Generational Accounting in Latvia:
Time for Action
GENERATIONAL ACCOUNTING IN LATVIA: TIME
FOR ACTION
Master Thesis
Olaf J. de Groot Groningen, July 2005
Supervisor: Prof. Dr. Jakob de Haan
University of Groningen, Faculty of Economics
Abstract
Table of Contents
Abstract 3 Table of Contents 4 1 Introduction 5 1.1 Outline 7 2 Background 9 3 Literature Review 113.1 Introduction of Generational Accounting 12
3.2 Criticism 15
3.3 Results from previous analyses 21
4 Methodology 23 4.1 Theoretical methodology 23 4.2 Practical application 25 5 Data 29 5.1 Demographic variables 29 5.2 Social variables 30 5.3 Fiscal variables 31 5.4 Economic variables 34 6 Results 36 6.1 Generational imbalance 36
6.2 Dangers if the imbalance is not solved 37
6.3 Policy implications 38
7 Alternative Scenarios 40
7.1 Results under different assumptions 40
7.2 Policy implications 42
8 Conclusion 44
Literature 46
Appendix 1: Tables 49
1. Introduction
The twentieth century has been an era of population growth. Global population has increased from 1.6 billion in 1900 to 6.2 billion in 2000, as a result of declining mortality rates. After 1960, the decline in mortality was followed by a strong decline in birth rates in many developed nations. As a result of this, population growth rates decreased strongly in many, if not all, developed nations. However, due to the particular features of the population growth spurt countries have gone through, there is a relatively large group of people in their middle age right now: the so-called baby-boom generation. This group is getting older and as a result, the average age is increasing in many countries. When this group is nearing their retirement age, this may have a profound impact on public finances. After all, these new retirees are eligible to receive pension payments, contribute less in taxes and need relatively large amounts of healthcare.
People have only recently started to realise the potential gravity of these effects. As the requirements of elderly differ from those of younger people, ageing can be expected to have a serious influence on the way our society is organised. In addition to the effects for the structure of our society, the implications for the finances of the public sector may be even more profound. However, determining the precise effects of ageing on public finances is complicated.
Determining the extent to which ageing will actually take place is the easiest part of the problem. Population projections have been made by national as well as international agencies (e.g. United Nations, 2005), using long-term prognoses of population determinants, such as birth, mortality and migration rates. Even though these projections are subject to rather wide margins, they all show that the population will indeed age strongly and elderly
dependency ratios1 will continue to increase. In Europe, elderly dependency ratios are
expected to more than double from 0.20-0.25 to 0.45-0.60 between 2000 and 2050, before settling on a long-term rate of 0.40-0.45. When determining the effect of ageing on public finances, most authors have focused on the question whether current policy can be continued indefinitely. Research concerning this so-called fiscal sustainability is very important, because it can say something about the long-run appropriateness of immediate policy change. A number of different methods have been applied in the past, with varying success.
Another important question regarding the way fiscal policy should handle the ageing problem is whether the burden of ageing is fairly distributed amongst different generations. It
1 The elderly dependency ratio is defined as the number of elderly divided by the number of people in the
can easily be seen that a pension system in which pensions are paid out of the contributions made by current workers (pay-as-you-go or PAYGO systems) would result in large payments that future workers will be required to pay when the elderly dependency ratios increase. However, the standard accounting techniques used in the field of public finance do not take these problems into account. In fact, current techniques actually hide some of the problems
that exist in current societies, by attaching particular labels to certain flows of money2.
As a solution to the problems described above, Auerbach, Gokhale and Kotlikoff (1991) have developed a completely new technique to address government accounting. This technique is known as Generational Accounting and is mainly concerned with the evaluation of sustainability and intergenerational fairness of fiscal policy in the long run. The method of Generational Accounting determines the average individual net tax burden for each generation. For the representative individual from each generation a so-called generational account is presented, which includes the Present Value (PV) of all future taxes it is expected to pay minus the PV of the future transfer payments it can expect to receive from the government. Summing all generational accounts of living generations will result in the net payment made by currently living generations. The PV of payments of living generations will be available to pay for the PV of government consumption minus the net wealth of the government. Typically, after subtraction of these payments, there will be a significant burden left, which has to be borne by future generations. Using some simple mathematics, it is easy to compute the average growth-adjusted burden that will rest on future generations if the government is to remain solvent. If future generations have to pay more than newborns from the current generation, an adjustment of current fiscal policy will be necessary to maintain
intergenerational fairness3. The size of the adjustment reflects the degree to which future
generations are better or worse off than their currently living peers under current policy arrangements.
Generational Accounting has been applied in many countries and is actually used for policy determination in some of these, such as Japan, Norway and New Zealand. However, most of the analyses have been applied to industrialised OECD member states. This is unfortunate, because there are a lot of other countries which could potentially benefit from the application of this technique.
2 Chapter 3 will show an example of how labelling may have a pronounced influence on fiscal balances under
the current accounting system.
3 Intergenerational fairness is achieved when future generations are confronted with the same net tax payment as
In this paper, Generational Accounting will be used to study the consequences of
ageing in Latvia. Latvia is a very interesting country in this respect, because the ageing
problems are quite strong, even compared to other countries. Contrary to many other countries, ageing will not only lead to a higher elderly dependency ratio, but also to a very significant population decrease. This is due to the fact that the decrease in Latvia’s birth rates has been more pronounced than anywhere else. Additionally, Latvia is currently suffering
from a significant budget deficit, despite the fact that economic growth is in fact very high4.
Ageing could severely aggravate these problems. The final feature that makes Latvia an interesting country to evaluate is the fact that it is a country with a lot of opportunities: GDP growth in Latvia is the highest in the European Union and there seem to be few obstacles ahead for this small country, as long as European integration continues the way it has done so far. All in all, the question addressed in this study is whether the current fiscal policy stance in Latvia is sustainable in the long run and whether the intergenerational distribution of fiscal burdens is fair.
1.1 Outline
The next chapter contains a short description of Latvia and the Latvian economy. The third chapter consists of an extensive literature review concerning Generational Accounting. An introduction about the methodology is followed by a discussion of criticism that has been put
forward by different authors. Finally, thischapter will also contain an overview of the results
and policy implications of Generational Accounting exercises done in other countries. Cross-country studies published by the European Commission (1999) and Auerbach, Kotlikoff and Leibfritz (1999) will be the most important sources of information.
Chapter four contains an extensive explanation of the methodology of Generational Accounting. In this chapter, the mathematical derivation of individual accounts will be explained as well as the calculations that are necessary to reach appropriate conclusions. The fifth chapter presents the extensive set of data necessary to perform an appropriate
Generational Accounting exercise. These data concern demographic, social, fiscal and other
economic variables.
In chapter six, an overview is given of the results from applying Generational Accounting to Latvia. In addition to the outcomes and their interpretation, a solution for any potential imbalance is also discussed. Of course, the chapter will also draw attention to what happens if current fiscal policy will actually be continued. Chapter seven contains similar
analyses as presented in chapter six, but with different scenarios, based on other assumptions. Of course, using different assumptions will lead to other outcomes and thus to different policy implications.
2. Background
The small (64,589 sq km) Baltic nation of Latvia gained its renewed independence in 1991, when it seceded from the Soviet Union. Its first democratic elections, held in 1993, were a great success, but the changeover from socialism to capitalism has not been trouble-free. This had led to a near-complete failure of the Latvian banking system in 1994-95 and a severe economic crisis. All in all, real GDP per capita decreased by about 48% between 1990 and 1995 (Summers, Heston and Aten, 2002). In the later nineties things started to look brighter for the Latvian people, when Latvia was invited to start accession talks with the European Union in 1999. In 2004 Latvia, as well as nine other countries, joined the European Union. The big steps that are taken forward now have led to a vibrant economy with great potential. As a result, economic growth has been extraordinary. Between 1996 and 2004 the growth rate of real GDP has averaged 5.9% (Ministry of Finance, 2004). Of course, it should be noted that Latvia remains the poorest country in the European Union, with a 2004 GDP per capita of only 43.2% of the EU-25 average (Eurostat, 2005). According to the neo-classical growth theory (Solow, 1956), this should create the possibility of a catch-up effect. Even those who
only believe in conditional convergence5 (e.g. Barro and Sala-i-Martin, 1992) should be
aware that Latvia has come into very favourable economic waters, with the stability of being part of the European Union and the export opportunities created by the Common Market. In fact, the conditional convergence theory dictates that convergence should take place between countries that are in similar situations. It is not difficult to realise that the Latvian situation differs little from the situation of countries like Finland or Denmark, which are both small open economies within the European Union. Another similarity between the Baltic and Scandinavian countries is related to the similarity of their social and demographic structures (Katus, Puur and Põldma, 2003). Of course, before convergence can take place completely, the structure of the Latvian economy will need to change, but this is a process that will take a number of years.
Socially, Latvia is going through a big change as well. Obviously the entire nation has been reshaped during the transition from socialism to capitalism. Some of those changes will be permanent, whereas others may be temporary. One thing that will hopefully have been temporary was the surge in emigration that followed the secession from the Soviet Union. Changes that are more likely to be permanent include a decrease in fertility rates, which were low to begin with. As a result, the Latvian population size has decreased from 2.7 million in
5 The conditional convergence theory states that low-income countries will indeed have a large growth potential,
1990 to 2.4 million in 2000. Estimations concerning 2050 vary between 1.4 and 2.0 million, with 1.7 million as the most likely scenario (United Nations, 2005). At the same time, the elderly dependency ratio is expected to increase from around 0.25 in 2000 to over 0.55 in 2050.
Relative to GDP, the Latvian government has seen both its revenue and its expenditure remain fairly stable. In absolute nominal terms, income and expenditures have increased by approximately 30% between 2000 and 2003. On the income side, social security contributions, personal income taxation and Value Added Taxes (VAT) make up nearly 80% of the total revenue. The expenditure side is mainly concerned with Social Security payments and the costs of education. All in all, between 2000 and 2003, the reported budget deficit was mostly between 1.5% and 3%, which is rather good compared to other new EU-members (Ministry of Finance, 2004a). However, one should be aware that past budget balances may have benefited from high growth rates, even though the structural budget balance remains poor. Unfortunately, structural budget balances are notoriously difficult to determine and can therefore not be reported easily. The level of government debt is very low, at around 15% of GDP. Again, this is relatively low compared to other EU member states. Inflation, finally has been between 2.0% and 3.0% from 1999 to 2003. This has also led to a fairly stable exchange rate and the Latvian Lats (LVL) was worth €1.56 in 2003 (Eurostat, 2005).
3. Literature
The concept of Generational Accounting was founded by Auerbach, Gokhale and Kotlikoff (1991), who presented it as an alternative to traditional government accounting. In their view, traditional deficit accounting methods are open to strong criticism due to some of the arbitrary assumptions underlying them. The classic example, taken from Auerbach and Kotlikoff (1987, p. 104), refers to the labelling of social security contributions. Labelling contributions as ‘taxes’ and pension payments as ‘transfers’ deserves no fundamental preference over labelling contributions as ‘borrowing’ and pension payments as ‘return of principal plus interest’, corrected by an old-age tax or subsidy. Changing these definitions does, however, have a huge impact on the reported budget deficit. Another example elaborated in nearly all Generational Accounting literature refers to a random citizen John, who transfers $100 to the government and receives $100 plus interest a year later. These transfers can be labelled in a number of ways, none of which deserve any particular preference. The results for the government budget balance and the level of government debt are very different, though:
i) The government borrows $100 from John and the year after, it repays the principal
plus interest, totalling $105.
ii) The government levies a tax of $100 on John and the year after it makes a transfer
payment of $105.
iii) The government borrows $1,000,000 from John and makes a transfer payment of
$999,900 in the first year, while the year after, it taxes John for $1,049,895 and repays his loan plus interest, amounting to $1,050,000
The effect of these three actions would be the same (John has $100 less in the first year and $105 extra in the second), but the effect on government debt and the budget balance differ strongly. In model i), the government balance will be 0 in the first year and -$5 in the second. Model ii) will lead to a government balance of +$100 during in the first and a balance of $105 in the second year. The extreme example is model iii), where the government balance will be -$999,900 during the first year and +$1,049,895 during the second. Government wealth will finish in balance, but differs strongly during the first year. Model i) would lead to -$100 government wealth, model iii) leads to -$1,000,000 end-of-year wealth and model ii) will have a zero-balance during both years.
liabilities mainly concern government promises to make future payments, such as pensions payments. Future pension payments are not recorded in a PAYGO system, even though it is
unlikely for the government to renege on their promise to pay them6. Of course, if a country
has a funded pension system, the future payments are not included in official government statistics either. But in such a system, people save for their own pensions and there is thus an additional stock of wealth, which is also unrecorded. These factors should balance each other, which makes the consequences of a funded pension system less troublesome than of a PAYGO system.
3.1 Introduction of Generational Accounting
To solve the issues associated with standard government accounting methods, Generational Accounting was developed. It ignores the labelling of different items on the government budget and takes into account the intergenerational transfers that (will) take place. In a nutshell, it can be described using the following formula, where NPV is the abbreviation of Net Present Value:
+ + =
This simple-looking equation, which will be discussed more thoroughly in chapter 4, states that total inter-temporal government wealth (i.e. the present value of all future income flows plus the current stock of wealth) should be equal to the inter-temporal government obligations (i.e. the present value of all future outflows). If not, the No Ponzi Game (NPG) condition, which says that governments must remain solvent, would be violated. This is simply impossible since investors would realise this and would thus no longer want to hold debt of the country in question. It also shows that there is no such thing as a free lunch: decreasing net taxes for currently living generations immediately has to be accompanied by either i) a decrease in future government consumption or ii) an increase in the taxes paid by
future governments7. Calculating the NPV of taxes paid by living generations involves setting
up individual accounts for representative members of each generation. These representative
6 Another example of the arbitrary nature of deficit reporting is given in Auerbach, Gokhale and Kotlikoff
(1994), who describe that the Congressional Budget Office publishes a whole range of budget deficits, to suit everyone’s needs. The 1992 US deficit is reported as $201 billion, $290 billion, $292 billion or $340 billion.
7 Of course, the third option is an increase in government wealth. This may seem unlikely, but is actually
possible. For instance, if a previously undiscovered natural resource, such as oil or natural gas, is uncovered, this can be included in government wealth and thus warrants a decrease in net taxes.
accounts are then multiplied with the size of each generation to determine the total net payment made by currently living generations.
The value of the generational account of unborn citizens is calculated as a residual needed to balance the previously mentioned equation. After the necessary balance has been determined, it can be split over all people who will be born in the future. The payment attributed to these people is adjusted for growth, i.e. the individual payment will increase at the same rate of productivity (and thus, the capacity to pay). This will spread the necessary payments fairly over all future generations.
Using this methodology, Auerbach et al. (1991) construct individual generational accounts for each living generation in the United States in 1989. In addition, government wealth, which must include both officially reported debt and the value of government possessions, is calculated. Finally, to evaluate current fiscal policy, future government consumption is calculated under the assumption that current policy will be maintained indefinitely. Finally, this leaves the NPV of the combined required payments that have to be paid for by future generations.
It must be noted, however, that the NPV of tax payments by currently living generations is based on their remaining lifetimes. It obviously makes no sense to compare the remaining lifetime payments of a newborn with those of a 30-year old. The first will benefit from huge transfer payments, in the form of education, whereas the latter has already received these in the past. In addition to that, the income taxes that will be paid by the newborn are discounted by an additional thirty years, which means that, at a 5% discount rate, these will amount to only a quarter of the 30-year old’s payments. In fact, the only realistic comparison that can be made is one between a newborn and a representative member of the future generations. After all, these accounts will both include lifetime benefits and costs. Auerbach et
al. (1991) thus calculate that a 1989 newborn male will pay a net $73,700, while future
generations will receive a net tax bill at birth of $89,500. This implies that future generations are worse off by $15,800 or about 70% of current annual GDP per capita. In addition, the study continues with a large number of different scenarios based on alternative assumptions, as well as various policy options that could close the gap between newborns and future generations.
conducted by the European Commission (1999), Auerbach et al. (1999) and Kotlikoff and Raffelhüsschen (1999). Two things become very clear when comparing the different findings of Generational Accounting exercises. First, there are only a limited number of non-industrial countries for which Generational Accounting exercises have been performed. In fact, serious attempts nearly always cover Western Europe, North America, Japan, Australia or New Zealand. The only notable exceptions are Argentina, Brazil and Thailand, which are included in Auerbach et al. (1999) and Kotlikoff and Raffelhüsschen (1999). Central or Eastern Europe is never used as a subject for performing Generational Accounting exercises.
The second major issue that can be noticed in the different reports is the fact that the conclusions vary strongly. As an extreme example, one can take Sweden. According to the European Commission (1999), the difference between the net payments of 1995 newborns and future generations equals an extraordinary 628% of GDP. On the other hand, Kotlikoff and Leibfritz (1999) report a negative difference amounting to -205% of GDP. This difference is even more remarkable when realising that both studies use 1995 as their base year. Quite surprisingly, part of the difference is explained by alternative labelling methods, even though the concept of Generational Accounting was supposed to solve that particular problem. There is, however, still considerable debate whether expenses such as education and child benefits should be classified as general government consumption or as transfers. Moreover, even if
they are considered as transfers, which is usually the case8, the question remains to whom
these transfers should be attributed. Should child support, for example, be attributed to
children or parents? Another explanation for the dissimilar conclusions of different authors
may be found in the definition of government wealth. For example, Auerbach, Gokhale and Kotlikoff (1994) sum up all general government deficits between 1900 and 1991, whereas the European Commission (1999) adds up, for the base year, the national debt and the PV of net
income of government-owned enterprises9.
Finally, an important source of debate concerns the underlying assumptions that are being used. These assumptions can be related to the data, such as demographic developments or to generic assumptions about economic states. Regarding demographic developments, for instance, the European Commission (1999) has assumed that no demographic changes take place after 2010, to reconcile the data from all countries. Kotlikoff and Leibfritz (1999) use
8 A notable exception is Auerbach et al. (1991), who classify education as government consumption. Later
analyses use double calculations, with different classifications of educational expenses (e.g. Kotlikoff and Leibfritz, 1999). Recent publications nearly always classify educational expenses as transfer payments.
9 The European Commission does realise this is a very ad hoc estimation. Of course, the input-oriented pricing of
full-scale long-term population estimations instead. Auerbach et al. (1991) already show that using alternative growth and discount rates also leads to large differences in the results. Increasing the future growth rate from 0.75%, which is their baseline scenario, to 1.5% enlarges the gap in net payments between newborn males and males of future generations from a growth-adjusted $15,800 to $21,000. Changing the assumed discount rate from 6% to 3%, results in an increase from $15,800 to $45,900 in the case of men. The original results of
the scenario analysis presented by Auerbach et al. (1991) are shown in table 110. This table
clearly indicates that higher growth rates and lower interest rates both result in larger generational imbalances. Another interesting phenomenon that can be derived from table 1 is the fact that female generational accounts are significantly lower than male accounts. This is partly due to the fact that females have lower labour participation ratios and thus pay less income tax. In addition, women have higher life expectancies and thus benefit from old-age welfare for a longer period of time.
3.2 Criticism
Obviously, when something new is proposed in economic science, there are people who disagree. Strong criticisms have been expressed by Bohn (1992), Haveman (1994) and Buiter (1997). These and other authors have identified a number of major objections against the use of Generational Accounting. Auerbach et al. (1994) refuted a lot of the objections, although recognised that some may indeed have some merit. In this section the different points that have been brought up will be discussed.
The first point of critique, discussed by Haveman (1994) and Buiter (1997), deserves some merit. The criticism concerns the incidence of particular taxes, as well as government expenditures. In the case of taxes, this refers to those taxes that are not distributed to certain ages, but averaged over the entire population. The more important problem lies with government expenditures. More particularly, a lot of government expenditures are not attributed to certain ages or generations, even though their benefits may in fact fall upon only certain generations. An example could be expenditures on environmental restoration. This restoration process may be a long-lasting expense, with benefits accruing only to future generations, due to the long time it takes for full restoration. The second problem in the attribution of government expenses is the disparity between the treatment of government income and government expenses. Nearly all taxes are attributed to generations in one way or another, whereas government expenditure is only in part attributed to generations. Haveman
(1994) considers this a sign that generational accounting assumes that government expenditures on public goods do not generate any utility. Due to this disparity between the treatment of government income and expenses, generational accounts tend to be more negative than if government consumption would be averaged out over all generations as
well11. The arguments by Haveman (1994) are countered by Auerbach et al. (1994). The
authors acknowledge the fact that it would be a good idea to attribute all government expenses to specific generations, but see no possibility of doing so. It is impossible to calculate the exact amounts of utility derived by different generations of public goods, such as national defence. The only semi-public goods for which this can be done are healthcare and education. The other point concerning the differing treatments of government income and expenses is not relevant. After all, averaging out residual government consumption over all generations would only have a level effect. The generational imbalance between current newborns and future generations will remain.
Buiter (1997) argues that “the usefulness of generational accounts…lives or dies with
the validity of the life-cycle model” (p.606). The life-cycle model, introduced by Modigliani
and Brumberg (1954) says that consumption at time t does not depend on income at time t, but depends on life-time income. During the first part of life, when income is relatively low, the consumer will therefore dissave. During mid-life, the consumer will repay his loans and save for the future, when expected income will decrease again. This way, the consumer is able to construct a stable consumption path, even though income actually varies. A simple representation of the life-cycle model is shown in figure 1. Buiter (1997) identifies the following base assumptions for the life-cycle model: i) finite individual lives, ii) no Ricardian intergenerational bequest motive and iii) complete markets. Obviously condition i) is met, but condition ii) has been a source of criticism before (Bohn, 1992 and Haveman, 1994). The problem lies with the fact that when there is a Ricardian bequest motive, the planning horizon of individuals becomes larger than their own lifetime. In other words, individuals derive utility from the utility of other generations. This would have serious consequences for tax policy. After all, assuming that the individual members of the society are fully rational, they would realise it when they were living on the credit of future generations. Thereby generationally unbalanced finances would lead to increasing bequests. A government that borrows heavily to stimulate the economy will be met by consumers who will reduce their
11 As could be seen in the first simple explanation of the generational accounts, on the right-hand side of the
spending to compensate future generations. This renders fiscal policy impotent and makes it useless to calculate generational accounts, according to Buiter (1997). Auerbach et al. (1994) respond to this criticism regarding the Ricardian bequest motive by quoting a number of studies which have found little or no evidence of any compensating private intergenerational transfers that offset government policy. In addition to that, even if the transfers would take place, generational accounting would still be worthwhile. It would then measure the amount of private transfers needed to offset government policy. This would also be interesting, because such private transfers carry a deadweight utility loss, due to transaction costs, and should thus be minimised. Condition iii) mentioned by Buiter (1997) concerns the
completeness of markets12. The critique related to market completion is the opposite of the
problems of the Ricardian bequest motive. The latter leads to planning horizons beyond individuals’ life times, whereas the first leads to planning horizons that are shorter than individual life times. After all, for generational accounts to be constructed, one would have to be able to make valid present value calculations, which will be difficult if markets are incomplete. Based on Bohn (1992), Buiter explains that under certain conditions, inter-temporal changes in government spending may influence the utility of individuals, despite the fact that generational accounts are left unchanged. However, the criticism by Bohn and Buiter can relatively easily be countered by adding that their model is simply different from the Auerbach-Kotlikoff model used in Generational Accounting.
A real and significant point was made by Haveman (1994) and concerns the disparity between fiscal allocation and fiscal incidence. The Generational Accounting literature assumes that taxes should always be attributed to those groups that they are fiscally allocated to. A major example in this case is the incidence of wage taxation. Generational Accounting attributes the full weight of this tax to individual workers, whereas economic theory shows that part of this weight will be borne by employers, and thus by capital owners. The issue of allocation also strongly occurs when families are introduced. Family spending on food (and the VAT and excise taxes related to this consumption) is often attributed to the income earner of the family. However, one could argue that part of the VAT paid by a family unit ought to be attributed to the children, who consume out of the family income. The issue of tax incidence in small open economies, as well as closed economies, has been discussed by Fehr and Kotlikoff (1997). Their conclusion refers mainly to the incidence of capital taxation and not. However, a number of small government inflows, which do not come from the population, such as foreign aid and taxes paid by foreign residents, are included in the government consumption account.
12 Buiter adds that market participation may be incomplete, as dead and unborn people are unable to participate
states that in the case of a closed economy and in the absence of transaction costs, capital taxes will be borne by capital owners. However, with the occurrence of transaction costs, or the use of a small open economy, the burden may shift away from capital owners and towards workers. The authors add that the existence of transaction costs is not firmly established, but conclude that methodological adjustments to Generational Accounting should be made in the case of small open economies. Tax incidence of other taxes than capital taxes must be estimated. Unfortunately it is simply not possible to estimate all the specific elasticities that would be necessary to be able to fully determine the exact incidence of each and every individual tax. However, Fehr and Kotlikoff (1997) show that generational accounts reflect utility changes of different generations very well.
Haveman (1994) has also voiced the critique that Generational Accounting does not include the behavioural changes that would occur in a General Equilibrium framework as a result of fiscal policy. The previously mentioned Ricardian bequest motive is not included and neither are any demand or supply elasticities. According to the author, Generational Accounting exercises always propose solutions that are totally out of order, because they do not include any of the feedback effects. Suggestions like those of Sartor (1999) that Italy should increase its income taxes with nearly 200% obviously do not take into account that such a tax increase would have incredible repercussions for the willingness to work, as well as the amount of tax evasion. The inclusion of such effects may actually necessitate tax changes that are either smaller or larger than the suggested numbers. Auerbach et al. (1994) counter Haveman’s critique by saying that Generational Accounting is only a method of evaluating
current policy. The outcomes simply state the generational imbalance that is in place. The
numbers that are often included in Generational Accounting exercises are only included to show how large the necessity is of a change of fiscal policy. The most important statistic coming out of the analysis remains the difference between newborns and future generations in the baseline scenario. Finally, it should be noted that Auerbach and Kotlikoff (1987) have argued that the macroeconomic feedback effects are often small. Auerbach et al. (1994) add that these effects nearly always exacerbate the generational imbalance that occurs.
would imply that in forty years time, a specific tax would be levied on those aged less than forty. This is obviously impossible. The explanation why the argument of Haveman is actually incorrect is similar to the explanation in the previous paragraph. Generational Accounting can only be used to evaluate the current policy stance. The conclusion that policy is generationally unbalanced simply implies that current fiscal policy should be changed. In addition to that, the actual policy suggestions that are made, do in fact include the influence these policy changes have on current generations.
Bohn (1992) tries to prove that the concept of Generational Accounting is only valid when governments possess the ability of levying lump sum taxes. He claims that when using the 1987 Auerbach-Kotlikoff 3-period life cycle model, one can only have irrelevance of labelling when non-distortionary taxing is available, i.e. lump sum taxation. Drazen (1992) on the other hand, argues that the possibly distortionary features of taxation have nothing to do with the relevance of Generational Accounting. He explains that Generational Accounting simply recognises that not only the classic measure of debt is relevant, but also the future commitments, as represented by the Present Value of government consumption and the value of the generational accounts of currently living generations. The question whether taxes are distortionary is simply unrelated to this. The author continues by explaining where the confusion comes from. He explains that Ricardian Equivalence achieved when non-distortionary taxes are introduced are one way to encounter confusion with regards to labelling, but certainly not the only one.
Auerbach et al. (1991), for example, only ranges between 2.5% and 7%, which Haveman (1994) says is as arbitrary as choosing rates varying from 1% to 10%. Haveman’s final discounting issue is recognised by Auerbach et al. (1991, 1994) as being a real problem. However, the difficulty of choosing different discount rates for transfer payments and taxes as well as private and public payments, is very large. Additionally, the resulting difference will most likely remain small. Auerbach et al. (1994) accept that the use of a single discount rate is an oversimplification, but disagree with Haveman about the size of the problem. With respect to possible inter-temporal changes of the discount rate,
In a similar vein, Haveman (1994) claims that the entire concept of Generational Accounting is based upon a myriad of assumptions that are open to question. In a way, the author has a good point, as projections about the future are always the result of scenarios deemed most likely. No-one is able to foretell the future, but that does not automatically mean we should not try. Population forecasts are never perfect, but they have been constructed to be the best we can come up with. Other assumptions are always open to debate, but when assumptions can reasonably justified there is no reason why Generational Accounting exercises should not be undertaken. It is, however, very important to take into account the uncertainty involved in the predictions. But these are always reflected in the alternative scenarios (with differing growth rates or different population forecasts) presented in nearly every Generational Accounting exercise. Finally, Haveman forgets that standard budget deficits also comprise assumptions and expectations. Reservations for uncertain future liabilities are also simple expectations, but he does not have a problem with that.
3.3 Results from previous analyses
An example of how serious Generational Accounting is taken nowadays can be found in the publication of European generational accounts by the European Commission (1999). The
European Commission shows that for 1995, most of the members of the European Union13
suffer from generational imbalances. The sizes of the imbalance vary strongly between the different countries. Ireland is the only country with net inter-temporal wealth, or in different words: the only country where future citizens are better off than newborns. At the other extreme, however, Finland’s future citizens are worse off by a stunning 797.9% of GDP per capita. To be more precise, a Finnish newborn in 1995 is a net receiver of €83,200, whereas a
citizen born after 1995 would have to pay net taxes of €71,60014.
One thing the study by the European Commission certainly shows is that there is a very wide variation, even within a region like EMU, which shares a lot of common features. The results of the analysis are shown in table 3. According to the 1999 study, Belgium and Ireland are the only countries that have little to worry. Denmark, France and the Netherlands belong to a medium group with respect to the size of their generational unbalance. The group consisting of Germany, Spain and Italy have troubles that need significant actions straight away. Finally, it is questionable whether Austria, Finland, Sweden and the United Kingdom will in any way be able to cope with the enormous burden that is upon them.
As shown in table 2, the study concerning the Generational Accounting in the Netherlands (Van Ewijk et al., 2000) shows an absolute difference between newborn and unborn generations of €14,600 in 1998, whereas the European Commission (1999) find an absolute difference of €40,300. The large difference between these two results illustrates a major problem still present in the field of Generational Accounting today. It was introduced to solve the problem of ordinary accounting methods in which the definition of specific terms of the budget may influence the apparent outcome of fiscal policy. However, even though the usage of Generational Accounting may solve that problem, it introduces the problem that the business cycle or other short term fluctuations of the economic state can have huge influences on the outcome of Generational Accounting exercises. Obviously, attempts have been made to correct for this (as was also done by van Ewijk et al., 2000), but the difficulties of
13 The study by the European Commission concerns Belgium, Denmark, Germany, Spain, France, Ireland, Italy,
the Netherlands, Austria, Finland, Sweden and the UK. Due to difficulties in finding comparable data, Greece, Luxemburg and Portugal were not studied.
14 It should be noted that the authors do add that the base year of 1995 was extremely badly chosen for Finland
determining output gaps and structural budget balances are notorious. This is a strong limitation for the usage of Generational Accounting in actual budgeting policy, but the results of Generational Accounting exercises still offer interesting indications about the actual inter-temporal fairness of a current policy stance and thus about fiscal sustainability.
4. Methodology
15As mentioned earlier, the original methodology of Generational Accounting was developed by Auerbach, Gokhale and Kotlikoff (1991). However, in this section I will explain the methodology as I use it, which differs somewhat from the original one. The reason for this is that, despite the short life-time of Generational Accounting so far, it has gone through a process of evolution and the original methodology of Auerbach et al. (1991) is not entirely appropriate anymore.
4.1 Theoretical methodology
The first step in Generational Accounting is to determine the government inter-temporal budget constraint for year t:
, , 1 b b t b d b t N N W G τ τ τ τ τ τ τ ∞ ∞ = − = + = + + =
∑
∑
∑
(1)where b stands for the year of birth and; d for the maximum age of death; τ represents the current year, to which all cash flows are discounted; t stands for the year under consideration;
N refers to the Present Value (PV) of net payments by a particular generation; W symbolises
government wealth and G stands for the PV of government consumption. Overall, this means that the first element on the left-hand side of equation (1) reflects the PV of all future net payments made by currently living generations. The second element covers the PV of net payments made by all future generations, and the third part stands for government wealth. On the right-hand side of equation (1), the PV of all future government consumption is found.
Some calculus shows that the PV of government consumption can be presented as:
( ) (1 ) (1 ) t t t t G G P τ τ τ τ γ δ − ∞ ∞ = = + = ⋅ ⋅ +
∑
∑
% (2)where G% stands for government consumption per capita and Pt is the size of the population at
time t16. It is assumed that government consumption per capita grows at rate γ, the growth of
productivity, and is discounted at rate δ.
Government wealth is another variable that is relatively easy to define. In Generational Accounting literature, it is uncommon to simply take government debt, as this misses out on a number of other components that contribute to government wealth. In this study, the method adopted in a study of the European Commission (1999) will be followed, in which government wealth is calculated as the sum of official government debt and the PV of profits of government owned enterprises, net of subsidies. The calculation of net government then looks as follows:
(
)
, 1 1 K t k k k W Dτ θ γ τ δ = + = +∑
⋅ ⋅ Π (3)In this equation, Dτ stands for the official government debt, and Πk,τ stands for the profit of
government-owned firm k at time t. The total number of government-owned enterprises is K
and the government share of ownership is θk. Clearly, profits are expected to grow at
productivity growth rate γ.
The most important part of the model is obviously the creation of generational accounts, both for living and future generations. Calculating generational accounts for current generations is a necessary prerequisite for making accounts for future generations, so these will be presented first. The Generational Account (GA) of a representative individual of gender g, who was born in year b is defined as follows:
( ) , , , , , , , , , 1 1 (1 ) t t d M b g m t b g m t b g t b g t m GA T E S τ τ δ − + = = = ⋅ ⋅ ⋅ +
∑∑
% (4)where m denotes the tax type and M is the total number of different taxes types17. T% denotes
the size of a particular tax payment by an individual of gender g, born in year b. E refers to the so-called eligibility for a certain tax, i.e. the percentage of the population group (defined by age and gender) that will be eligible to pay a tax (or receive a transfer). S refers to the so-called Survival rate, i.e. the proportion of the people born in year b who are still alive in year
t.
16 It is obviously possible to use lump sum government consumption as well, but it seems very reasonable to
assume that government consumption is for a greater part dependent on the size of the population. A combination of lump sum and marginal consumption is sometimes used, when appropriate.
17 When the word tax is used, in this context, this refers to either a tax or a transfer. Transfers are regarded as
However, GAg,b only reflects an individual account. Of course, there are a large
number of such accounts that have to be made18 and the totals have to be multiplied with the
size of each of these groups in order to calculate the total payments by living generations:
2 , , , , 1 t b b g b g b d b d g N P GA τ τ τ τ τ = − = − = = ⋅
∑
∑ ∑
(5)Obviously, Pb,g,τ stands for the total population at time τ of a particular gender and age.
Having calculated all but one of the elements of equation (1), since this is an identity, it is now possible to calculate the necessary payment to complete it: the net payment by future generations. This completes equation (1), but the total payment by future generations should still be split, to show the necessary payment per individual:
(
)
( ) 2 , , 1 1 1 1 1 b t b b g g b t g b N P N τ τ γ δ − ∞ ∞ = + = = + + = ⋅ ⋅ + ∑
∑ ∑
(6)Equation (5) should appear reasonably clear: the total payment of all future generations is a
summation over generations and future birth years of the payment Ng, which grows at growth
rate γ and is discounted at discount rate δ. It is the value of Ng, which is of interest as it is the
value that can be compared by the GA of a newborn individual at time τ.
4.2 Practical application
The practical application of the methodology described above has some challenging features, especially with regard to the determination of generational accounts of living generations. The accounts have been calculated using the following equation for each tax type m and gender g:
, , ,
m g m g m g g
N
=
T
o
E
o
S
o
D
(7)In this equation, the capital letters stand for different matrices. The elements of the Taxation
matrix Tm,g show the amount of tax an individual born in year b has to pay in year t. The
payment shows up in the matrix at row t- τ and column τ-b. In the case of a model with
period living individuals, the Taxation matrix will look as follows, where the elements
,
b t
T% refer to the payment by an individual born in year b, paid at time t:
(
)
(
)
(
)
, 1, 2, , 1, 2, , , 1 1, 1 1, 2, , 2 2, 0 1 1 0 0 0 1 0 0 m g T T T T T T T T T T T T T τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ γ γ γ − − − − + − + − − + − = = + + + % % % % % % % % % % % % (8)As can be read from the matrix, the payments of individuals are expected to increase at growth rate γ.
The Eligibility matrix, Em,g remains steady in the long run, as long as there are no strong
changes to be expected. Of course, such changes, like an increasing retirement age are possible, and these changes can easily be incorporated in the model. In the long run, such variables are expected to stabilise, making the Eligibility matrix look as follows, where the
elements Eb,t represent the eligibility in year t of individuals born in year b:
, 1, 2, , 1, 2, , , 1 1, 1 1, 2, , ` 2 2, 0 0 0 0 0 0 m g E E E E E E E E E E E E E τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ − − − − + − + − − + − = = (9)
Finally, the Survival matrix Sg contains the percentage of a certain age group born in year b
that is still alive in year t, where Pb,t is the size of the population group born in year b at year
t: , 1, 2, , 1 1, 1 , 1 1, 1 , 1, , 2 , 2 , 1 1 1 0 0 0 0 0 0 g t t t t S S S P P S S S P P S P P τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ − − + − + + − + − + + = = (10)
The final feature in equation (6) is the Discount matrix D, which simply has columns with
(
)
( ) 1 1 t τ δ − + summed to create a matrix that shows the NPV of all payments to the government that will be made by all representative individuals during each year of their potential lives. When the columns are then summed, the individual Generational Accounts show, and when the first row is summed, the budget balance in year t is shown (excluding government consumption).
The challenging part of Generational Accounting is calculating the Eligibility and Taxation matrices. After all, the starting point must be a reflection of current fiscal patterns in
a country. These data are not always readily available19, and estimations must be made.
Micro-economic data are used to make estimations about the incidence of taxation over different age groups, which are then finally scaled to fit the macro-economic data provided by the government budget.
The final issue relevant to performing a Generational Accounting exercise is the criterion used to pass an actual judgement about the severity of the problems. It has been mentioned before that different authors use different criteria. First, one can check whether current fiscal policy is unsustainable by examining whether the burden carried by future generations is heavier than the burden carried by newborns. There is no discussion as to whether this is a relevant measure, but the size of the imbalance can be measured in different ways.
i) One can easily look at the absolute difference between future generations and
newborns. However, this does not give any clues as to whether the difference is small or large.
ii) One can look at the relative size of the difference, compared to the Generational
Account of the newborn generation, as Auerbach et al. (1999) do. This is an inappropriate method, because generational accounts near zero will yield
differences that are very high, even though the imbalance might be small20.
iii) One can look at the Intertemporal Public Liabilities (IPL), as the European
Commission (1999) does. The IPL shows the residual necessary to balance the intertemporal budget constraint if the GA for future generations is calculated in the same way as the GA of current generations is. This will leave equation (1) unbalanced and the size of the imbalance is called the IPL. This method shows clearly what the size of the problem is and is internationally comparable. However, the outcome in itself may not be interpreted that easily.
19 Even if the relevant data would be readily available, the incidence of taxation as discussed in chapter 3.2
iv) Another method is to calculate the necessary changes in taxation or expenditures (European Commission, 1999), which is often done as a policy warning anyway. This is a very clear indicator that states the size of the problem, but a result may reflect only the size of the government and not the size of the generational imbalance. In addition, this indicator seems to be most appropriate for the policy section, and should not necessarily be used as an indicator of generational imbalance.
v) The final indicator, which will be used in this study, is to show the difference
between newborns and future generations as a percentage of GDP per capita. This method is superior, because it has a straightforward interpretation, is simple to calculate and retains its comparability.
20 For example, the difference between paying $1,000 and paying $5,000 should obviously be less severe than
5. Data description
In this chapter, the data and their necessary transformations are described, including their sources.
5.1 Demographic data
Obviously, it is necessary to construct a complete data-set regarding the demographic features of Latvia. The most important source for these data is the United Nations (2005), which publishes forecasts with regard to population size, fertility and mortality rates, life expectancies and the resulting age structure of the population. The relevant data are shown in table 4. In addition to that, predictions regarding Latvian immigration statistics are also presented by the United Nations. In Western Europe, immigration tends to slightly reduce the effect of ageing and decreasing fertility rates. However, Latvia does not benefit from immigration. In the past, there has been very strong emigration (between 1995 and 2000, 11,000 people left Latvia each year, i.e. 0.5% of the population) and although the United Nations expect net migration to improve, it will remain negative at -2,000 people per annum.
The original United Nations estimates run until 2050. and these have been extended by determining age-specific death rates and combining these with the age-specific birth rates. Assuming age-specific death rates to stabilise after 2050 effectively means that life expectancy remains constant (Turner et al., 1998), which can then be combined with other demographic data that is assumed to remain constant after 2050 as well. In fact, the only change allowed for after 2050 is an increase of immigration (decrease of emigration), which
will change net migration to 021. Doing this yields a population forecast until 2200, which
5.2 Social data
To perform the Generational Accounting exercise in an accurate way, a number of social variables have to be observed. These variables are necessary to determine the matrices regarding tax and eligibility, as described in chapter 4.
First of all, the official age of retirement in Latvia is 62 for men and was 58.5 for women in 2001, which will be increased to 62 in 2008. Moreover, the government is planning to increase the age of retirement even further after 2008. Until July 1, 2005, both men and women can stop working two years before their official retirement age. For men, the actual average retirement age has been increasing slowly, up to 61.1 in 2003. Women, on the other hand, have an average retirement age of only 57.7 in 2003 (Ministry of Welfare, 2004). During the next few years, retirement ages are expected to continue rising, as a result of current policy intentions. In equilibrium, men are assumed to have an average retirement age of 64.6 and women of 61.2. This implies an additional 3.5 years in the labour force for both genders.
Labour participation rates have risen strongly during the last few years, except for the lowest age group (15-19), due to the increased importance of education. In 2003, the overall labour participation rate for the ages 15-64 was 64.7% for women and 74.0% for men. In higher age categories, labour participation also rises, but is still very low. Females aged 55-64, for instance, only have a participation rate of 41.7% (Ministry of Welfare, 2004). Over the first few years, the increase in participation rates is expected to continue, and it is assumed to
remain stable after 201522.
Using data from Eurostat (2005), exact educational participation rates for each age group for the year 2003 have been uncovered. Education participation rates are expected to increase further, even though they have already risen quite strongly during the past few years. Women will remain more highly educated than men, although the male educational deficiency will be reduced slightly. At age 18, education participation among women is 76.1% in 2003, which will increase to 77.5% in 2010. Among men, the participation rate of 18-year olds will increase from 67.5% to 70% between 2003 and 2012.
21 Without changing the net migration, the population of Latvia will actually be negative in 2200. This is due to
the fact that the United Nations (2005) assume an absolute migration of -2,000, although all other variables have been expressed as relative variables.
22 As these are age-specific participation rates that increase, it is not actually possible to give an equilibrium
5.3 Fiscal data23
In 2003, the Latvian budget balance showed a small deficit of 2.1% of GDP, which is a bit lower than it has been in the past. Government income can be divided in six categories, as can be seen in figure 3. The main category, direct taxes, can be split further into four subcategories: taxes on corporate income (9% of all direct taxes), direct income taxes (33.5%), social security contributions (52.5%) and property taxes (5%). The second-most important category, indirect taxes, is split into three categories: value added taxes (VAT, 67% of all indirect taxes), excise taxes (30.5%) and customs taxes (2.5%).
The income categories of self-earned revenues and foreign assistance are not burdens to the population and should thus not be included in the generational accounts. After all, these are not payments levied on the population, so they can not be attributed as such either. On the other hand, these items do need to be included in the final equation, which can be done by recording them as negative government consumption. The self-earned revenues are expected to grow at the same rate as productivity, as is all other government consumption. Foreign assistance, on the other hand, can not be expected to grow at that same rate. After all, Solow’s convergence theory suggests a high growth rate for Latvia, which will lead to a decreasing gap between Latvia and other developed nations. This would imply that Latvia should become less eligible for receiving foreign assistance. In this study, the growth rate of foreign assistance is expected to be half of that of the rest of government consumption. Non-tax revenues and other taxes are burdens for the population, but cannot be attributed to particular
ages24. These are spread out evenly over all ages. Per capita non-tax revenues and per capita
other taxes are also expected to grow at the same rate as productivity.
Social security contributions are officially set at 33.09% of income (9% for employees and 24.09% for employers), but it seems as if employers have found ways around paying the full contributions. Macro-economic data show that the total contributions are only 24.6% of wages. The separation between employees and employers does not matter, as the argumentation of Fehr and Kotlikoff (1999) shows that in small open economies, the burden will fall upon employees anyway. Using the 24.6%, age-specific profiles of social security contributions can be constructed.
Since the income tax rate is 25%, income taxes are calculated in the following way:
23 Fiscal data have been obtained from a number of different sources. Eurostat (2005), CSB (2005) and the
Ministry of Finance (2004a) are the main ones.
(
)
0.25Tax= ⋅ wage contributions threshold− − −deductibles
The tax-free threshold was 252 LVL25 in 2003, plus 180 LVL per dependant. Deductibles
include medical and schooling expenses. Unfortunately, no data are available concerning the age-specific size of these deductibles. Therefore macro-economic data have been used to fit the details of the income tax system to the micro-economic data available. Eventually, for 2003, the tax-free threshold has been set at 432 LVL per capita and the tax rate at 20.95%. This lower value reflects the deductions from the wage tax that can be made. The tax rate is assumed to remain constant in the future, but the tax-free threshold will increase at the same rate as productivity.
Finally, property and corporate taxes depend on wealth. Approximately one-third of corporate taxes is paid by foreigners and is thus treated as negative government consumption as well. The rest of corporate taxes, as well as property taxes are distributed according to estimated wealth. On the basis of approximate wealth profiles, the corporate tax and property tax have been attributed. Auerbach et al. (1991) and others use complicated procedures to attribute the taxes on property to the appropriate generation, but as the amount of property taxes is relatively small anyway (53 million LVL), these are simply attributed according to wealth.
For the indirect taxes, similar procedures have been followed. Customs taxes cannot be attributed to any particular age, so these are spread evenly over all age groups. Excise taxes, on the other hand can be attributed more precisely. McKee et al. (2000) present data regarding alcohol use among different ages and genders. Using these data, it is possible to make age- and gender-dependent profiles for alcohol consumption and thus for alcohol excise taxes. Tobacco taxes have been attributed on the basis of smoking prevalence, as presented by the World Health Organisation (2003). Finally, excise taxes on oil and other products could not be properly attributed to the appropriate generations, but have instead been split over the population of 15 years and older. Due to the low prevalence of alcohol and tobacco use amongst women, there is a very large difference between men and women in the case of excise taxes. On average, women pay only 66.5% of what men pay. Towards the future, the age-dependent level of excise taxes is expected to grow at the same rate as productivity.
it is unclear who is paying what percentage. Therefore, macro-economic data have been used to determine the average VAT rate of 16.1% and all ages will pay the same uniform rate of VAT. Consumption itself has also been attributed to particular ages. As discussed in chapter 3, other Generational Accounting exercises have encountered difficulties in attributing consumption. If consumption is considered to be a share of income, this would imply that young generations, who do not earn any income yet, do not consume and are not thus burdened with VAT. This does not seem reasonable and a middle solution has been sought here. In 2003, all individuals are assumed to consume 561.24 LVL, plus a percentage of their income. The fixed part of consumption is equal to 50% of the reported subsistence level (CSB, 2005) and the flexible component has been fitted to reconcile the micro-economic data with the macro-economic outcomes. Consumption is then assumed to follow the same pattern as productivity.
Figure 4 shows government expenditures per category. Social security, education and healthcare expenditures form the largest shares. These are exactly the three categories that are considered transfers, whereas the other categories of government expenditure are classified as government consumption. All elements of government consumption grow at the same rate as productivity.
Educational expenditures total 405.5 million LVL in 2003. Eurostat (2005) publishes data that attributes budgets to specific levels of education and also provides information about the exact age-dependent education participation rates for each level of education. These data can be combined to create age- and gender-dependent costs of education. These costs per student are expected to increase at the same rate as productivity. Note that this does not mean that educational expenditures will rise at that rate. After all, the number of people in student-age will decrease over the next few years, which will actually reduce the projected outlays significantly.
Healthcare is another category of major importance. The Ministry of Finance (2004b) provides age- and gender-dependent costs of healthcare in 2002. This information has been combined with macro-economic data to create age- and gender-dependent estimations of healthcare costs in 2003. In this case, total outlays will outrun the productivity growth, even though the age-dependent costs will grow at the rate of productivity growth.
The final and largest category of transfers is social security. Social security includes maternity, child, family, disability, sickness, guardian, unemployment and funeral benefits, as well as pensions (Ministry of Welfare, 2004). These have been assigned to the appropriate
