Experiments With Fiscal Policy Parameters on a Micro to Macro Model of The Swedish Economy

It is impossible to explain all algorithms of the model here.¹⁴ To illustrate, let us define a subset [Ɵ] of endogenous variables (micro and macro) that we are interested in, say those macro variables that are studied in the tax experiments and listed vertically in Tables 3 and 4 (W, PDOM, M,…).

Each of these variables chosen (and others as well) can be represented by the general function:

This function is complex and not well behaved. It is based on selection, highly nonlinear and initial state dependent, and hence, of course, generally not continuous, and differentiable. Most of the theoretical and empirical work of this project will consist of ascertaining the properties of Γ at various points in X space. This work is now going on and its empirical side will dominate the next phase of the project (see below). To get a feeling for the properties of Γ let us divide (X) into subsets of variables:

X1 : hierarchical ordering of algorithms, illustrated diagrammatically in Figures 1 and 2.

X2: structural (technical) parameters. Macro parameters and identities. Coefficients that are kept constant throughout simulation or are varied exogenously.

X3: time reaction parameters¹⁵ (measuring how the individual firm transforms its price, wage, and sales information into expectations, how fast it changes its exports ratio in response to foreign-domestic price differentials, how it reacts when raiding or being raided in the labor market, how it determines its offering wages and prices, the number of searches it is allowed in labor and product markets, and so on for in all 20 parameters).

X4 : starting positional data matrix (micro and macro). Exogenous.

XS : historic input matrix. Exogenous until starting time

X7 : other exogenous input variables.

The Ɵ = Γ(X) structure of the model is graphically illustrated in Figure 4. If we look at one algorithm, at one particular point in time, we will find that most decisions in the model respond directly to classical expected price differentials (market disequilibria) leading to an expected improvement in business profits in the short run as well as in the long run, that will be realized ex post, if not countered by market reactions. The general target of a business unit in the model is to maintain or improve its expected profit margin (MIP) positions, and to stop responding when certain criteria are satisfied and/or when a certain time has elapsed. These price-reaction functions are very simply expressed. But they are numerous, and when all interaction within a period (a quarter) is over most Ɵ will have become functions of most X and past Ɵ. This complexity is further increased when we move from period to period.

Figure 4

Download asset Open asset

Hence the causal ordering of the entire system at the micro level (X1) is critical for understanding the dynamics of the entire model economy. The decision structure within the firm, as depicted in Figure 2, and the search order in the labor and product markets are most important. Profit targeting, for instance, has a strong impact on cyclical as well as long-run growth properties of the entire system. Feeding back information from the employment or inventory parts of the production system of the firm means less intensive enforcement of targets and results in very different behavior. Similarly, some simple experiments have shown that the way we loop together the sales expectations/ production planning, and inventory adjustments in Figure 2 can generate very different cyclical patterns. Since product, labor and financial market expectations are integrated withing the business plans of firms, the core of the integrated cyclical and growth dynamics of the entire model economy is defined by the links (over time and within Figure 2) between price expectations and profit targeting on the one hand, and wage setting in the labor market and interest sensitive investment determination, on the other. Any change in the interlinking of these modules will strongly impact economy wide behavior. Perhaps the most illustrative example is when we close off labor market search between groups of firms or markets entirely. If relative product prices in foreign markets (exogenous) develop differently, a very uneven labor income distribution soon develops between closed-off sectors, and profit and growth patterns change significantly. This “theoretical" property alone emphasizes the importance of the market price setting in total economic systems behavior and the need for more empirical research in this area. On the internal planning and decision structure of the firm the empirical specification is as well founded as it can be, owing to a 5-year interview study (Eliasson, 1976a) that preceded this project and that has strongly influenced the design of the firm model. On the market side the availability of relevant, systematized information is embarrassingly scanty. This means that an effort to become more knowledgeable about wage setting and mobility in the labor market at the micro level is planned and will be designed on the format of the model.

Figure 5 shows how the public sector interacts with the rest of the model via its taxes, transfers, spending, and borrowing (lending).

Figure 5

Download asset Open asset

Firms’ profits are taxed at the individual firm level and we apply calculated effective tax rates to account for Swedish fiscal depreciation rules. Likewise, the payroll tax (WTAX) is applied as wages are paid out. We here use the nominal rates on a properly specified tax base.

Similarly the value-added tax (VATAX) is applied at the going rate when the households purchase goods.

The problem of relevant specification refers mainly to the income tax (ITAX), and its progressiveness, as well as to transfer payments (TRANS). The model is equipped with a macro income tax function estimated by a simulation method in the highly disaggregated tax model developed by Jakobsson and Normann (1974). The tax rate function is of the following form:

where L is the number of taxpayers; W is average "total" income; A is a tax parameter (exogenous); and e is a tax elasticity representing the degree of progression.¹⁶

With these notations the average income tax paid will be A∙W^e. By entering A and e exogenously for each year, the fraction TXI of income paid as income tax will depend on how much income is endogenously generated in the model. Hence, varying the value-added or payroll tax or the TRANS rates will exercise a feedback effect on total household income both through the whole economic system and through the ITAX function above.¹⁷

Similarly, TRANS payments are entered exogenously as a fraction of total public expenditure. It should be noted here that in Sweden some transfer payments are subject to income taxation (e.g., pensions), and some are not. We disregard this and apply a uniform income tax rate on all taxable income each period. All transfer payments are treated as taxable income in the model to keep detail within limits. It should not affect the empirical results to be reported.

We should also remember that the public sector includes the large obligatory, supplementary pension scheme, which means that payments to and from the pension (ATP) fund are treated as payroll taxes and transfer payments respectively. TRANS also includes unemployment compensation, assumed to be 60% of the current average wage and available to all unemployed.

To give the reader an idea of the dimensions of the Swedish tax world, Table 3 presents the tax conditions of two families in 1969 and 1975. In both cases there are two children and only one income earner. We have chosen the average income of a skilled (male) worker and a well-paid salaried worker with a before tax income about twice that of the worker.

Even though simple and transparent in principle, by virtue of the large number of heterogeneous micro agents that compete in markets the model requires complex specification and an enormous amount of data. Since market competition cannot occur without this heterogeneity there are no ways to avoid this complexity. Microsimulation is also the perfect tool to deal with it both theoretically and empirically. The full potential of the model in this respect (for instance "tailor-made" parameter specifications of individual firms) will however have to wait for data to become available.

Numerical specifications required by the model are directly observable. All needed data is, however, not currently available. Since the start of the theoretical part of the project a parallel database project has therefore been going on that aims at gathering the necessary information. The idea of this model has been to build it on the format of national accounts (NA) statistics at the macro level and the internal accounting systems of the firms at the micro level. We believe that potential access to high quality firm data and internal "technical" coefficients will compensate handsomely in the long run to alleviate the estimation problems associated with these complex models (see below). To run the model over historic time, however, we lack much of the required data. Also, some of the required data on expectations is currently not being collected. There is no good conventional method of estimating the parameters of the expectations functions, except by directt questioning of firms.

Parallel to the start of work on specifying the model, a data-gathering process began. The annual Planning Survey of the Federation of Swedish lndustries was designed on the format of the model in 1975 and in close contact with firms. We now have a 3-year time series on a large number of in dividual decision units with more than 200 employees in Swedish manufacturing. The sample consists of approximately 250 units that are "measured" each of the three years. Among other things this survey collects data on the coordinates of points A, B, and D in Figure 3. This information allows us to approximate the production frontier QFR(L) at initialization for each firm.¹⁸ In a few years we should have enough time series panel data to estimate the relationships between investment and γ and QTOP for individual firms as well.

A parallel project to collect a detailed database on the 40 largest Swedish corporations has been carried out at the lndustrial lnstitute for Economic and Social Research (IUI). We will also be able to draw on two other IUI studies to handle the foreign operations of these firms. (Space does not allow for more de tail on the micro databases to be presented here). To be complete, we also need a matching and consistent macro database to define the operating levels of the model in terms of Swedish national accounts statistics.

One thing that we have learned from work on this model is that a consistent demand and production measurement system is imperative for the proper functioning of a model with many linkages across firms and over time. It has been necessary to make substantial adjustments of public Swedish NA statistics to obtain satisfactory consistency in this respect,¹⁹ even though the ambition has been to calibrate the model so that its scale conforms with the NA size of the Swedish economy. We have learned that high quality of measurement is imperative especially when inter­ firm markets and inventories with input-output coefficients are entered exogenously. Thus, exogenous coefficients brought in from official statistical sources have been found not to be consistent with the rest of the database.

The macro database has been prepared simultaneously with the model. This has not been possible for the micro database. To be able to start experimenting with the model and to calibrate it against historic time series data, an interim synthetic firm database has been put together. This database currently consists of a very small number of real firms and split-up macro aggregates for each market. Whenever we have had the information on cross-sectional features, we have entered it. The experiments to be reported on below have been run on this mostly synthetic, micro data base that (for historic time) adds up to Swedish national accounts statistics supplemented with available cross-sectional information. On the estimation side, most of the parameters have been ascertained in the form of ratios from direct measurement. This goes for all the input-output coefficients. Each firm in one market has the same input-output structure on the purchasing side, even though the ratio of purchases to value added varies between firms.²⁰

The household demand system is a modif1ed version of a Stone-type expenditure system (1954). We rely on somewhat modified parameter estimates on such a system on Swedish macro data from Dahlman and Klevmarken (1971). These parameters are entered as a priori hypotheses. The system has a habit-formation feature for each consumption category which rests on the same idea as the permanent-income hypothesis of Milton Friedman. A simultaneous handling of savings and stocks of consumer durables has been added as a theoretical novelty. It should be recalled here that modeling short-period change at the micro level rids us of a number of coefficients that normally appear in macro models. They are replaced by sequential orderings and feedbacks. The estimation problem centers on the 20 time-reaction parameters called X3 and exemplified above. One satisfactory property of these 20 parameters is that they can be grouped into some that operate only (or roughly so) on long term trends, some that affect the cycle only, and a residual group that affects both dimensions. We will be able to estimate some of them for individual firms in the future by conventional econometric techniques.

For the time being we can only ascertain them -by trial-and-error experiments and checks against macro data (manually or by automated computer algorithms).²¹ It sometimes (not often) happens that we find double or multiple parameter combinations that satisfy test criteria, and we cannot discriminate between them, except by a priori judgment. This is time-consuming work. An element of subjectivity also enters in a more manifest way than in conventional econometric work. I would like to add as a personal note, however, that in large-scale modeling of this kind, experience and judgment should take priority over mechanical, statistical procedure. It is easy to understand how important it is to have a high-quality and consistent statistical base to work from, since this database contains most of the numerical specification that determines the properties of the entire model economy.

This should give a flavor of the current status of the model which we regard having something empirically meaningful to tell on a Sweden like industrial economy. For the time being the model is capable of generating 5 to 20 year growth trends for a spectrum of macro variables with satisfactory accuracy. It traces price, wage and profit cycles well. We are currently somewhat uncertain about the model’s capability of catching short-term (quarter to quarter and year to year) change well. One satisfactory feature of the model, however, is that the 20 parameters can be grouped into some that operate only (or roughly so) on long term trends, some that affect cycles only and a group that affect both. One piece of observational evidence, which is comforting as regards realism on the cyclical side, is that exogenous step impulses (not large ones) in exogenous variables and policy parameters tend to spin off gradually dampened 4- to 6- year cycles in the model economy. This property has a bearing on the experiments to follow.

We have chosen to present a series of fiscal policy experiments on the entire model as described above, and more particularly those on the Swedish tax system. This should be of special interest for two reasons. First, the share of total resources taken out by the public sector (50% of GNP in 1976), which is partly used up and partly redistributed, is one of the highest in the world. This means that even minor modifications in the tax structure may have sizable effects upon the economy. Second, the very heaviness of the average tax burden raises particular problems of tax evasion and control. For that reason, there has been a lively discussion about modifying the tax system or even changing it radically.

We will study the effects on growth, inflation, and industrial structures of a shift in emphasis from one tax type to another. To keep the number of pages down we will limit ourselves to two tax categories, the value-added tax (VATAX) and the pay roll tax (WTAX).

We will also carry out a sensitivity analysis to study the relative importance of the two fiscal parameters. The simulation results will be compared with those of our current reference case, which tracks the growth paths of the variables studied 1968-1975 satisfactorily for this experimental purpose. The model has been loaded with the actual nominal tax parameters for the period 1968 through 1975, and all fiscal and other economic data have been correctly and consistently entered at the micro firm level and at the national accounts level at starting time. We should recall that numerical consistency throughout at starting time has been found to be a critical prerequisite for the proper functioning of the model. We recall that the model forecasts historical, long-run growth trends of major macro variables well and prices, wages, and profits quite well, both in the short and the long runs. We think that we can fairly safely say that the results indicate what would have happened to a Sweden like economy over the 8-year period 1968 through 1975 if this or that fiscal measure had been enacted as described. We are, however, somewhat uncertain about the quarter-to-quarter effects and will not report such information except for illustrative purposes.

Recall that the model, being a disequilibrium system, responds quite differently to policy measures or whatever exogenous disturbances it is subjected to, depending upon its positional description when this happens. This is a most desirable property. The problem is only to see to it that the data set that positions the model economy initially in 1969 and the exogenous inputs it runs on are consistently and relevantly specified. The year 1969 witnessed the recovery phase of a business cycle that peaked in 1970.

There are three important systems properties that we should mention before we proceed, since they explain some unexpected experimental results. First, under normal circumstances most firms hoard labor to some extent (see Figure 3). This means that firms experiencing a demand increase can increase supplies and productivity simultaneously for a while. This is a typical cyclical phenomenon. Only on the production frontier does productivity decrease as more labor is recruited to work with on the margin less productive machinery (above B in Figure 3). We have learned from model experiments that individual firms only rarely operate on that end of their production frontiers, and these experimental results are supported by preliminary evidence from the Planning Survey of the Federation of Swedish Industries. Second, an exogenous step impulse tends to spin off a 4- to 6-year cycle in the model economy. The system hence behaves according to the Le Chatelier principle at least in some time dimensions. The initial effect in one direction will reverse itself through an endogenous counter-response of the system transmitted through price responses. These properties are conventionally assumed away in mainstream economic modeling that lack explicit markets: in Keynesian models through the absence of supply feedbacks and in neoclassical models through assumed perfect foresight. The property arises in this model through mistaken expectations, a consequent disturbed market price signaling system, and corrective effects on the rates of return, investment and capacity growth. Third, one particularly important instance of this property is that sudden, unexpected inflationary shocks (not necessarily large ones) tend to have long-lasting downward effects on economic growth after an initial positive effect.²² lt is difficult to test for this property directly. So far, my own attempts to ascertain its relevance have however turned out evidence in the affirmative.²³ This property was furtheermore suggested by Professor Friedman in his 1976 Nobel lecture. Some econometric models report negative growth coefficients for the rate of change in the rate of inflation but this is too crude a test for the phenomenon we are discussing to be more than slightly helpful. Cagan's (1974) results that common stock maintains its real value in the Iong run but that "stocks may take many years to catch up to an inflationary episode" (15 years is the median time in an international comparison) refer more directly to the machinery in our model and hence is more supportive.

Throughout this empirical application we will be particularly interested in the effects of a fiscal parameter change on the absolute price level (inflation), total output, and the functional income distribution (profits and wages shares).²⁴

We have defined most policy changes as a once-and-for-all step change, and allowed the mode! economy to run for 7 years on that step input under the ceteris paribus assumption that firms knew about the duration of the policy when it was enacted. For a short period like 2 or 3 years that is a reasonable assumption to make. However, over a longer time the same assumption is not acceptable. For one thing, if bad side effects are generated, government will certainly enact measures to counter them. This makes the ceteris paribus assumption strange for longer periods.

First we will study the response patterns of ceteris paribus changes in two fiscal policy parameters, one at a time. They are the payroll tax rate (TXW), and value-added tax rate (TXVA). This sensitivity analysis serves two purposes. It tells the power of the two parameters in affecting the macro economy, and the effects can be compared with similar studies on a sector model of the Leontief-Keynesian type also developed at the IUI by Dahlberg and Jakobsson (1977).

Second, it serves the pedagogical purpose of telling what happens in the model, step by step when we enact the parameter changes. The mechanics will certainly be very different from what you can learn from Keynesian macro models or static general equilibrium models that lack explicit markets. This time you can follow the time transmission of the effects through the three integrated product, labor and financial markets.

6.2.1 Value-Added Tax (TXVA)

We first raised and lowered the value-added tax rate in 1969 with ± 2.5 and 5 percentage points, respectively. Since the value-added tax rate was 10% on purchases of goods in 1969, the larger parameter change was a substantia] one.

The effects of the increase in the value-added tax rate are quite straightforward. The experiment can be regarded as a case of restrictive fiscal policymaking. There is no corresponding feedback through increased public spending. There is a negative and not very large GNP effect that stabilizes at around 2% from the third year for parameter changes between +2.5 and 5 percentage points (Figure 6 and also Table 4). Industrial output is affected similarly. The initial first year (2nd and 3rd quarter) effects in Figure 6 depend on the habit-formation hypothesis embedded in the consumption system. The lowered purchasing power of households due to the increase in the value-added tax is buffered through less saving in the first round. It is interesting to notice that the two parameter changes result in output effects of similar magnitude (see Figure 6). In both cases, the employment level is back to the initial level in year 7 (1975). The value-added tax rate increase (Table 4), furthermore, allows us to push wage change (and temporarily employment) down, gradually realizing an improvement in the return to investment.

Figure 6

Download asset Open asset

Table 4

	GNP		Industrial Output (Q)		Industrial Employment (L)			CPI		Producer Prices (PDOM)		Wage Costs (W)		Profit Margins (M)
	1969- 1970	1969- 1975	1969- 1970	1969- 1975	1969- 1970	1969- 1975	Year 7	1969- 1970	1969- 1975	1969- 1970	1969- 1975	1969- 1970	1969- 1975	1969- 1970	1969- 1975
VATAX
+5a	98.5	97.5	98.4	97.6	98.1	97.8	100.1	104.6	104.4	105.3	106.0	100.1	94.4	100.6	109.3
-5	101.3	98.6	100.6	96.6	93.3	92.4	100.0	96.6	95.5	98.2	95.8	100.1	88.9	116.4	125.5
WTAX
+5a	99.2	98.2	99.2	98.4	99.6	98.4	100.1	99.6	99.5	99.7	99.1	100.0	95.4	98.3	106.1
-5	101.0	101.2	101.0	100.9	100.3	100.7	100.0	100.3	100.3	100.5	100.8	100.0	99.6	102.1	102.0
W = Labor cost to firm (wages + all social charges incl. supplementary pension charge)								RSAVH = Household savings ration out of disposable income
PDOM = Domestic wholesale prices, industrial goods, incl. of VATAX								INVFIX = Investment in industry in constant prices
M = Operating gross profit margin								VATAX = Value-added tax (money terms)
CPI = Consumer price index, incl. of VATAX								ITAX = Income tax, income earners (money terms)
PROD = Average labor productivity								WTAX = Payroll tax including all social charges and ATP-charge (money terms)
Q = Industrial output								TXVA
GNP = GNP								TXI = The corresponding nominal rates
L = Labor input in industry; effective man-hours								TXW
If nothing else is said, all tables and figures are on index form where the endogenous variable reported (Q, PROD, etc.) is indexed with the reference run as the base. On a current basis: If Q(1972) = 3.363 in the reference run and Q(1972) = 3.681 in the experimental run, the index will say index = 3681/3363 x x100 = 109.5 (Diagrams). On an average basis: The indexes have been averaged from the beginning year 1969 to the year you read in the table.

1. W = Labor cost to firm (wages + all social charges incl. supplementary pension charge); PDOM = Domestic wholesale prices, industrial goods, incl. of VATAX; M = Operating gross profit margin; CPI = Consumer price index, incl. of VATAX; PROD = Average labor productivity; Q = Industrial output; GNP = GNP; L = Labor input in industry; effective man-hours; RSAVH = Household savings ration out of disposable income; INVFIX = Investment in industry in constant prices; VATAX = Value-added tax (money terms); ITAX = Income tax, income earners (money terms); WTAX = Payroll tax including all social charges and ATP-charge (money terms); TXVA, TXI, TXW = The corresponding nominal rates;

2. If nothing else is said, all tables and figures are on index form where the endogenous variable reported (Q, PROD, etc.) is indexed with the reference run as the base. On a current basis: If Q(1972) = 3.363 in the reference run and Q(1972) = 3.681 in the experimental run, the index will say index = 3681/3363 x x100 = 109.5 (Diagrams). On an average basis: The indexes have been averaged from the beginning year 1969 to the year you read in the table.

The surprise results come when we lower the value-added tax (expansive fiscal policy making). The first 2-year effect on industrial output and GNP is up. The GNP effect, however, extends for a somewhat longer period. Consumers spend more, and especially on goods from sectors 3 (durables) and 4, owing to a higher real disposable income. However, some of the increased purchasing power of households in the initial phase leaks into increased savings and some into more imports. In the first round (2 years) exports are left fairly untouched (not shown).

Since, however, firms do not lower their domestic prices (before value-added tax) by the full amount of the value-added tax rate change, they gain initially on the profit margin side. This is normal pricing behavior among firms. It is difficult to compensate for the tax by raising prices fast, and firms only reluctantly lower prices by the full amount of the reduction in the tax. This produces an asymmetrical profit response to value-added tax rate changes for the 7-year period of investigation that is supported by knowledge of firm pricing behavior. The higher price level (after tax) can be maintained for some time, and firms gain on selling more at home and relatively less in foreign markets. Export performance is coming down slowly from the third year and even though savings also decrease, the import effect persists. More import price competition hurts business profits, and firms begin to get tough on the wage side to satisfy their profit targets. The net impact is a worsened external balance, less domestic supplies in domestic markets (compared to the reference case) and a slowdown in industrial growth. Employment and real wages suffer in the medium term, but business profit margins are still maintained at a higher level by the end of the 7-year period, so the results correspond to rational responses by firms. We should note here that profit margins are on their way down strongly in 1975, probably to undershoot the reference case after a few years. This again underlines the new realistic and hence important feature of disequilibrium models of this kind. Depending upon initial conditions and the time period studied, the results may be very different. The possibility ty that the incidence and the allocation effects of changes in, for example, payroll tax rates may be very different depending on the cyclical situation is also emphasized in Normann and Södersten (1978). A deflationary spiral sets in from the fourth year that works counter to the initial fiscal measure. Firms come out all right with higher profit margins, but output suffers slightly. This deflationary cycle is just about to reverse itself at the end of the 7-year period. Note also that employment recovers faster than output, and the reason of course is a lowering of real wages. The back side of this situation is a lower productivity growth than in the reference case. Quite in keeping with the dominant property of our model, this dynamic reversal is stronger and faster the larger the initial parameter change. The reason is that the early overoptimism generated by the fiscal stimulant has created overexpansive output and recruitment plans. Since profit targets are based on excessive price and sales expectations, wage drift was allowed to occur without violating ex ante targets. Firms then experienced profit disappointments ex post and consolidated their positions by laying off labor and pushing up profit margins at lower output levels. This is what normally happens in the early phases of a business recession, so in fact the initial boom, created by the fiscal stimulus, has generated its own "extra" slump after a few years, albeit a minor one.

These results may seem surprising to a reader used to think in terms of equilibrium models that lack such market dynamics. The results, however, only tell that growth is not the only way to maintain rates of return for firms, and when markets are not so competitive and when profits come easily, it may be better for firms to slow down investment and growth. The results depend in part on the degree of competition in markets but also in part on the feedback, profit-targeting device (maintain or improve profits-MIP) in the firm model. As a result of improved profit performance, firms upgrade their targets and vice versa. This set of propositions is difficult to test owing to the long time dimensions involved, but one has to remember that these unconventional results are conventionally assumed away in most econometric models. So, there is no need to question the results except on clear empirical evidence. A case in point here is Pratten's (1976) results from comparisons of matched Swedish and British firms. He found that while British firms were far below their Swedish counterparts in productivity and growth rates, they attained higher rates of return. Again note (from Figure 6) the second quarter temporary reversal of the GNP effects. It depends partly on households not adjusting their consumption immediately to their lower or higher real purchasing power; they prefer to take part of the adjustment by way of their savings account.

In the earlier sections we have studied the economy wide effects of differently sized countercyclical fiscal parameter variations. In this section we are interested in the effects of a sizable modification of the entire tax system. The conventional way of conducting such a study is to compare two equilibrium states of the economy, one with the old, and one with the new tax system, using some type of general equilibrium model, and assuming the transition to be engineered in a socially “painless” way by a fully informed policy maker. In reality the transition will have to be intermediated by the imperfect information support of market prices. Since our model runs on a market-price information-response system, where performance of the total economic system depends on the ability of agents (decision makers) to interpret these signals. Compared to artificial general equilibrium analyses we can more realistically simulate such systemic transitions. Because of the market disturbances created by major and fast fiscal changes, they must be enacted in a gradual way. Even so we should expect some economic effects (a “transactions cost”) of a negative nature to arise because of disturbances in the market information system, which demand a long time for firms to learn and correct for. This, no doubt, is a relevant aspect of the problem we are about to study.²⁵

We know that no general policy measure can be enacted in a structurally heterogenous economy leaving the structures unchanged. Conventional macro models abstract from such effects, and in neoclassical equilibrium models structures are assumed to remain unaffected by the price change. Such partial analyses become misleading in a longer term perspective. Our model accounts for market induced endogenous structural change through investment and reallocations of workers among firms, and also captures the economy wide consequences of a population of firms that changes because of entry and exit, a changing size distribution of firms, and shifting relationships between firm rates of return, investments, and growth such that the aggregation assumptions that underlie a macro model will vary with the experimental design. In this study we will, however, have to limit our attention to a few illustrations of the changes in relative sectoral sizes brought about by the above fiscal parameter changes via relative changes in prices and rates of return among individual firms. Again, we confine ourselves to a 7-year time span. To generate observable sectoral effects, we need fairly strong and, as it turns out, sudden parameter changes, and parameter changes have different market selection consequences. The value-added tax change hits households directly, while the payroll tax hits the firms.

We are interested in how the effects of one "general measure" are carried through the economic system by market adjustments among actors, and want to study how the carry-over (or through) process diversifies unevenly between firms and sub industries. A change from a value-added tax to a payroll tax system, for instance, will hurt (or help) consumer goods producers more than other industries.

One thing comes out clearly in Table 2. When we make a sudden, once-and-for-all change from a payroll tax to a value-added tax, it hits household demand. Consequently, growth comes down in the consumption goods industry (4) and increases among export intensive firms. This relative growth reallocation between industries is reinforced by the fact that investment spending also decreases somewhat. Consequently, total industrial output growth decreases as well. On the other hand, the comparable, public budget neutral lowering of the payroll tax should be a positive stimulant on manufacturing industry, albeit with a delay compared to the negative demand effect of the increased value added tax, so the question is if the long run growth effects on the model economy cancel, and only leave a changed industrial structure in its wake?

When the tax system is changed in the other direction, the effects are also reversed. This time the differential impact “looks" somewhat bigger in the table, since the positive saving-investment growth effects combine with the direct consumption effect to favor growth in the consumer goods industry. One should however note that raised demand in domestic households as well as raised investment demand partly occur at the expense of export deliveries. In the beginning imports increase, as should be expected, and the foreign balance deteriorates (not shown). In the longer term both exports and imports balance off better in the downward direction compared to the reference case, but the long-term outcome is a worsened external balance.

Lower payroll taxes and a higher value-added tax as compensation, however, mean a long-term improvement in the external balance, even though both exports and imports decrease for the first few years. Interestingly enough, however, when we allow for a smooth gradual replacement of the entire payroll tax system by a more dominant value-added tax system (Table 2 and Figure 7) we can study reactions of similar direction during the seven-year period. But the internal dynamics of the system this time are efficient enough to make relative prices and activity levels interact in such a way that the initial relative price structure is approximately restored toward the end of the simulation, and so is also industrial structure, as measured by relative growth rates in output. Again, we should remember that if actors in the markets are allowed enough time to reinterpret the new signals, no adverse effects on total growth need to occur. The long­ term effect on the structure of the value-added and payroll tax switch seems to be fairly “neutral."²⁶

Figure 7

Download asset Open asset

When the direction of fiscal change is reversed (the pay roll tax gradually replaces the value-added tax) the effects mentioned above are also reversed (see Figure 7). It is worth noting that the combined effect on household durable goods demand and business investment goods demand (both goods categories being supplied by industry 3 in Figure 7) is slightly positive on the average when going from a payroll to a value-added tax system and vice versa. Furthermore, when the effect is negative, the cyclical swings generated in the durable goods producing industries are much stronger. And, since the effect on total industrial production for the same period (the fiscal change going both ways) is approximately nil, part of the explanation lies in a reshuffling of investment spending patterns between firms and industries. None of these results are a priori obvious but depend on the numerical specifications of the model.

The upper Figure 7 however shows that even tough total manufacturing output has returned to the level in the reference case after a change from payroll to value added tax by the end of the period, during several years in between total output runs significantly below that in the reference case. One may perhaps refer to this as a temporary “transactions cost” for the economy at large incurred because of the market disturbances associated with the systems change. Perhaps also the growth effect enjoyed by the typically export intensive firms in the RAW industry would have persisted if the simulations had been extended for a longer time. The opposite (lower figure) appears to have occurred in the change from value added to payroll tax system, benefitting consumption goods.

We also note that the changes in growth rates are quite small despite the large institutional change, and that initial sectoral balances are roughly restored by the end of the 7-year period. That foreign trade should play a role in the adjustment process, as witnessed by the opposite movements of the export intensive industry 1 (raw materials), and the domestically oriented nondurable, consumer goods industry 4 is perhaps not surprising. Even though it looks as if the two effects roughly cancel over time, and movements are in opposing directions and depending upon the direction of the fiscal change over a longer time than the seven simulated years one should expect a permanent relative product price change for Swedish firms depending on the relative importance of exports, and even though small also a permanent change in production structure over the longer term.

Figure 8 gives a micro illustration from a new experiment clarifying what market prices occasion within the model. We have chosen to raise transfer payments to households and the payroll tax simultaneously in 1969 so that the initial (1969) public budgetary impact is approximately nil. The expected change would be a shift in total consumption demand towards relatively more private consumption. Domestic producers of consumption goods would benefit, and typical export firms would be hurt. How exactly business investment spending is affected cannot be foretold by principal reasoning. Since the fiscal parameter change is sudden, substantial, and of a one-shot type, we expect a negative total growth effect over the seven­ year period owing to market disturbances. This conclusion is based on our experience from interpreting Tables 2 and 4. This would suggest less investment spending, but perhaps more household spending on durable goods. Increasing transfer payments, while increasing the payroll tax, however, are very similar to lowering the value-added and increasing the payroll tax. Tables 1 and 2 would thus suggest slightly more investment and hence more growth in sector 3.

Figure 8

Download asset Open asset

The outcome is not exactly as anticipated. Over the 7-year period both the raw materials firms, and the investment and consumer goods industries lose growth on average, as a result of the fiscal change. For raw materials producers this would be considered normal as they export the bulk of their output. They do not benefit indirectly from the transfer payments to households, but are nevertheless affected by the payroll tax increase. Investment goods producers’ output decreases slightly, even though earlier results would suggest a slight demand benefit. For consumer goods the result is odd at first sight. It is however, due to the fact that firms respond fast to the payroll tax increase by reducing cash wage payments after pay roll tax, so the net effect on disposable income (after tax and transfer payments), everything considered, is a reduction.

Figure 8 illustrates what has happened at the micro level for investment goods producers. The upper scatter and dashed line in Figure 8 give the relationship between average rates of return and output growth rates for the 5-year period 1969-1973. As should be expected the relationship is significantly positive. When the same relationship is extended to 1975, the regression line (shown by the solid line in the upper and the dashed line in the lower part of the figure) pivots for two reasons. The high-performance (in terms of profitability) firms essentially stay with their relatively higher rates of return for the longer period. Since they reached that position in the inflationary period 1972-1973 and have plowed back profits in investments, and/or have borrowed more to invest because of the higher rate of return, the higher growth in capacity that enters production with a delay explains part of the pivoting. However, some of the firms in the low-performance end have instead suffered, owing to the wage drift induced by the generally higher inflation that they have to share. This reinforces the pivoting in the left-hand end of the diagram. The discriminating impact on firms in different performance brackets of sudden inflation bursts (1973 and 1974) followed by a sudden lowering of price increases (1975) is illustrated even better in Figure 9. The reason is that wage drift is governed by a margin of high-performance firms but spreads to all firms through competition for workers in the labor market. High­ performance firms bid up wages to get more workers to improve their profit position, so arrows above 10% rates of return in Figure 9 tend to go up and right. Low­ performance firms have to take these wage increases in order not to lose workers, and their rate-of-return position deteriorates. Arrows point upward and left in Figure 9.

Figure 9

Download asset Open asset

When the fiscal change defined above is enacted, the effect is to shift down the whole cluster of (RR, DQ) points in Figure 8 (lower diagram) relative to what it would have been without the change. The same average rate of return for the 1969-1975 period is now associated with a lower average growth rate. The shift is in both dimensions, however, and the fiscal measure enacted both lowers the average growth rate for the industry and increases the rate of return. The reason is again indirectly clear from Figures 8 and 9. The two fiscal measures combined shift the composition of demand suddenly. Some firms are hurt and slow down their investment spending. Others benefit, but there is not time enough to catch up fully during the period, and total growth is lowered.

This project would have been impossible without the generous back up of technical skills from IBM Sweden, a partner in this research venture.

Several persons have been actively involved in this project. In particular I want to mention Gösta Olavi, IBM Sweden, who has provided indispensable and efficient support on the mathematical and programming side, as well as Louise Ahlström and Thomas Lindberg at IUI, who have been responsible for putting together and adjusting the data base to the statistical format of the model.

A complete mathematical specification of an earlier version of the model without taxes and transfers is found in Eliasson (1976b).

This means that the next stage of model development will be predominantly devoted to enlarging and refining the micro database. See below.

See e.g. Albrecht (1978a) who entered U.S. macro data into the model and compared the development of a “Sweden like” economy with that of a “US -oid” economy under the same exogenous circumstances.

For that the reader is referred to Eliasson (1976b).

We plan to enter explicit market imperfections at some later time.

We have also tried to estimate some of the individual firm coefficients above by direct interviewing of execu tive staff in one very large Swedish firm. The results turned out quite successful in terms of improving historic tracking of data for the same firm. Further efforts of this kind are planned. To this can be added the possibility to impose an exogenous adjustment of expectations in individual firms. The profit-targeting function used in the model is similar in form to HIST(r) above. For further details on specification see Eliasson (1976b).

See further Eliasson (1976a), especially chapter XI where observed budgeting practices are reformulated as a theory of firm behavior..

See the discussion in Eliasson (1976a), Eliasson (1977).

To see how QFR(L) is determined see the paragraph on the data base below.

Capacity to produce is expressed in terms of QFR(L) in Figure 3. Roughly speaking, new technology enters through an endogenous upgrading of γ (see formula in Figure 3) by merging the old and the new (additional) QFR(L) through an harmonic average. We prefer to work with a production system with capacity expressed as potential output and no explicit aggregate measure of capital stock. The reason is that the estimation procedure is geared to direct use of individual firm data on the format of each firm’s own internal accounting and planning system. There one seldom or never meets with the concept of aggregate capital stock in the context of production planning See Eliasson (1976a). The mere fact that QFR(L)- as estimated on data from the Planning Survey - shifts by investment of course means that one can define an implicit measure of capital that has accomplished that shift. If we do so, a generalized CES-type production function can be shown to appear (See Eliasson (1978). The QFR(L) specification combined with its updating by new investment vintages is however perfectly OK for our purposes, since we are neither concerned with modelling observable firm production planning, nor being compatible with neoclassical micro theory.

It is based directly on a generalized version of the Meyer and Kuh (1957) residual funds-accelerator theory as reformulated to include the consequentes of Stockholm School feed backs of anticipatory mistakes and estimated in Eliasson (1969).

This part has been designed, but not coded. See Eliasson (1976b).

For this the reader is referred to Eliasson (1976b).

The two step procedure to calibrate these parameters is desceribed in Eliasson and Olavi (1978b). All the X3 parameters are listed and defined there. The procedure begins with getting the simulated trends right. Then a different set of parameters is varied to get the business cycle around the trend right.

e can be written e=m/t, where m =marginal tax rate and t = average tax rate.

An inconsistent experimental design that we enacted by mistake illustrates some macro properties of the model. The scales of the Swedish income tax system are now indexed. During inflationary times like the 1970s absurd things occur if the scales are not frequently adjusted. This has been done several times since 1970.Inflation is endogenous. These corrections in the model must be entered through e in the above TXI function. We forgot to enter such corrections after 1976 in a 20-year experiment that started in 1968. A gradual slowdown in total economic growth occurred; not extreme, but unexpected, and it led us to look for the reason. We found a rapidly decreasing current, after tax income inflow into the household sector, and a tremendous public budget surplus piling up and being deposited in the money system.(Public employment and expenditures were fixed in volume terms).Household private consumption was still being maintained at high levels but at a diminishing growth rate. Why? Households were simply borrowing back the public surplus at the going interest rate to maintain their living standards. In the current version of the model the only hindrance to borrowing is the upward drift in the interest rate that occurs if total demand for funds grows faster than total supply of funds. Interest costs are fully tax-deductible in the model as well as in Sweden. The slight slowdown in growth in private consumption was dictated by the permanent income or habit formation part of the household expenditure system that determined how fast households reduce their addicted consumption levels, that are in turn determined by past consumption levels. This reduction was strongly dampened by the efficient inflation and income tax hedge that borrowing constitutes in the Swedish tax world, a hedge that is also effectively enjoyed by income earners possessing highly valued and mortgageable property. This property of the entire model economy that we did not foresee, and discovered accidentally, however, is quite in line with the current political discussion of the problems afflicting the Swedish tax system.

See Eliasson (1976b) and Albrecht (1978b).

See Ahlström (1978).

My experience is that firms do not have the kind of internal acconting system that makes direct questioning on their I/O coefficients meaningful.

See Eliasson and Olavi (1978).

These systems properties have been extensively investigated on an earlier version of the model in Eliasson (1978).

Together with Dr Hans Genberg, Institute Universitaire des Hautes Etudes Internationales in Geneva. Some results will soon be reported. Also, cf. some earlier research on the rate of foreign-domestic price transmission in Genberg (1974).

In analyzing these experiments I have been fortunate in being able to draw directly on a recent IUI study (Normann and Södersten, 1978) that includes a broad analysis of the Swedish tax system.

Individual decision makers (/firms) in the model are equipped with a feedback learning mechanisms for all their (mostly price) expectational signals that correct the interpretation of “new” signals asymptotically but quite fast. (See above description of micro part of model). Erratic (“random”) signals contain no information and cannot be interpreted, and by definition, the more of such noisy price signaling the more disorderly markets. However, if the change to a new tax system is gradual and smooth the more interpretable price signals. We can then isolate the effects due to price disturbances during the transition, from the structural effects of the change in tax system if we run the experiment long enough. We have not done that due to the time constraint, and because we then have to leave the domain of known historical experience (1969-1975) and interpret our results in a new, hypothetical economic environment. That would be a major, but interesting research effort that we will have to leave for another occasion.

Note that the exogenous foreign relative price change is the same in all experiments. Government and households do not change their demand patterns except through endogenous feedbacks because the tax changes have affected wage incomes in firms and the distribution of demand on different markets.

Published in Haveman and Hollenbeck (1980).

1. Version of Record published: December 2, 2024 (version 1)

© 2024, Eliasson

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.