In 1973, the first oil shock meant a tipping point for Dutch public debt (that had been declining since WWII). Unemployment and economic stagnation meant a further deterioration of public finances. The danger to the long-run sustainability of public finances had by then been fully recognized, and a commission (Dutch: “Studiegroep Begrotingsruimte”, founded in 1971) was given the task to conquer the worsening of public deficits and debts. “Economy, efficiency, and effectiveness” were guiding principles throughout the 1980s, when serious contractionary reforms were conducted and strong rules concerning national debt and deficit were put into place. Subsequently, public expenditures and the deficit fell. Debt rose for a longer time though, and peaked in 1993 at around 78% of GDP. Discipline was thus still key. Every segment of the Dutch government was held responsible (as it already had been since 1970) for its own financial balance, with its respective spending ministers being held personally accountable. Close watch was explicitly kept when “exogenous” factors were pleaded for to have caused overspending. In the meantime, the Maastricht treaty of December 1991 and its EMU debt- and deficit norms came into place in view of Europe’s unification. The government’s debt fell, and from 1994 the budgeting process was guided by trend-based fiscal policy that deliberately took a cautious mid-term perspective into the future. Mid-1990s, the parliament criticized the cabinet on the lack of transparency of its budgeting and justification procedures. This led to the adoption of the above mentioned policy oriented budgeting reform in 1999 and the start in 2000 of a formal and public control on the annual fiscal planning3_{. The current crisis and accompanying fiscal stimuli} have again put public finances in the Netherlands under great pressure.

_____________________________

1

Next to the fact that the bilateral negotiations with his spending ministers took days (Postma, 2006), a more extravagant example of his determination is that once on a World Bank mission, he pulled a gun on a group of robbers in the middle of the Turkish mountains in order to protect his secretary from getting hurt… Without eventually having to fire the gun, he saved the day (Ziegler, 1987). 2

In 1950, public expenditure was 28,9% of GDP. In 1970, it was already 44,1,%. It peaked in 1983 with 60,9%. 3

In Dutch: “Gehaktdag”; every third Wednesday in May.

To recap, since about 1994 the budget in the Netherlands is determined on the basis of trend-based economic assumptions with a medium to long term forward focus. Four other main institutional periods can be distinguished in the development of Dutch fiscal policy: from 1945 to 1952, the government’s central (and thus fiscal) policy was to rebuild the Netherlands. Within this framework, the view was that the budget should be balanced. Also, the government should be limited and should not interfere too much in the economy. In contrast, the period of 1952 to 1959 was characterized by Keynesian policy: the “manageable” economy called for stabilization by anti-cyclical government policy. Proving hard to

put into practice, Keynesian policy was abolished in 1959. At that time, structural budgeting became the general vision (until 1982). This approach had a medium-term focus and was built upon the notion that the Dutch economy would grow by about five percent each year. By 1982, the indispensable viewpoint had become to regain control of public finances (that had drastically deteriorated in the meantime). The nearly unsustainable deficits and debts, built up by the ever-growing government, had to be reduced. This policy was in place until 1993. Throughout all five periods, common features are that excessive debt is not acceptable and that taxes should be low to stimulate entrepreneurship and economic growth – the Keynesian period being an exception to the latter guideline (Bos, 2007; Debets, 2007; MinFin, 2009). Of interest is whether these periods and institutional perspectives can be identified from the data, and if so: how they compare over time. Does the error between the planning and realization of fiscal policy in the Netherlands differ significantly over time, with the adoption of different fiscal rules and institutions? Section 5 addresses this question.

Next to the more institutional perspective described above, there is also politics. The Dutch political system is typified by coalition governments: no political party enjoys a majority in the parliament12. Although the budgeting process is contract-based (anticipated to counter overspending), the coalition set-up might make political fragmentation a more relevant determinant of fiscal policy. Also, the effects of elections and political partisanship are examined.

In between institutions and politics lies the general criticism on fiscal policy; whenever rules exist, people try to evade them. Even when these rules are installed by the people themselves. There is no government with only “benevolent dictators”. So, there is off course a (political) flipside to all the seemingly cutting edge Dutch budgeting procedures. Taking a closer look, for example the recent output-based budgeting system has led politicians to formulate targets in an increasingly abstract fashion. To an extent that even in retrospect it can hardly be verified whether targets have been accomplished or not. To cite Dutch public finance specialist Mr. De Kam about the opaqueness of the fiscal planning for 2009: “In light of the information age, it has become ever more important for policymakers to conceal unwanted information from the public and wordily twist any potential adversity into something positive” (De Kam & Ros, 2009). This once again indicates that the mere adoption of sound fiscal institutions is not enough (Debets, 2007; Nispen, 2006). And even though the CPB’s forecasts are neutral, they too may prove to be off target or systematically biased13 (e.g. Kranendonk et al., 2009). In the following sections this speculation ends and it will become clear how the fiscal framework of the Netherlands has truly performed over the past 55 years.

12

New governments announce their major policy objectives in the multi-party coalition agreement, formulated at the start of the government’s four-year term of office. Although this agreement already involves broad budgetary consent, the annual budget planning (“Miljoenennota”) is financially more detailed and serves as the main policy document in this research.

13 _{Next to general uncertainty, another explanation for this bias is that the Netherlands is a small and open economy. As such,}

3. Errors in Fiscal Policy

This section first describes the data set that is used in this thesis. Second, it gives the methods and results regarding the first research question. The section ends with a discussion of the findings.

3.1. Data Set

The primary source used to construct the data set of this research is the “Miljoenennota”. This is the annual fiscal policy plan of the government in the Netherlands, holding detailed information on fiscal and economic variables. Real-time data were collected from vintages from 1945 to 200914. Regarding the explanatory variables, next to the “Miljoenennota” also the Comparative Political Data Set 1960-2007 was used (CPDS; Armingeon et al. (2009). This is a collection of annual political, economic and institutional data for 23 OECD countries. This study only uses political data on the Netherlands. The CPDS code book was used to extend the data to match the longer sample in this research. Inflation figures (and an extra control for the other parameters) were taken from the independent CPB’s annual macroeconomic explorations (MEV’s), also in real-time. Following close inspection of the raw data, several outliers were excluded. Before 1955, a few subsequent unexplained and very large revisions occurred concerning the balance, revenues, and expenditures. Given their size (around 8% of national income) it can be excluded that genuine fiscal information updates caused these revisions. More likely, financial transactions or definition-changes of the parameters under examination laid at their origin15. Therefore, these values are not incorporated in this study, as they do not resemble fiscal policy and make a valid comparison over time impossible. Consequently, the final sample is 1955 to 2010. An extensive description of the data set and an explanation on how to read it can be found in the Annex.

A general remark regarding the data, is that the key parameter reflecting “fiscal policy in the Netherlands” is the government’s annual balance. Three historical headline figures of this annual balance were aggregated into one single parameter (BAL). A justification of this choice is given in Annex A5. The balance is given by BB/NNI from 1955 to 1992, by BB/GDP from 1993 to 1998, and by EB/GDP from 1999 to 2010. As the aggregated parameter BAL is generated on the basis of all headline measures and covers as many observations as possible, all questions regarding fiscal policy in the Netherlands are answered with the highest possible precision. The aggregation in BAL does not unrightfully extrapolate present findings to past times or vice versa: it leaves various fiscal frameworks intact while merely creating a larger frame of reference. Conclusions on the planning, realization, errors, and revisions in fiscal policy are drawn in a way that represents the policymaker’s historical perspective at all times. To test for its robustness, this aggregated framework is decomposed again later into its original elements.

14

Data prior to 1945 is less relevant for the purpose of this study because the assumption of homogeneity in fiscal definitions and relationships over time may prove faulty. Excluding pre-WWII data thus aims to prevent a potential generalization-bias (par. 2.1). The budget has been published on paper since 1906. Before, it was since 1806 presented in speech (Postma, 2006).

15

3.2. Methodology Question 1: Errors in Fiscal Policy

This part deals with the question “Is there a systematic error between planning and realization of fiscal policy?” Both the ex-ante plan and different estimates of the ex-post realization can be directly observed from the data. To obtain the error, equation (1) can then be shifted to form equation (2):

planning

n

realizatio

error

≡

−

The ex-ante budget plan for year t remains unchanged after it is published in year t-1. Therefore, the calculation of the above error fully depends on which estimate of the ex-post realization is used. These range from a first-release estimation made in the current year t, to the more definite lagged figures published in “Miljoenennota”-vintages of years t+1, t+2, and t+3. The crucial guideline in the choice of the primary error is the real-time character of fiscal policy decision making. Policymakers base their budget plan for the next year on fiscal and economic estimates of the current year. Hence, first-release estimates are key in the assessment of policy plans. Consequently, the primary error in this research is the difference between the initial budget plan and the first-release estimate of the realization of the balance. This first-release error in the balance may be given either in changes or in outcomes. This thesis describes both, but the error in changes is considered to be the most relevant. The difference between the two errors is best explained with the help of a formula (Beetsma et al., 2009a):

(

) (

)

(

) (

) (

)

−

−

=

−

−

−

−

t

BAL

BAL

BAL

BAL

BAL

BAL

BAL

BAL

Error in change Realized change Planned change

This is the detailed version of equation (2). Box 1 below explains the double indices. Equation (3) shows annual fiscal changes. It states the first-release error in the change of the balance equals the difference between its realized and planned change (this holds by definition). A first-release estimate of the realized change in the balance is made in year t. Then, estimations are made for the realization of the balance in both the previous year t-1 and the current year t. Thus, realized change is calculated by subtracting the current estimate of the balance of t-1

(

BAL

)

(

BAL

)

(

BAL

)

(

BAL

)

Double indices are used to indicate which estimate is considered exactly. The superscript refers to the year in which the

estimation is made, while the subscript indicates the year to which the estimation refers. For example,

BAL

BAL