## Donogh McDonald

### Background

Substantial resources will be needed to rebuild the economy of east Germany. Over time, as productivity rises, a growing proportion of these resources is likely to come from domestic (i.e., east German sources). But, in the early years of German economic, monetary, and social union (GEMSU), it is likely that net saving in east Germany will be negative. From the perspective of the German economy as a whole, the large external surplus that has existed in the western part of the country will be more than adequate to finance resource needs in the eastern part. But the redirection of this surplus will have implications both for west Germany and the rest of the world. The transmission of demand shock associated with GEMSU to the rest of the world, which has been analyzed in Chapter VI, will depend, inter alia, on the extent to which the west German economy is able to generate additional resources for investment in the east, that is, over and above its existing surplus of saving. In examining this last issue, there are three important questions concerning how the west German economy responds to GEMSU: How much additional output will be stimulated? How will saving respond? And what will be the reaction of investment? Some of the important considerations bearing on the first question were analyzed in Chapters VII and VIII. Taking an assumed scenario for the effects of GEMSU on output, interest rates, and other macroeconomic variables, this chapter uses econometric estimates of saving and investment behavior in west Germany to look at aspects of the second and third questions. In addition, the influence of changes in government saving is examined closely, in order to evaluate the potential role for budgetary measures in generating additional saving at the national level.

The broad conclusion is that higher saving in west Germany rather than reduced investment is the most likely source of additional resources that could be channeled to east Germany. However, particularly in the early years, the rise in saving, while far from insignificant, will be small in relation to the prospective net import demand of east Germany.^{1} Initially, the boost to saving will come largely from higher output. A shift in the distribution of income away from labor (relative to baseline) is likely to occur more gradually but has potential to be a relatively important source of additional saving. Quantification of the effects of GEMSU on income shares is, however, rather difficult at this stage. As regards the role for policy, the econometric work indicates that perhaps 60 percent of any change in government saving is offset, in the short term, by adjustments in private saving, but that this offset is reduced to about one fourth in the third year and is negligible in the longer term. On investment, the simulations suggest that, in the early years of GEMSU, capital spending by businesses will be negatively influenced by a rise in real interest rates, but that these negative effects are likely to be offset by the boost given by faster growth. On balance, therefore, changes in business investment in west Germany are unlikely to free resources for capital formation in the east.

### The Effects of GEMSU on Saving Behavior in West Germany

#### Econometric Estimates

A marked rise in national saving in west Germany since 1982 has been the principal factor in the emergence of a large current account surplus (Chart 1 in Chapter I). A previous study (McDonald (1989)), seeking to explain this, developed a model that tracked saving quite well, both in and out of sample, including the sharp change in direction that took place after 1982. The model—which related the net national saving rate (*s*_{n}) to the growth rate of real NNP (*g*_{v}),^{2} gains in the terms of trade *(deltot)*, the labor share in NNP *(labsh)*, and government saving in relation to NNP (*s*_{g})—forms the basis of the analysis in this chapter.^{3}

Two sets of estimation results are reported in Table 1: the first (equations 1.1 to 1.4) incorporates a correction for first-order autocorrelation, while the second (equations 1.5 to 1.8) does not. The second set of equations produces higher standard errors, reflecting the omission of the autocorrelation correction; the point estimates also seem somewhat less stable to the eye. What is striking about the estimates in Table 1, however, is the degree of parameter stability across the various sample periods, even for the second set of equations.^{4} This stability is all the more notable when one considers the major shifts in the behavior of some of the explanatory variables in the sample period, particularly in *s*_{g} and *labsh*(Chart 1). The overall results are consistent with the good in-sample and out-of-sample tracking ability of the model found in McDonald (1989).

**Federal Republic of Germany: Equations for the National Saving Rate (s ^{n})^{1}**

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; ρ is the first-order autocorrelation coefficient; D-W is the Durbin-Watson statistic; and

**Federal Republic of Germany: Equations for the National Saving Rate (s ^{n})^{1}**

Influence of s_{g} | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

Equation | Period | const | labsh | tot | g_{y} | S_{g} | S_{n}(−1) | SEE | ρ | D-W | Third year | Long run | |

1.1 | 1962–89 | −0.106 (1.31) | −260 (4.96) | 0.238 (2.09) | 0.311 (5.70) | 0.305 (3.66) | 0.673 (16.07) | 0.0050 | 0.443 | 0.982 | 0.65 | 0.93 | |

1.2 | 1965–89 | −0.080 (1.01) | −0.272 (5.34) | 0.219 (1.99) | 0.295 (5.25) | 0.302 (3.39) | 0.683 (17.40) | 0.0048 | −0.492 | 0.981 | 0.65 | 0.95 | |

1.3 | 1970–89 | −O.116 (1.23) | −0.240 (4.37) | 0.241 (2.12) | 0.309 (4.56) | 0.329 (3.59) | 0.652 (15.18) | 0.0048 | −0.478 | 0.971 | 0.68 | 0.95 | |

1.4 | 1962–85 | −0.137 (1.31) | −0.216 (2.51) | 0.3.33 (1.90) | 0.312 (5.02) | 0.318 (3.09) | 0.668 (13.38) | 0.0054 | −0.432 | 0.982 | 0.67 | 0.96 | |

1.5 | 1962–89 | −0.140 (1.30) | −0.234 (3.12) | 0.301 (2.08) | 0.332 (4.79) | 0.375 (3.49) | 0.614 (10.96) | 0.0056 | 2.37 | 0.980 | 0.75 | 0.97 | |

1.6 | 1965–89 | −0.101 (0.91) | −0.248 (3.34) | 0.284 (1.99) | 0.303 (4.19) | 0.396 (3.43) | 0.617 (11.16) | 0.0055 | 2.52 | 0.978 | 0.79 | 1.03 | |

1.7 | 1970–89 | −0.173 (11.49) | −0.212 (2.93) | 0.297 (2.22) | 0.351 (4.32) | 0.402 (3.54) | 0.606 (9.55) | 0.0051 | 2.29 | 0.976 | 0.79 | 1.02 | |

1.8 | 1962–85 | −0.216 (1.53) | −0.141 (1.15) | 0.380 (1.72) | 0.345 (4.45) | 0.401 (3.54) | 0.608 (9.55) | 0.0059 | 2.30 | 0.980 | 0.79 | 1.02 |

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; ρ is the first-order autocorrelation coefficient; D-W is the Durbin-Watson statistic; and

**Federal Republic of Germany: Equations for the National Saving Rate (s ^{n})^{1}**

Influence of s_{g} | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

Equation | Period | const | labsh | tot | g_{y} | S_{g} | S_{n}(−1) | SEE | ρ | D-W | Third year | Long run | |

1.1 | 1962–89 | −0.106 (1.31) | −260 (4.96) | 0.238 (2.09) | 0.311 (5.70) | 0.305 (3.66) | 0.673 (16.07) | 0.0050 | 0.443 | 0.982 | 0.65 | 0.93 | |

1.2 | 1965–89 | −0.080 (1.01) | −0.272 (5.34) | 0.219 (1.99) | 0.295 (5.25) | 0.302 (3.39) | 0.683 (17.40) | 0.0048 | −0.492 | 0.981 | 0.65 | 0.95 | |

1.3 | 1970–89 | −O.116 (1.23) | −0.240 (4.37) | 0.241 (2.12) | 0.309 (4.56) | 0.329 (3.59) | 0.652 (15.18) | 0.0048 | −0.478 | 0.971 | 0.68 | 0.95 | |

1.4 | 1962–85 | −0.137 (1.31) | −0.216 (2.51) | 0.3.33 (1.90) | 0.312 (5.02) | 0.318 (3.09) | 0.668 (13.38) | 0.0054 | −0.432 | 0.982 | 0.67 | 0.96 | |

1.5 | 1962–89 | −0.140 (1.30) | −0.234 (3.12) | 0.301 (2.08) | 0.332 (4.79) | 0.375 (3.49) | 0.614 (10.96) | 0.0056 | 2.37 | 0.980 | 0.75 | 0.97 | |

1.6 | 1965–89 | −0.101 (0.91) | −0.248 (3.34) | 0.284 (1.99) | 0.303 (4.19) | 0.396 (3.43) | 0.617 (11.16) | 0.0055 | 2.52 | 0.978 | 0.79 | 1.03 | |

1.7 | 1970–89 | −0.173 (11.49) | −0.212 (2.93) | 0.297 (2.22) | 0.351 (4.32) | 0.402 (3.54) | 0.606 (9.55) | 0.0051 | 2.29 | 0.976 | 0.79 | 1.02 | |

1.8 | 1962–85 | −0.216 (1.53) | −0.141 (1.15) | 0.380 (1.72) | 0.345 (4.45) | 0.401 (3.54) | 0.608 (9.55) | 0.0059 | 2.30 | 0.980 | 0.79 | 1.02 |

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; ρ is the first-order autocorrelation coefficient; D-W is the Durbin-Watson statistic; and

In view of the large fiscal demands that are to emerge in the course of GEMSU and the debate concerning the need for new budgetary measures, Table 1 provides additional information on the influence of the fiscal variable showing the estimated three-year effect of a change in *sg* as well as the long-run effect. Again, the stability across the various sample subperiods is notable. The broad picture that emerges from the first set of equations (1.1 to 1.4) is that, in the first year, about 30 percent of an increase in *s*_{g} is transmitted to *s*_{n}, with the rest offset as the private sector adjusts its saving behavior. As time passes, the effect on *sn* gets larger; in the third year, *s*_{n} has increased by about two thirds of the increase in *s*_{g} and in the long run by almost all of the rise in *s*_{g}. The equations without autocorrelation correction produce an estimate of the effect of *s*_{g}, which is higher both on impact (close to 40 percent) and after three years (three fourths) but not dramatically different. While the short-run offset in the private sector to changes in *s*_{g} reported in equations 1.1 to 1.8 might seem Ricardian in nature, the results appear to be more consistent with a Keynesian interpretation. The weaker influence of changes in *sg* in the short run may well reflect the slow adaptation of private consumption to changes in disposable income;^{5} as time passes, however, with lower disposable income, private agents find that maintaining consumption at the expense of their financial position becomes less feasible and/or less desirable and the adjustment of private consumption is intensified. It is, nevertheless, surprising to find that in the long run almost the entire change in the government saving rate gets absorbed in the national saving rate.

While tests do not indicate a problem with serial correlation in equations 1.5–1.8,^{6} the relatively high value for the first-order autocorrelation coefficient in equations 1.1–1.4 suggests the desirability of exploring alternative lag patterns for the lagged dependent variable. In Table 2, equation 2.1, an additional lag of the dependent variable is added to equation 1.5. The estimated parameter on *s*_{g} is only a little larger in the short run and essentially the same in the medium and long run. Estimates over different subperiods (one of which is reported as equation 2.2) show this model producing lower overall standard errors than is the case for the equations in Table 1. There is, however, less parameter stability across estimation periods. Moreover, over some subperiods, the long-run effects of changes in government saving are implausibly high.^{7}

**Federal Republic of Germany: Additional Equations for the National Saving Rate (s _{n})^{1}**

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The estimate for the constant term is not reported. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; D-W is the Durbin-Watson statistic; and R^{2} is calculated based on original levels, not levels as transformed by the autocorrelation adjustment.

**Federal Republic of Germany: Additional Equations for the National Saving Rate (s _{n})^{1}**

Influence of s_{g} | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

Equation | Period | labsh | tot | g_{y} | S_{g} | S_{g}(−1) | S_{n}(−1) | S_{n}(−2) | SEE | D-W | Third year | Long run | |

2.1 | 1962–89 | −0.316 (0.455) | −0.289 (2.34) | 0.296 (4.88) | 0.418 (4.49) | 0.408 (4.89) | 0.192 (3.01) | 0.0048 | 1.93 | 0.985 | 0.74 | 1.05 | |

2.2 | 1965–89 | −0.374 (6.10) | 0.241 (2.33) | 0.228 (4.11) | 0.524 (5.89) | 0.338 (4.40) | 0.249 (4.26) | 0.0040 | 2.43 | 0.988 | 0.89 | 1.27 | |

2.3 | 1962–89 | −0.208 (2.67) | 0.348 (2.31) | 0.293 (3.76) | 0.432 (3.62) | −0.166 (1.08) | 0.692 (7.55) | 0.0056 | 2.59 | 0.980 | 0.66 | 0.86 | |

2.4 | 1962–89 | −0.294 (4.01) | 0.326 (2.51) | 0.257 (3.94) | 0.461 (4.47) | −0.129 (0.97) | 0.475 (4.37) | 0.186 (2.91) | 0.0048 | 2.15 | 0.985 | 0.68 | 1.00 |

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The estimate for the constant term is not reported. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; D-W is the Durbin-Watson statistic; and R^{2} is calculated based on original levels, not levels as transformed by the autocorrelation adjustment.

**Federal Republic of Germany: Additional Equations for the National Saving Rate (s _{n})^{1}**

Influence of s_{g} | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

Equation | Period | labsh | tot | g_{y} | S_{g} | S_{g}(−1) | S_{n}(−1) | S_{n}(−2) | SEE | D-W | Third year | Long run | |

2.1 | 1962–89 | −0.316 (0.455) | −0.289 (2.34) | 0.296 (4.88) | 0.418 (4.49) | 0.408 (4.89) | 0.192 (3.01) | 0.0048 | 1.93 | 0.985 | 0.74 | 1.05 | |

2.2 | 1965–89 | −0.374 (6.10) | 0.241 (2.33) | 0.228 (4.11) | 0.524 (5.89) | 0.338 (4.40) | 0.249 (4.26) | 0.0040 | 2.43 | 0.988 | 0.89 | 1.27 | |

2.3 | 1962–89 | −0.208 (2.67) | 0.348 (2.31) | 0.293 (3.76) | 0.432 (3.62) | −0.166 (1.08) | 0.692 (7.55) | 0.0056 | 2.59 | 0.980 | 0.66 | 0.86 | |

2.4 | 1962–89 | −0.294 (4.01) | 0.326 (2.51) | 0.257 (3.94) | 0.461 (4.47) | −0.129 (0.97) | 0.475 (4.37) | 0.186 (2.91) | 0.0048 | 2.15 | 0.985 | 0.68 | 1.00 |

^{1}Variable definitions are as follows: *s*_{n} is the net national saving divided by net national product; *g*_{y} is the real rate of growth of net national product; *s*_{g} is general government saving divided by net national product; *labsh* is the share of labor in net national product; and *tot* is the gain in the terms of trade divided by the previous year’s net national product. The equations are estimated by ordinary least squares. The estimate for the constant term is not reported. The numbers in parentheses are t-statistics; SEE is the standard error of the estimate; D-W is the Durbin-Watson statistic; and R^{2} is calculated based on original levels, not levels as transformed by the autocorrelation adjustment.

A feature of the equations that have been reported thus far is that the same pattern of adjustment is forced on all variables; to see how this might have affected the estimated coefficient on *s*_{g}, the lagged value of *s*_{g} was added to equations 1.5 and 2.1 and the results are reported as equations 2.3 and 2.4. This had little material effect on the fit of the equations.

On balance, therefore, it would seem that, other things being constant, about 40 percent of a change in government saving might get reflected in national saving in the short run, with the effect rising to about three fourths in the third year. However, in the short run, other variables are likely to be influenced by changes in budgetary policy in a way that dampens the effect of *s*_{g} on *s*_{n}. In particular, budgetary measures to increase government saving may not only reduce *g*_{y} but also shift the distribution of income toward labor.^{8} They might also produce a terms of trade loss. Over the medium and long term, however, there is no reason to expect that the other explanatory variables will be b>influenced significantly by a change in *s*_{g}. Finally, in interpreting the estimate on *s*_{g}, one should consider the possibility that private saving responds differently, depending on the nature of the change in government saving. It seems plausible, for example, that the effects, particularly in the short term, of changes in government consumption (i.e., direct current expenditure on goods) might be somewhat different than for other elements of the fiscal balance, with, for example, a decline in government consumption having a much greater effect on national saving than an increase in taxes. An attempt to identify such a differential influence in the econometric work was, however, not successful.

#### Simulation of the Effects of GEMSU on Saving

All of the explanatory variables that appear in the equations reported in Tables 1 and 2 are likely to be affected by GEMSU. In this chapter, the saving model discussed above is simulated under assumptions as to how GEMSU might affect these variables. The illustrative baseline scenario adopted is as follows: (1) the labor share in net national product recovers somewhat over the coming years, returning to its 1987 level by 1995;^{9} (2) net output grows by 3½ percent in 1990 and by 2¾ percent a year subsequently; (3) the terms of trade are constant; and (4) government saving, in percent of NNP, changes little from its 1990 level.^{10}

Table 3 presents, in the form of deviations from baseline, an alternative scenario for these variables reflecting considerations of how they might be affected by the process of GEMSU:^{11}

**West Germany: Assumptions for the Effects of GEMSU on Selected Variables**

(Relative to baseline)

^{1}Percentage points.

^{2}Percent of net national product.

**West Germany: Assumptions for the Effects of GEMSU on Selected Variables**

(Relative to baseline)

Growth Rate of Net National Product^{1} | Change in Terms of Trade^{1} | Real Long-Term Interest Rate^{1} | Labor Share^{2} | |
---|---|---|---|---|

1990 | 0.7 | 1.5 | 0.7 | −0.5 |

1991 | 1.0 | 0.3 | 0.8 | −1.0 |

1992 | −0.2 | 0.3 | 0.8 | −1.5 |

1993 | −0.2 | — | 0.7 | −2.0 |

1994 | −0.2 | −0.5 | 0.7 | −2.0 |

1995 | −0.1 | −0.5 | 0.5 | −1.5 |

1996 | — | −0.5 | 0.3 | −1.0 |

1997 | — | −0.4 | 0.1 | −0.5 |

1998 | — | −0.2 | — | — |

1999 | — | −0.2 | — | — |

2000 | — | — | — | — |

^{1}Percentage points.

^{2}Percent of net national product.

**West Germany: Assumptions for the Effects of GEMSU on Selected Variables**

(Relative to baseline)

Growth Rate of Net National Product^{1} | Change in Terms of Trade^{1} | Real Long-Term Interest Rate^{1} | Labor Share^{2} | |
---|---|---|---|---|

1990 | 0.7 | 1.5 | 0.7 | −0.5 |

1991 | 1.0 | 0.3 | 0.8 | −1.0 |

1992 | −0.2 | 0.3 | 0.8 | −1.5 |

1993 | −0.2 | — | 0.7 | −2.0 |

1994 | −0.2 | −0.5 | 0.7 | −2.0 |

1995 | −0.1 | −0.5 | 0.5 | −1.5 |

1996 | — | −0.5 | 0.3 | −1.0 |

1997 | — | −0.4 | 0.1 | −0.5 |

1998 | — | −0.2 | — | — |

1999 | — | −0.2 | — | — |

2000 | — | — | — | — |

^{1}Percentage points.

^{2}Percent of net national product.

(1) The expansion of output is initially faster than under the baseline, owing to the surge in demand from east Germany: growth is boosted by ¾ of 1 percentage point in 1990 and by 1 percent in 1991. Subsequently, growth falls a little below baseline, as the demand impulse from east Germany lessens and the effects of higher real interest rates and the appreciated exchange rate continue to pass through to other components of demand. The level of output remains above baseline throughout the scenario.

(2) The higher rate of return to capital and the closer links with the labor market in east Germany are likely to put downward pressure on the labor share in west Germany. Thus, this scenario assumes that, instead of recovering somewhat over the next few years, the labor share stays broadly unchanged before starting to rise toward its longer run level in the second half of the 1990s.

(3) The large increase in demand for west German goods is expected to result in a real appreciation of the deutsche mark, producing a terms of trade gain. As the demand shock subsides, the terms of trade gain is likely to be reversed. The illustrative scenario used here assumes a 2 percent gain in the terms of trade gain in 1990–92, which then unwinds gradually in the following years.

It should be stressed that the alternative scenario is purely illustrative, as there are many uncertainties concerning how GEMSU will affect west Germany, including any effects it might have on saving behavior. Nevertheless, it can be used to give some sense of the possible response of saving. The simulations of saving under the assumptions of Table 3 are presented in Table 4 using the coefficients in equation 1.5.^{12} The broad conclusion is that the boost to saving coming from higher output is significant, representing close to one half of the increase in the level of NNP in west Germany in 1991 (relative to the baseline); however, it still represents a relatively small share of the possible resource demands from east Germany and, moreover, subsides after 1991. The stimulus to saving from the changes in the terms of trade is relatively small and is reversed as the increase in the terms of trade unwinds.^{13} The calculations would suggest that, over the medium term, changes in the labor share have the greatest potential for stimulating saving, though it should be stressed that the calculations are sensitive to the assumed scale of the effect of GEMSU on the labor share. Given the lags in the adjustment of saving behavior (a high value for the lagged dependent variable) and the assumption that changes in labor share resulting from GEMSU are spread over a number of years, the implications for saving of these changes materialize gradually, reaching a peak in the mid-1990s. Putting all three variables together (the effects on growth, the terms of trade, and the labor share), one sees that the boost to saving amounts to over 1 percent of NNP a year in the period 1991–95; by the mid-1990s, it is due almost entirely to the assumed path of the labor share.

**West Germany: Effects of GEMSU on Net National Saving ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

^{1}Excluding fiscal transfers to east Germany.

**West Germany: Effects of GEMSU on Net National Saving ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

1990 | 1991 | 1992 | 1993 | 1994 | 1995 | 1996 | 1997 | 1998 | 1999 | 2000 | ||
---|---|---|---|---|---|---|---|---|---|---|---|---|

Higher growth | 7 | 16 | 10 | 5 | 3 | 2 | 2 | 2 | 3 | 3 | 3 | |

Lower labor share | 2 | 7 | 12 | 18 | 22 | 23 | 20 | 16 | 10 | 6 | 4 | |

Changes in terms of trade | 3 | 3 | 2 | 2 | — | −1 | 1 | −1 | −2 | −2 | −2 | |

Higher growth, lower labor share, and changes in terms of trade | 12 | 25 | 24 | 26 | 25 | 23 | 21 | 17 | 11 | 7 | 5 | |

Memorandum items: | ||||||||||||

Effect of measures to increase government saving(1 percent of NNP) | 8 | 13 | 17 | 20 | 21 | 23 | 24 | 25 | 26 | 26 | 27 | |

Effect of higher growth scenario on the level of NNP | 14 | 36 | 32 | 29 | 25 | 23 | 24 | 24 | 25 | 26 | 27 |

^{1}Excluding fiscal transfers to east Germany.

**West Germany: Effects of GEMSU on Net National Saving ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

1990 | 1991 | 1992 | 1993 | 1994 | 1995 | 1996 | 1997 | 1998 | 1999 | 2000 | ||
---|---|---|---|---|---|---|---|---|---|---|---|---|

Higher growth | 7 | 16 | 10 | 5 | 3 | 2 | 2 | 2 | 3 | 3 | 3 | |

Lower labor share | 2 | 7 | 12 | 18 | 22 | 23 | 20 | 16 | 10 | 6 | 4 | |

Changes in terms of trade | 3 | 3 | 2 | 2 | — | −1 | 1 | −1 | −2 | −2 | −2 | |

Higher growth, lower labor share, and changes in terms of trade | 12 | 25 | 24 | 26 | 25 | 23 | 21 | 17 | 11 | 7 | 5 | |

Memorandum items: | ||||||||||||

Effect of measures to increase government saving(1 percent of NNP) | 8 | 13 | 17 | 20 | 21 | 23 | 24 | 25 | 26 | 26 | 27 | |

Effect of higher growth scenario on the level of NNP | 14 | 36 | 32 | 29 | 25 | 23 | 24 | 24 | 25 | 26 | 27 |

^{1}Excluding fiscal transfers to east Germany.

The discussion up to now has not focused on how GEMSU could affect national saving in west Germany through its effect on government saving. While transfers to east Germany are likely to reduce government saving,^{14} it is questionable whether this will have a major influence on consumption in west Germany as it does not affect the income streams (including public goods) of private individuals. ^{15} Thus, it would seem inappropriate to mechanically include in the simulation changes in *s*_{g} due to these transfers.^{16} A memorandum item in Table 4, however, shows how national saving might be expected to adjust if new budgetary measures (affecting income streams in west Germany) were enacted to boost government saving by 1 percent of NNP per annum. While a large part of the increased government saving is initially offset in the private sector, the induced change in national saving would over time represent a significant source of additional saving.

### The Effects of GEMSU on Investment in West Germany

#### Econometric Estimates

The effects of GEMSU on investment are examined using a study of machinery and equipment investment reported in McDonald (1988). While this focused only on machinery and equipment, it may be a good basis for discussing the likely effects on business sector investment. ^{17} In 1989, however, the business sector accounted for only 63 percent of total fixed investment, with government (11 percent) and residential construction (26 percent) also being significant components. Estimates for the equation explaining machinery and equipment investment, using semiannual data, are presented in Table 5. The dependent variable is gross fixed investment divided by the net capital stock and this is related to the output-capital ratio, the user cost of capital, and accelerator terms to capture short-run cyclical effects on investment.^{18} The coefficient estimates are relatively stable up to 1985 (the time span explored in the original study), but the model has more difficulty in 1986–87, a period of almost unprecedented slow growth in exports; as a result of this slow export growth, there was significant business sector pessimism, particularly in parts of the manufacturing sector dependent on exports. Adding a dummy variable for the four observations in 1986–87 improves the fit and results in parameter estimates closer to those for the earlier estimation period.

**Federal Republic of Germany: Equations for Machinery and Equipment Investment ^{1}**

^{1}The dependent variable is *I/K*. Variables are as follows: *I* is real gross investment in machinery and equipment by the private sector; *K* is the real net stock of machinery and equipment at the beginning of the period; *r/p* is the user cost of capital relative to the GNP deflator; *Y* is rear GNP; *D74* is a dummy to capture the effects of the first oil shock on investment; *D8384* is a dummy to capture the effects of the temporary investment scheme for machinery and equipment that was introduced in 1982 and expired at the end of 1983; *D8687* is a dummy to capture the “export pessimism” of 1986–87; the estimated coefficients on the seasonal dummy and constant are not reported. The estimate of the elasticity of substitution is presented in the last column. The numbers in parentheses are t-statistics. The Chi-square test in the second last column is for first-order correlation. The critical value at the 5 percent level is 3.84. A delta sign Δ in front of a variable denotes a first derivative and two deltas denotes a second difference. Equations are estimated with semiannual data, using ordinary least squares. The estimate for the constant term is not reported.

**Federal Republic of Germany: Equations for Machinery and Equipment Investment ^{1}**

Equation | Period | In(r/p)_{−1} | In(Y/K)_{−1} | ΔΔInY | ΔInY(−2) | D74 | D8384 | D8687 | (I/K)_{−1} | SEE | D-W (Chi-sq.(1)) | σ | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

6.1 | 1965–76 | −0.065 (2.77) | 0.108 (2.38) | 0.159 (4.46) | 0.094 (1.81) | −0.007 (2.68) | 0.397 (2.98) | 0.0030 | 0.945 | 1.44 (1.96) | 0.60 | ||

6.2 | 1965–80 | −0.069 (4.47) | 0.113 (3.80) | 0.156 (5.26) | 0.103 (3.03) | −0.007 (4.00) | 0.340 (3.39) | 0.0027 | 0.945 | 1.58 (1.27) | 0.61 | ||

6.3 | 1965–85 | −0.066 (5.48) | 0.112 (4.94) | 0.157 (6.53) | 0.108 (3.86) | −0.007 (4.72) | 0.007 (3.03) | 0.318 (4.22) | 0.0024 | 0.953 | 1.60 (1.84) | 0.59 | |

6.4 | 1965–89 | −0.045 (5.36) | 0.090 (4.41) | 0.141 (5.83) | 0.117 (4.18) | −0.007 (4.41) | 0.007 (3.01) | 0.362 (4.91) | 0.0025 | 0.941 | 1.51 (3.87) | 0.50 | |

6.5 | 1965–89 | −0.060 (6.30) | 0.109 (5.39) | 0.144 (6.40) | 0.103 (3.90) | −0.007 (4.87) | 0.007 (3.12) | −0.004 (2.72) | 0.321) (4.57) | 0.0024 | 0.950 | 1.52 (2.83) | 0.55 |

^{1}The dependent variable is *I/K*. Variables are as follows: *I* is real gross investment in machinery and equipment by the private sector; *K* is the real net stock of machinery and equipment at the beginning of the period; *r/p* is the user cost of capital relative to the GNP deflator; *Y* is rear GNP; *D74* is a dummy to capture the effects of the first oil shock on investment; *D8384* is a dummy to capture the effects of the temporary investment scheme for machinery and equipment that was introduced in 1982 and expired at the end of 1983; *D8687* is a dummy to capture the “export pessimism” of 1986–87; the estimated coefficients on the seasonal dummy and constant are not reported. The estimate of the elasticity of substitution is presented in the last column. The numbers in parentheses are t-statistics. The Chi-square test in the second last column is for first-order correlation. The critical value at the 5 percent level is 3.84. A delta sign Δ in front of a variable denotes a first derivative and two deltas denotes a second difference. Equations are estimated with semiannual data, using ordinary least squares. The estimate for the constant term is not reported.

**Federal Republic of Germany: Equations for Machinery and Equipment Investment ^{1}**

Equation | Period | In(r/p)_{−1} | In(Y/K)_{−1} | ΔΔInY | ΔInY(−2) | D74 | D8384 | D8687 | (I/K)_{−1} | SEE | D-W (Chi-sq.(1)) | σ | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

6.1 | 1965–76 | −0.065 (2.77) | 0.108 (2.38) | 0.159 (4.46) | 0.094 (1.81) | −0.007 (2.68) | 0.397 (2.98) | 0.0030 | 0.945 | 1.44 (1.96) | 0.60 | ||

6.2 | 1965–80 | −0.069 (4.47) | 0.113 (3.80) | 0.156 (5.26) | 0.103 (3.03) | −0.007 (4.00) | 0.340 (3.39) | 0.0027 | 0.945 | 1.58 (1.27) | 0.61 | ||

6.3 | 1965–85 | −0.066 (5.48) | 0.112 (4.94) | 0.157 (6.53) | 0.108 (3.86) | −0.007 (4.72) | 0.007 (3.03) | 0.318 (4.22) | 0.0024 | 0.953 | 1.60 (1.84) | 0.59 | |

6.4 | 1965–89 | −0.045 (5.36) | 0.090 (4.41) | 0.141 (5.83) | 0.117 (4.18) | −0.007 (4.41) | 0.007 (3.01) | 0.362 (4.91) | 0.0025 | 0.941 | 1.51 (3.87) | 0.50 | |

6.5 | 1965–89 | −0.060 (6.30) | 0.109 (5.39) | 0.144 (6.40) | 0.103 (3.90) | −0.007 (4.87) | 0.007 (3.12) | −0.004 (2.72) | 0.321) (4.57) | 0.0024 | 0.950 | 1.52 (2.83) | 0.55 |

^{1}The dependent variable is *I/K*. Variables are as follows: *I* is real gross investment in machinery and equipment by the private sector; *K* is the real net stock of machinery and equipment at the beginning of the period; *r/p* is the user cost of capital relative to the GNP deflator; *Y* is rear GNP; *D74* is a dummy to capture the effects of the first oil shock on investment; *D8384* is a dummy to capture the effects of the temporary investment scheme for machinery and equipment that was introduced in 1982 and expired at the end of 1983; *D8687* is a dummy to capture the “export pessimism” of 1986–87; the estimated coefficients on the seasonal dummy and constant are not reported. The estimate of the elasticity of substitution is presented in the last column. The numbers in parentheses are t-statistics. The Chi-square test in the second last column is for first-order correlation. The critical value at the 5 percent level is 3.84. A delta sign Δ in front of a variable denotes a first derivative and two deltas denotes a second difference. Equations are estimated with semiannual data, using ordinary least squares. The estimate for the constant term is not reported.

#### Simulation of the Effects of GEMSU on Investment

To illustrate possible effects of GEMSU on machinery and equipment investment, the investment equation described above has been simulated on the basis of the following assumptions: (1) Output growth follows the path laid out in Table 3; (2) the real interest rate rises by about ¾ of 1 percentage point in 1990–94 and then falls back gradually toward the baseline level;^{19} and (3) the price of investment goods rises by 1 percent relative to other goods.

Table 6 shows the effects of these alternative assumptions relative to a baseline for machinery and equipment investment which assumes a constant real cost of capital and the baseline growth pattern assumed for the saving simulations. The broad conclusion is that in the early years the output-accelerator effects offset the investment dampening effects originating in the higher cost of capital.^{20} It would thus seem that lower business sector investment in west Germany is unlikely to be an important source of increased resources for investment in east Germany. It is also clear that assuming a sharper rise in interest rates and in the prices of capital goods would moderate this conclusion but not fundamentally alter it.

**West Germany: Effects of GEMSU on Gross Machinery and Equipment Investment ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

^{1}Based on assumptions in Table 3.

^{2}Higher real interest rates and relative price of investment goods that are subsequently reversed.

**West Germany: Effects of GEMSU on Gross Machinery and Equipment Investment ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

Effect of Higher Growth | Effect of Higher Cost of Capital^{2} | Combined Growth and Cost of Capital Effects | |
---|---|---|---|

1990 | — | — | — |

1991 | 4 | −2 | 2 |

1992 | 4 | −4 | — |

1993 | 2 | −3 | −1 |

1994 | — | −2 | −2 |

1995 | — | −1 | −1 |

1996 | — | — | — |

1997 | 1 | 1 | 1 |

1998 | 1 | 1 | 1 |

1999 | 1 | 1 | 1 |

2000 | 1 | 1 | 2 |

^{1}Based on assumptions in Table 3.

^{2}Higher real interest rates and relative price of investment goods that are subsequently reversed.

**West Germany: Effects of GEMSU on Gross Machinery and Equipment Investment ^{1}**

(Deviations from baseline, in billions of 1990 deutsche mark)

Effect of Higher Growth | Effect of Higher Cost of Capital^{2} | Combined Growth and Cost of Capital Effects | |
---|---|---|---|

1990 | — | — | — |

1991 | 4 | −2 | 2 |

1992 | 4 | −4 | — |

1993 | 2 | −3 | −1 |

1994 | — | −2 | −2 |

1995 | — | −1 | −1 |

1996 | — | — | — |

1997 | 1 | 1 | 1 |

1998 | 1 | 1 | 1 |

1999 | 1 | 1 | 1 |

2000 | 1 | 1 | 2 |

^{1}Based on assumptions in Table 3.

^{2}Higher real interest rates and relative price of investment goods that are subsequently reversed.

With regard to government investment, while some areas for economy might be found in light of the high infrastructural demands in the east, it is also likely that the authorities will have to make additional investments to improve infrastructure in areas in west Germany close to the former border between the FRG and the GDR (particularly to enhance infrastructural links with the east) and to respond to demand resulting from immigration. Thus, the scope for a significant fall in investment to provide additional resources for east Germany would seem to rest with residential homebuilding; however, given that residential construction is only one fourth of total investment, the impact would have to be quite large to have a marked effect. Moreover, increased immigration in 1990–91 can be expected to have a positive influence on residential construction activity.

## References

Kaldor, Nicholas, “Alternative Theories of Distribution,”

, Vol. 23 (1955), pp. 83-100.*Review of Economic Studies*Laursen, Svend, and Lloyd A. Metzler, “Flexible Exchange Rates and the Theory of Employment,”

, Vol. 32 (November 1950), pp. 281-99.*Review of Economics and Statistics*McDonald, Donogh, “An Econometric Analysis of Machinery and Equipment Investment in Germany” (unpublished; Washington: International Monetary Fund, 1988).

McDonald, Donogh “The German Current Account from a Saving-Investment Perspective,” (unpublished; Washington: International Monetary Fund, 1989).

^{}3

A notable feature of the model is the role played by the distribution of income and the terms of trade, following ideas originating, respectively, in Kaldor (1955) and Laursen and Metzler (1951).

^{}4

The levels of the parameter estimates, however, differ to some extent between the two sets of equations. The somewhat greater variability in the estimated coefficient on *deltot* than for the other variables should be seen in the context of the less precisely estimated coefficients (high standard errors relative to coefficient size): within each set of equations, the various coefficient estimates for *deltot* are not significantly different from one another. Similarly, the coefficient on *labsh* in equation 1.8 is not significantly different from the estimates in equations 1.5–1.7.

^{}5

In part, this process is likely to depend on the perceived permanence of the change in *s*_{g}; no attempt has been made in the econometric work to distinguish between temporary and permanent budgetary developments.

^{}6

A Lagrange-multiplier test rejects the presence of serial correlation of first and higher orders.

^{}7

In some cases the short-run effect of *s*_{g} is in the region of one half, for example in equation 2.2 in Table 2, but this is associated generally with unacceptably high estimates for the long-run effect of *s*_{g}.

^{}9

This represents a relatively moderate recovery compared with the sharp decline that occurred in the 1980s; see Chapter II, section on “Employment, Wages, and Prices”.

^{}10

The calculations of the effects of GEMSU on saving reported below are not, however, very sensitive to the choice of baseline.

^{}11

The alternative scenario outlined in Table 3 is designed to reflect broadly some of the macroeconomic effects of GEMSU simulated in the different scenarios in Chapter VI. It should be noted, however, that the savings and investment functions used here are different from those in the model used to generate the simulation output in Chapter VI.

^{}12

Using one of the other estimated equation>s does not produce calculations that are markedly different.

^{}13

A change in the terms of trade has no long-term effect on the level of saving (expressed in terms of domestic output); however, a change in output growth does have a long-term effect as it results in a higher level of output.

^{}14

For illustrative purposes, it is assumed that the fiscal accounts in east and west Germany remain completely separate, except for transfers.

^{}15

As noted earlier, the results of the study of saving suggest that slow adjustment to changes in income (rather than “Ricardian” influences) may be the principal factor underlying the pattern of private saving adjustment to changes in government saving.

^{}16

National saving declines one for one with the increase in government transfers to east Germany but this fall in saving directly supplies part of the external resource needs of east Germany. If increased government transfers from west to east Germany were to increase private saving in the west, in the short run, (as would be suggested by a mechanical application of the econometric estimates), additional resources would be freed for use in east Germany.

^{}18

The theoretical basis for the model is discussed in McDonald (1988). Essentially, investment is determined by the difference between the actual capital-output ratio and the desired capital-output ratio (which is related to the user cost of capital), with the speed of adjustment being influenced by cyclical factors.

^{}19

It is assumed that the increase in the required yield on internal funds of enterprises is one half of the increase in the cost of external funds, assuming that the marginal effective tax rate is about half. This assumption on the change in the cost on internal funds also seems reasonable in light of sample magnitude of these variables: in the 1980s, the proxy for the internal cost of funds (the dividend yield) was about one half of the real bond rate. Moreover, for the large number of small and medium-sized firms in Germany, the link of investment decisions to yields in the capital market may not be particularly tight.