This Selected Issues paper and Statistical Appendix presents a number of studies designed to probe in more depth issues of fiscal adjustment, the public sector’s creditworthiness as judged by foreign lenders, and the banking system in Greece. The paper highlights that fiscal adjustment in Greece has followed a gradual path, relying to a significant extent on revenue enhancement and eschewing primary expenditure reduction. The paper also examines the evolution and tax determinants of the most widely followed indicator of the health and competitiveness of a banking system—the lending-deposit spread.

Abstract

This Selected Issues paper and Statistical Appendix presents a number of studies designed to probe in more depth issues of fiscal adjustment, the public sector’s creditworthiness as judged by foreign lenders, and the banking system in Greece. The paper highlights that fiscal adjustment in Greece has followed a gradual path, relying to a significant extent on revenue enhancement and eschewing primary expenditure reduction. The paper also examines the evolution and tax determinants of the most widely followed indicator of the health and competitiveness of a banking system—the lending-deposit spread.

II. Interest Rate Convergence and Household Consumption: How Important is the Income Effect?5

A. Introduction

16. The implications of the substantial easing of monetary conditions over the next fifteen months in the case of Greece, associated with interest and exchange rate convergence in the run-up to EMU, is an area of considerable analytical and policy interest. In particular, from a policy perspective, understanding the effects in question would allow a better assessment of the policy mix and suggest the extent to which other macroeconomic policies may need to be adjusted to maintain the desired overall policy stance. In making this judgment, both the size of the effects of a monetary easing and the time horizon over which they operate are relevant.

17. The previous chapter implemented a structural VAR model, based on a high degree of aggregation, and offered empirical evidence suggesting that a decline in short-term interest rates, and a concomitant depreciation of the drachma, can be expected to entail a substantial increase in the rate of inflation. Moreover, this effect turned out to be quite persistent, extending over the entire four-year simulation horizon. These results would imply that a monetary easing of the type envisaged for Greece in the run-up to EMU would require a quite sharp increase in the primary fiscal balance and/or a very rigorous program of structural reforms to mitigate its impact on prices.

18. It could legitimately be questioned, however, whether such an inherently simple model, utilizing variables pertaining to overall economic activity and not explicitly controlling for variables pertaining to the patterns of financial intermediation and financial asset holdings, can do justice to all the various factors that could be potentially relevant. Specifically for the issue at hand, it has been suggested that, owing to the large stocks of bank deposits and government paper held by Greek households, a decline in interest rates could entail a large reduction in household income that could depress consumption to an extent that might offset partially, or even fully, the expansionary impact of lower interest rates on other components of aggregate demand (notably investment and the external balance). This chapter attempts to shed some light on this question by looking at the experience of, and model predictions for, a number of current euro area participants that underwent a process of interest rate convergence in recent years and that, to varying degrees, bear some relevant similarities to Greece’s situation.

B. Some Stylized Facts and Theoretical Considerations

19. The likely impact of a reduction in interest rates on the interest income of Greek households is quite substantial, given the currently large interest rate differential vis-à-vis the euro area at the short end of the maturity structure and the large share of household financial assets held as bank deposits or treasury bills. On the basis of the stocks of bank deposits and repos and of short-term government paper held by households at end-1998 (Dr 18.2 trillion and Dr 13 trillion respectively), the Bank of Greece estimates that interest rate convergence to euro-area levels will entail a reduction in interest income accruing to households of some Dr 1.2 trillion, or 3¾ percent of GDP.6 At first sight, therefore, the relevance of the income effect of an interest rate change on Greek household consumption would appear likely to be nontrivial.

20. As is often the case, however, economic theory does not provide an unambiguous answer to the question of the impact of a change in the real interest rate on consumption. Thus, while the income effect of a reduction in interest rates would tend to reduce consumption, other factors work in the opposite direction. It is important to keep in mind in this connection that it is rather misleading to view interest income as an exogenous determinant of household consumption, in the sense that, at least in the absence of liquidity constraints, intertemporally optimizing households would in general simultaneously determine their holdings of financial assets (and hence their “interest income”) and consumption expenditure. In that sense, there is a fundamental difference in the response of consumption to a decrease in wage income resulting from loss of employment or a reduction in the real wage, which can to a reasonable approximation be viewed as exogenous, and the corresponding response to a reduction in interest rates. Thus, in response to a decline in real interest rates, households could conceivably move their savings away from interest-bearing assets and into assets whose rate of return is either unrelated or inversely related to the interest rate (stocks come immediately to mind in this regard), or indeed substitute out of saving and into current consumption altogether, thus directly boosting economy-wide consumption expenditure.

21. The ambiguities resulting from the interaction of the income and the substitution effect are illustrated in Figure 1. These well-known diagrams illustrate the choice between current and future consumption (C1 and C2, respectively) for a household with current and future income streams of Y1 and Y2; for simplicity, current and future income are assumed to be independent of the rate of interest. Line AA represents the household’s intertemporal budget constraint: it passes through the (Y1, Y2) point and its slope depends only on the real interest rate (the lower the interest rate, the flatter the budget constraint). Point E represents the household’s equilibrium at the original interest rate, with the upper panel depicting the situation of a household in a net creditor position and the lower panel the situation of a household in a net debtor position. A reduction in the interest rate will tilt the budget constraint through the (Y1, Y2) point to A’A’, with the new equilibrium represented by point E’ For a household originally in a net creditor position, current consumption could increase or decrease, depending on whether the substitution or the income effect dominates. A similar ambiguity exists for a household that is originally a net debtor—although in this case current consumption will increase if both current an future consumption are normal goods. An important asymmetry between the two types of households in this highly simplified setting may be worth noting. For a net debtor, an interest rate decrease will unambiguously raise welfare; on the other hand, the welfare of a net creditor will be lower as long as the household remains a net creditor (but will rise if it switches to a net debtor position in response to the lower interest rate—not shown in the diagram).

Figure 1.
Figure 1.

Greece: Interest Rates and Consumption

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A002

22. A number of additional factors slightly complicate the picture, but can be expected to reinforce the substitution effect in boosting current consumption in the wake of a decline in real interest rates. First, if the interest rate decline acts to raise household current and/or future wage income (via an increase in aggregate labor demand), the budget constraint will not only become flatter but it will also shift out. In that case, a net creditor household will be more likely to raise current consumption, and at the same time may enjoy a welfare improvement as a result of the interest rate decline, even if it remains a net creditor. Second, a similar set of considerations apply to a situation where consumption depends not only on current and future income streams but also on an inherited stock of wealth, whose real value increases as interest rates decline—for example, in the form of capital gains on bond holdings. In the presence of this wealth effect, the intertemporal budget constraint will also shift out in the wake of an interest rate decrease, with very similar implications for current consumption and the welfare level of net creditor households.

23. In the face of these ambiguities, the issue at hand needs to be resolved empirically. Unfortunately, however, such an investigation for the case of Greece is seriously hampered by data limitations. In order to formally test the empirical validity of the effects outlined above, one would need rather detailed household income and financial accounts, which are unavailable for Greece. Even a second-best approach that would entail estimating the impact of interest rates on household consumption is precluded by the unavailability of data on aggregate demand components at higher than annual frequency—working with annual data would necessitate looking at a very long sample period over which Greek financial markets underwent fundamental change, so that even the information content and economic function of interest rates in affecting the behavior of economic agents is far from uniform.

24. While these limitations preclude the empirical investigation of the question at hand for the case of Greece, it can be hoped that that one can draw relevant inferences by looking at the experience of other countries that bear important similarities to Greece. In what follows, we perform model simulations on three southern European economies that currently participate in the euro area: Italy, Portugal, and Spain. All three economies were characterized by significant interest rate premia vis-à-vis the rest of the euro area and hence experienced substantial interest rate convergence in the run-up to EMU—albeit to a smaller extent than what is in store for Greece. Moreover, these economies have a broadly similar economic structure to Greece, as well as a similar level of development of their financial systems; an important common feature for the issue under consideration is that, at least until very recently, the bulk of household financial assets was held in the form of bank deposits or short-term government paper.

25. A cursory look at the behavior of consumers in Italy, Portugal, and Spain in the run-up to EMU casts doubt on the quantitative importance of the income effect on consumption of a reduction in real interest rates. Figure 2 plots the evolution of consumer spending and consumer credit for each of these three countries immediately before, during, and in the immediate aftermath of interest rate convergence.

Figure 2.
Figure 2.

Consumption and Consumer Credit, 1995:I-1999:II

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A002

Source: Data provided by the authorities.

26. The trends depicted in Figure 2 fail to detect a dampening impact of interest rate convergence on consumption for each of the three countries considered. If anything, consumption appears to accelerate in the wake of (or just prior to) the decline in real interest rates, and continues to grow vigorously thereafter, especially in the case of Portugal and Spain. Even in Italy, in the context of a very weak economy, consumption shows signs of picking up, growing significantly faster than real GDP and thus constituting one of the few sources of support to economic activity. In addition, for all three countries, one can observe a clear acceleration of consumer credit with the onset of interest rate convergence. This may suggest the emergence of a relatively new transmission channel of interest rate changes, to which we return in the concluding section.

27. Suggestive as they may be, such simple correlations cannot provide a definitive answer to the question of the impact of interest rates on private consumption, as it can plausibly be argued that other factors could have been at work in boosting consumption in the three countries under consideration. For instance, at least for Portugal and Spain, a case could be made that improved consumer confidence and sentiment, linked to the prospect of EMU participation itself and the associated boost to investment and employment, may have been more important factors accounting for the acceleration in consumption than the real interest rate reductions. In order to obtain a more satisfactory answer to the question at hand, we turn to simulations based on two well-established macroeconometric models.

C. Oxford Economic Forecasting Model Simulations

28. The first set of policy simulations for the three countries under consideration was conducted on the basis of the Oxford Economic Forecasting model (OEF),7 which is among the tools utilized by EU1 desks for their WEO projections. OEF is a relatively small-scale quarterly macroeconometric model of most OECD countries (Greece being a notable exception for lack of quarterly national accounts data) specifying Keynesian short-run dynamics and a classical long-run steady state. It is a well-specified general equilibrium model, with backward-looking expectations and an error-correction structure, which imposes a gradual adjustment of a variable to its steady-state level. The main advantage of its relatively small size and fairly common structure across countries is that it renders cross-country comparisons fairly transparent and easy to interpret. Its main drawback is that it could miss certain country-specific factors or transmission channels that may be quite important in reality.

29. As regards real household consumption, its short-run dynamics are modeled to be driven by changes in real disposable income, changes in the real interest rate (with the corresponding terms including variable lags), and lagged real household consumption. In the long run, real consumption is modeled to depend on real disposable income (with its coefficient constrained to 1 in accordance to the permanent income hypothesis), the wealth to disposable income ratio, and the real interest rate. In terms of the theoretical discussion of the previous section, it would thus appear that OEF should in principle capture the substitution effect of a change in the real interest rate. On the other hand, with wealth specified to enter the picture only in the long run, the contribution of corresponding wealth effects in the short-run dynamics of consumption can be expected to be rather limited.

30. The model’s estimation results for all three countries considered points to a negative and statistically significant impact of real interest rate changes on consumption, holding real disposable income constant, over a number of lags.8 With the income effect presumably reflected in the disposable income term and the wealth effect reflected in the wealth term (albeit with the restrictions mentioned above), it would thus appear that the real interest rate terms should capture the underlying substitution effect. It should also be noted that the lag structure of the interest rate terms and, especially, the error-correction structure of the model, should impart a substantial degree of persistence on the effect of the real interest rate on consumption.

31. The simulation experiment performed in this section (and the next) seeks to determine the impact on consumption and real GDP of a permanent reduction in the real short-term interest rate by 1 percentage point. The impact of this change was simulated over a four-year horizon. The choice of the length of the simulation horizon was governed by two main considerations. First, a horizon of four years should capture the actual estimation results, before the long-run equilibrium conditions that serve to close the model (which are to some extent arbitrary) kick in. Second, a four-year horizon seems long enough to be relevant for policy purposes—for instance, if one were to attempt to estimate the amount of fiscal tightening needed to offset the impact of a given interest rate reduction, it would be unlikely that one would be looking for the relevant impact beyond four years.

32. Finally, it should be emphasized that the version of OEF used for the simulations of this section is the post-EMU one. As such, it treats the exchange rate of each country under consideration with its euro area partners as fixed, a restriction that would not fit Greece which can expect exchange rate convergence to accompany interest rate convergence in the run-up to EMU. The implications of this feature of the OEF model will be compared and contrasted with the simulations of the next section. Moreover, the post-EMU version of the OEF model specifies an interest rate change as pertaining to the entire euro area, rather than as a change in a country’s interest rate premium vis-à-vis the rest of the euro area which is what is in store for Greece.9

33. The table below presents the simulated impact on real household consumption and GDP of a permanent 1 percentage point reduction in the real short-term interest rate for each country under consideration.

Simulations Using OEF

(Percent deviations from baseline)

article image

34. The simulation results of the above table cast considerable doubt on the dominance of the income effect of a change in the real interest rate on household consumption. For all three countries considered, a permanent interest rate reduction is shown to entail a positive impact on real household consumption, which, together with real GDP, remains above its baseline path throughout the four years of the simulation horizon. In fact, the simulation results seem to suggest that household consumption constitutes a major transmission channel of interest rate changes to economic activity in southern European countries, much more important than in the rest of the EU.10 As the following section suggests, however, this may be an implication of the restriction in the OEF model of fixing intra-EU exchange rates, which would tend to dampen the importance of the external current account as a transmission channel.

35. Another noteworthy feature of the simulations is that the cumulative (four-year) expansionary impact of a real interest rate reduction on real household consumption turns out to be smaller for Italy than for either Portugal or Spain. Given the model’s restriction of constant intra-euro-area exchange rates, this probably does not mainly reflect Italy’s lower degree of openness to international trade relative to either Portugal or Spain, even though Italy’s extra-EU exports are among the highest in the EU. Rather, it could reflect the impact of Italy’s higher government debt ratio, which is difficult to capture explicitly given the OEF’s linear structure.

D. Bank of Italy Model Simulations

36. This section conducts simulations on the basis of the Bank of Italy’s macroeconometric model. Focussing especially on Italy as a comparator to Greece for the issue at hand appears justified in view of their similarities in the fiscal area. Specifically, the two countries are characterized by a similar level of the public debt-to-GDP ratio and a very similar pattern of its financing: at least until very recently, both Italy and Greece resorted mainly to short-term financing of their debt, with domestic households holding the bulk of treasury bills issued—the potential implications of recent changes in this regard are discussed in the concluding section.

37. Compared to the OEF model, the Bank of Italy model also has Keynesian short-run properties, but it is much larger, specifying very detailed linkages between the various sectors of the economy;11 as such, it may capture more Italy-specific institutional features, including those pertaining to the pattern of public debt financing that are of particular relevance for the problem under consideration.

38. Regarding the modeling of household consumption, the Bank of Italy model specifies a number of its components separately, notably distinguishing between durable and nondurable goods. Moreover the Bank of Italy model contains a number of important differences relative to the OEF model in the specification of the income, substitution, and wealth effects of a change in the real interest rate. On the income effect, consumption is specified to depend on both wage income and household interest income, with the coefficients of the two components restricted to be identical; given the problems with treating interest income as exogenous to consumption mentioned in Section B, this is a rather strong restriction which may result in overstating the importance of the income effect. In contrast to the OEF model, the Bank of Italy model assumes that the real interest rate directly affects only the consumption of durable goods—hence, it may tend to underestimate the substitution effect. On the other hand, it allows household wealth to affect both the short-term dynamics of consumption and its long-run steady state.12 Finally, adjustment to the long-run steady state is specified to be faster in the Bank of Italy model than in the error-correction OEF model, so that the effects generated by the former can be expected to be less persistent than the corresponding effects generated by the latter.

39. The Bank of Italy model was used to simulate the same experiment as in Section C above, namely the effects on real consumption and GDP of a 1 percentage point reduction in the real short-term interest rate.13 The simulations were run by alternatively treating the lira exchange rate as exogenous (an assumption similar to that made by the OEF model) or endogenous (which would appear to be more relevant to the situation that Greece is faced with). The simulation results are summarized in the tabulation below.

Simulations Using Bank of Italy Model

(Percent deviations from baseline)

article image

40. Looking first at the simulation that treats the lira exchange rate as exogenous, which are more comparable to the simulations on the OEF model of the previous section, the Bank of Italy model appears to suggest a smaller expansionary impact on consumption of a real short-term interest rate reduction, but the overall picture is qualitatively rather similar. Specifically, an interest rate reduction is estimated to boost consumption over a horizon of three years after the shock, with the effect turning negative only in the fourth year as the income effect begins to make itself felt—in cumulative terms, the impact is clearly expansionary over the entire simulation horizon. Moreover, even in the fourth year after the shock, the income effect turns out not to be strong enough to offset the expansionary impact of the interest rate reduction on other components of aggregate demand (notably private investment), and real GDP is estimated to remain above its baseline path throughout the simulation horizon—in fact, the simulated impact of the interest rate reduction on GDP turns out to be remarkably similar to the predictions of the OEF model.

41. We now turn to the simulation that treats the lira exchange rate as endogenous—in this setting, an interest rate reduction would entail a lira depreciation. This version of the Bank of Italy model renders comparisons with the OEF simulations of the previous section more difficult, but at the same time should be of more relevance to the Greek case, where interest rate convergence would be accompanied by a depreciation of the drachma to its ERM2 central parity. Compared to the simulations that hold the exchange rate constant, the boost to consumption brought about by the interest rate reduction now turns out to be smaller, as the adverse terms of trade shock has a further dampening effect.14 Thus, consumption rises above its baseline path in the first two years after the interest rate shock, returns to baseline in the third year, and falls below the baseline in the fourth year. Still, over the entire simulation horizon, the cumulative impact on consumption remains expansionary, albeit to a smaller extent than in the constant exchange rate case. On the other hand, it should be emphasized that the overall expansionary impact on real GDP, which after all is what is relevant for policy purposes, is clearly stronger in the case where the exchange rate is treated as endogenous, notably during the first two years after the interest rate shock, as the improvement in the external current account associated with the exchange rate depreciation far outweighs the dampening impact of the terms of trade change.

E. Concluding Remarks

42. This chapter attempted to shed some light on the likely impact of the convergence of Greek interest rates to euro area levels in the run-up to EMU participation on household consumption. This question is of high policy relevance, given the large share of bank deposits and short-term government paper in Greek households’ portfolios. In this setting, one could legitimately ask whether the substantial reduction in real interest rates that is in store for Greece, by depressing household disposable income, could conceivably bring about a reduction in private consumption to an extent that might dampen a good part of the expansionary impact of the lower interest rates on other components of aggregate demand, thereby attenuating the need for offsetting fiscal and structural policies.

43. In view of the paucity of relevant data in the case of Greece, this chapter sought to draw some lessons from three other southern European countries (Italy, Portugal, and Spain) that have already gone through the process of EMU-related interest rate convergence. Casual inspection of the post-convergence trends in these three countries cast some initial doubt about the quantitative importance of the income effect: the decline in real interest rates was not accompanied by a perceptible deceleration of consumption, while consumer credit accelerated markedly.

44. Simulation results of a real interest rate reduction on the basis of two standard macroeconometric models tended to confirm this impressionistic picture. Thus, on the basis of the Oxford Economic Forecasting (OEF) model, a reduction in the real interest rate turns out to actually boost consumption in all three countries throughout the (four-year) simulation horizon, with this effect being particularly strong in the case of Portugal and Spain. Simulations for the case of Italy on the basis of the much larger Bank of Italy model yielded similar, if a bit less stark, results: a reduction in the real interest rate tended to boost consumption two to three years after the shock (depending on the treatment of the exchange rate), with the income effect becoming dominant only during the fourth year.

45. As with any econometric estimates, the results of this chapter need to be interpreted with caution. A question is whether some feature of the models used, or the estimation period, could bias the estimated magnitude of the income effect in either direction. A potential factor that could lead to under estimation of the income effect may be the presence of nonlinearities, that is, a situation where the “true” dampening impact of an interest rate reduction on consumption increases with the level of interest-bearing household assets. Since the models considered are linear in nature, they cannot capture the possible impact of the rising trend over time of the public debt held by domestic households, a trend also relevant for the case of Greece.15

46. As regards the estimation period, it could be argued that the far-reaching changes in the financial structure of the countries considered may render the empirical results less reliable for the current context. For the problem at hand, two such factors, relevant for the countries considered in the simulations as well as for Greece, could actually lead to the overestimation of the income effect.

  • First, the last years of the sample saw an important change in the maturity structure and pattern of holding of public debt. Whereas during most of the period public debt was predominantly short-term and the bulk of it held by domestic households, more recently the average maturity of the debt has lengthened, and there are indications that households have been switching out of treasury bills and into long-term bonds16 (also, the share of debt held by foreigners increased). In this context, and taking into account the sharp reduction in the longer term inflation outlook associated with EMU participation, households may have locked into interest rates that generate a higher ex ante stream of real income over the life of the security compared to the alternative of holding short-term paper. This factor could already be sustaining demand, and thus make a simple comparison of household interest income between one year and the next quite misleading for the purposes of explaining consumption behavior.

  • Second, a notable development of the early 1990s was the liberalization of consumer credit, which has since grown very rapidly in importance as a source of financing of consumption spending. This development could add another channel, namely the credit channel, through which a decline in interest rates can boost consumption, which would not be fully captured by the sample period over which the models used in this chapter were estimated.

47. Taken together, the results of this chapter would suggest that policy formulation should not rely on the income effect of the interest rate reduction that is in store for Greece to offset the latter’s expansionary impact on other components of aggregate demand. Rather, they point to the likelihood that recourse to other policy instruments (notably fiscal and structural) is necessary to keep inflation under control. Particularly as regards fiscal policy, the result that the expansionary impact of an interest rate reduction makes itself felt fairly quickly would suggest a front-loaded response, given that the lags typically associated with fiscal policy are rather short.

References

  • Banca d’ Italia, 1986, “Modello Trimestrale dell’ Economia Italiana,” Temi di Discussione No. 80.

  • Bank for International Settlements, 1994, National Differences in Interest Rate Transmission, (Basel).

  • Dornbusch, R., C. Favero, and F. Giavazzi, 1998, “Immediate Challenges for the ECB,” Economic Policy, April.

  • Gaiotti, E., A. Gavosto and G. Grande, 1997, “Inflation and Monetary Policy in Italy: Some Recent Evidence”, Banca d’ Italia Temi di Discussione No. 310, July.

    • Search Google Scholar
    • Export Citation
  • Galli, G., D. Terlizzese, and I. Visco, 1990, “Short- and Long-Run Properties of the Bank of Italy Quarterly Econometric Model,” in Dynamic Modelling and Control of National Economies, ed. by N. M. Christodoulakis (London, IFAC).

    • Search Google Scholar
    • Export Citation
  • Levy, J. and I. Halikias, 1997, “Aspects of the Monetary Transmission Mechanism Under Exchange Rate Targeting: The Case of France,” IMF Working Paper 97/44, April (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Nicoletti Altimari S., R. Rinaldi, S. Siviero and D. Terlizzese, 1997, “I Canali di Trasmissione della Politica Monetaria nel Modello Econometrico della Banca d’ Italia”, Banca d’ Italia Temi di Discussione No. 316, September.

    • Search Google Scholar
    • Export Citation
  • Oxford Economic Forecasting, 1996, The New Oxford World Model: An Overview (Oxford).

5

Prepared by Ioannis Halikias.

6

This amount is in net terms, that is, it adjusts for the lower tax liabilities on household interest income.

8

For instance, for the case of Italy, the estimated consumption function is of the form:

Δ ln C =0.02083 + 0.1766Δ ln DY 0.00087Δ R 0.00073Δ R1 0.00075Δ R2+ 0.6732Δ ln C1 0.0750 [ln C1 ln DY1 + 0.0300 W1/DY1 + 0.00466 R1 + 0.3056]

where C is real consumption, DY is real disposable income, R is the real interest rate, W is real household wealth, and Δ is the first difference operator. The first six terms in the equation represent the estimated short-run dynamics, while the term in brackets is the estimated long-run relationship.

9

See, however, Levy and Halikias (1997) and Dornbusch and others (1998) for a theoretical discussion and empirical evidence on possible differences in the impact on aggregate demand between changes in the interest rate premium and changes in the anchor currency interest rate under the ERM regime.

11

For a description of successive vintages of the Bank of Italy’s macroeconometric model, see Banca d’ Italia (1986) and Galli and others (1990).

12

The large estimated coefficient of the wealth variable may suggest that this variable may be actually capturing part of the substitution effect that is ignored by not including the real interest rate as a direct determinant of the consumption of nondurable goods.

13

The simulations are included in Nicoletti Altimari and others (1997). See also Gaiotti and others (1997) for very similar simulations (which however focus on inflation rather than on real activity) on the basis of a VAR model.

14

In the Bank of Italy model, the terms of trade shock affects both real disposable income and real household wealth.

15

In this regard, one could consider re-estimating the models over successive subperiods, but would very quickly run into problems of degrees of freedom.

16

Although no data on household financial accounts exist for the case of Greece, it should be noted that mutual funds, generally thought to be held predominantly by households, increased from 3 percent of GDP in the early 1990s to 25 percent of GDP by end-1998.

Greece: Selected Issues
Author: International Monetary Fund