Kari E.O. Alho, ETLA, The Research Institute of the Finnish Economy
Abstract
Income policies are a key element of labour market negotiations and economic policy in Finland. Despite this, there has been a lack of macroeconomic analysis of it, which is the focus of this paper. First we will discuss at a theoretical level the possibilities to achieve a centralised income policy agreement and derive the characteristics of such an agreement compared to union-level agreements. Thereafter, we will briefly assess the connections between collective bargaining and firms’ wage formation needs. At the end, we will present a quantitative macroeconomic assessment of the significance of centralised income policy agreements by estimating the dampening effect of them on rise of real wage costs and by feeding this then into a model depicting the behaviour of the labour market and determination of the equilibrium rate of unemployment. According to the results, centralised income policy agreements boost employment and lower the unemployment rate, on average, in the medium term by about one percentage point, compared to union-level agreements.
Key words. Income policies, wage formation
Introduction
Income policies have been practiced in Finland in various forms in the post-war decades and since the late 1960s this term has been in common use. It is nonetheless surprising that there is a lack of quantitative research on their impacts, even though it is a key component of labour market negotiations and economic policy in Finland. The general attitude towards income policy is quite favourable. It is thought that a coordinated settlement on wages is more efficient than uncoordinated wage competition between unions.
From the perspective of trade unions centralised income policy agreements have to be judged by three main goals:
- favourable development of real earnings
- favourable development of employment
- controlling inflation,
with respect to which centralised income policy agreements (CIPA) may be deemed better than the alternative, i.e. union-level agreements. These three goals are nevertheless mutually conflicting since centralised income policy agreements may promote employment by dampening the rise in real labour costs, so that achieving the first and second criteria simultaneously requires facing a trade-off between these two goals. A centralised income policy agreement also appears to have virtue of its own from the standpoint of the prestige of the trade union movement.
Moderate wage settlements are in the interest of fiscal policy decision makers, the government, in terms of controlling inflation and boosting employment, so that it is inclined to give tax concessions to wage earners in order to foster centralised income policy agreements. This may partially compensate for the lower income rises attached to a moderate wage settlement. Thus fiscal policy can be looser when a centralised income policy agreement is in place than otherwise would be the case, so that a fourth goal can be added to the list:
- favourable development of disposable income.
From the perspective of employers the goals for wage negotiations are partly the same as those above supplemented by the goal that it is necessary in them to ensure:
- the competitiveness and profitability of enterprises.
The employers do not have a clear opinion of the virtues of centralised income policy agreements as such, but rather their attitude depends on the cost level achieved in individual agreements. The interest of the employers has perhaps shifted – as EMU membership imposes an effective constraint on wage hikes, see figure 5 – more toward questions regarding efficiency of wage formation on a micro-economic level such as the impact of wage formation on incentives in firms, and there is a growing tendency to try to steer agreements on wage increases more toward the firm level. From the perspective of employers these trends have, however, not led to significant changes in the negotiation set-up, but rather centralised income policy agreements are still one alternative.
Income policy is in a continuous state of flux as one of the key elements of the labour market. It is continuously “competing” with other types of agreements, and its main alternative in Finland is the union-level agreement. Even in this case there can, of course, be co-ordination between unions or employers.
Centralised income policy agreements have thus included other economic policy elements such as tax concessions. We do not seek here to evaluate assorted mixes of policy elements and their joint impact, but rather we focus only on the wage settlements and on their impact and on the differences between centralised and union-level decentralised labour market agreements. Connections between taxes and wage formation are presented elsewhere as a separate part of this research project.
The paper proceeds as follows. At first, we discuss wage agreements on a theoretical level and whether union-level or centralised income policy agreements are “better” and when we are prone to end up having one or the other. Readers who are more interested in the significance of income policies from a quantitative macroeconomic standpoint can skip directly to the final section. Next, we briefly evaluate the mutual compatibility of collective bargaining and firm-level wage formation needs. Thereafter, we empirically investigate the dampening impact of centralised income policy agreements on the rise in real labour costs in Finland. This estimate is fed into a recently built model (Alho 2002a) depicting the behaviour of the labour markets and determination of the so-called equilibrium rate of unemployment. This exercise seeks to shed light on the macroeconomic impact of CIPAs on employment.
Comparison of centralised income policy agreements and union-level agreements
Wage formation is ordinarily studied in labour market theory so that wages are agreed upon in negotiations between employers and unions representing employees. In the following we will look at a more complicated situation where the employers’ federation negotiates wages with two trade unions, which can be assumed to represent skilled and unskilled workers. This breakdown is not significant in the following; what is important is that there are two trade unions, the workers of which work together in firms and that their unions negotiate wages with the employers’ federation representing these enterprises.
Static wage negotiations
Let us start at first with wage negotiations between one trade union and an employers’ federation representing firms. The employers’ goal is to maximize the firms’ profits π,
π = PQ - WL - rK, (1a)
where P is the price of production, Q is its volume, W the wage level, L labour input (employment), r interest rate and K is the capital stock. Capital is a fixed factor of production in the short run. The trade union, on the other hand, seeks to maximize the total income U of its members,
U = WL + b(N-L), (1b)
where b is unemployment compensation and N the number of union members, which is split between employed persons (L) and unemployed persons (N-L). The result of the firms π0 when nothing is produced (so-called fall-back situation), is the loss stemming from interest expenses on capital, i.e. - rK. The trade union’s fall-back situation U0 is one where all members receive unemployment compensation, which is bN. The negotiations between the employers and trade unions are ordinarily described using the so-called Nash bargaining model, where the counter-parties maximize their utility relative to that achieved in the fall-back situation,
(π - π0)θ(U - U0)1-θ = (PQ - WL)θ(L(W - b))1-θ, (2)
where θ is the firms’ and 1-θ is the employees’ bargaining power. Wage negotiations are assumed to take place under the constraint that the firms decide unilaterally about the level of employment and production after the wage agreement is finalised. This arrangement is known by both counter-parties, when the wage negotiations are initiated.
From these points of departure we can in a few steps derive from equation (2) the optimal condition for wages W,
(Illustrasjon mangler)
(3)
Here εL is the firms’ elasticity of demand for labour L with respect to real wages W/P (and is negative with an absolute value assumed to be smaller than one[25]) and Ω is the ratio of bargaining powers (= θ/(1-θ)). If there were no external unemployment compensation (b = 0), the wage level would be a certain share of productivity PQ/L. The more elastic the demand for labour is with respect to real wages (the greater the absolute value of elasticity εL), the smaller this share. The existence of unemployment insurance, on the other hand, changes the situation so that the ratio of wages to productivity rises above this threshold.
Let us then extend the analysis so that the firms use two types of labour, L1 ja L2,
Q = F(K,L1,L2), Fi > 0, Fii < 0, Fij > 0, (4)
where F is a production function with normal assumptions.
Both employee groups are represented by their own trade union, which negotiates with the employees’ federation. A further difference with that above is that both trade unions seek to maximize, not only their own rent in (1b), but also their wages relative to the other trade union. This is a fairly natural extension of the trade unions’ objectives stemming from the notion of wage bargaining in practice. In this respect our framework differs from that described in the seminal article by Calmfors and Driffill (1988). Thus the objective function of trade union i, i =1,2, could now be
(Illustrasjon mangler), (5)
where μ depicts the weight given in this negotiating framework to the relative wage (μ > 0). The negotiated settlement is an extension of equation (3) and rather complicated in form, so it is fruitful to illustrate it graphically, as we do below. The wage negotiations are carried out with the employer either non-cooperatively, which we will call the union-level solution, or some sort of a jointly agreed solution is reached, which we call a centralised income policy agreement. We will describe the latter in more detail a little later.
In the case of a so-called monopoly trade union, the union has ultimate power to decide about wage increases, but it is constrained by the fact that after the agreement firms decide about the level of employment. Thus in equation (2) the power parameter θ = 0, i.e. the negotiation result is determined only by the goals of the trade union. On the other hand, now the other workers’ union also decides about its wage at the same time.
The utility of union 1, U1, as defined in (5), depends on the wage in their own union as well as in the neighbouring union, i.e. U1 = U(W1,W2), where UW2 < 0, because a rise in the neighbouring union’s wages reduces the employment in union 1 and decreases its relative wage. The same applies to union 2.
The reaction curve R1 of union 1 with respect to union 2 is defined as the wage W1 of union 1 that generates the greatest utility for union 1 when the wage for union 2 is given, i.e.
W1 = R1(W2), where typically R1’ < 0. (6)
The reaction to the wage of union 2 is deemed here to be negative because when the wage of union 2 rises, union 1 witnesses a decline in the income to be shared with the firm and also the demand for its type of labour slackens. On the other hand, the pressure to raise their own wages increases, because the wage ratio in equation (5) becomes worse for union 1. Below we will assume that this wage competition will not weigh so much in the goals of the trade union as the employment of their own members. But it is possible that in reality the interaction between wages is more complicated than depicted here. Let us assume further that the reaction in equation (6) is smaller than unity in absolute terms, because then the trade union can compensate for the fall in its wages by ensuring that its relative wage does not decrease to the full extent. In this case the union-level agreement is from a technical standpoint unique and stable. The situation is depicted in figure 1, where the reaction curves are simply assumed to be straight lines.
An indifference curve of union 1 is defined to be U1(W1,W2) = c, where c is a constant. From this we can solve for the dependence of W2 with respect to W1,
(Illustrasjon mangler) (7)
At the maximum the numerator for the reaction curve R1 of union 1 is, according to the definition, zero, and because the denominator is negative, we can see that the indifference curves for union 1 and union 2 are of the form depicted in figure 1.
The intersection R of the reaction curves in figure 1 is the wage negotiations’ so-called Nash solution, which is the game’s non-cooperative solution. It can be called the union-level agreement situation. Let us assume that the counter-parties can also engage in cooperative centralised income policy agreements (CIPA). In principle they are agreements where the trade unions share the rent, i.e. the sum of the maximum solutions to the objective functions (5) is divided amongst themselves through the so-called side payments. In practice, these kinds of payments and agreements are not usually possible, thus we assume that centralised income policy agreements are constrained to those that do not change the relative wages between the unions compared to the non-cooperative situation. The income policy solutions are then depicted in figure 1 as the line OR. In this case a cooperative centralised income policy agreement is always better for both unions than a union-level agreement, which is easy to see since the indifference curves at point R are vertical and horizontal, respectively, so that moving along the line OR generates a better situation for both unions.[26]
Simultaneously, the centralised income policy agreement is more advantageous for the employers than the union-level agreement. This becomes evident when we take the derivative of the enterprises’ profits π = PQ - W1L1 - W2L2 - rK with respect to wages Wi and substitute into this the above-mentioned condition that the enterprises decide about the demand for labour, which means that the so-called marginal revenue product and wage level are the same in the equilibrium. Thus we obtain
dπ = -(dW1)L1 - (dW2)L2. (8)
This is positive since both wage levels are smaller in centralised income policy agreements than in union-level agreements. We thus obtain the following result:
Outcome 1. Two monopoly trade unions should jointly seek a moderate centralised income policy agreement. From the standpoint of the firms centralised agreements are better than union-level agreements. From the standpoint of employment centralised agreements are better than union-level agreements.
This analysis applies to the case of two monopoly unions where the employers are passive in the negotiations. On the other hand, under wage negotiations described above where the employers have power, the trade union has to be satisfied with a lower wage than would be the case with a monopoly union.
Now the situation is like that in figure 2. The dotted reaction curves describe the negotiating situation and its change relative to the case of monopoly unions in figure 1.
Now we see that the new outcome at point R’ is located closer to the origin than the previous union-level outcome at point R. Point R’ can in fact be closer to the above-described coordinated outcome than the union-level outcome R. Surprisingly enough, the result at point R’ can be better also from the standpoint of the trade unions than point R.[27] We obtain the following result:
Outcome 2. Employers’ power in wage negotiations reduces benefits from wage coordination by the unions, and can make a centralised income policy agreement unnecessary.
Agreement on wage increases
The analysis and comparison of the advantages and disadvantages of various types of wage agreements is nevertheless insufficient because in practice both centralised income policy agreements as well as union-level agreements are made regarding changes in wages over time, from one year to the next, not with respect to a hypothetical non-cooperative situation. Let us further assume that the centralised income policy agreement can only mean that the wages of all unions are increased by the same amount. Let us evaluate the basic two monopoly union situation. The case of wage bargaining is analogous. The situation may be as follows, see figure 3.
The dotted lines in figure 3 depict the transition from the situation of figure 1 in period (year) 1 to period (year) 2 via growth in productivities PQ/Li. In this figure it is assumed that the non-cooperative union-level solution would mean that the relative wage W1 of (skilled) labour L1 would rise in year 2 compared to year 1 relative to the wage W2 of (unskilled) labour L2. Despite this, the emergence of a new centralised income policy agreement between the counter-parties is worthwhile, because it is possible to find a solution at the extension of the line OR1 that would be better for both counter-parties than the new union-level solution. This is designated in figure 3 as “New CIPA”. In this solution the wages rise by the same percentage for both unions and relative wages remain unchanged.
On the other hand, if there are profound structural changes in the labour markets, it is not necessarily rational to bring about a new centralised income policy agreement, but rather the counter-parties are driven toward a union-level agreement. This situation is depicted in figure 4.
Now the extension of the centralised income policy agreement is no longer located in the area for centralised agreements that are better than the union-level solution for both counter-parties.
Thus we obtain the following result.
Outcome 3. Two trade unions should make a centralised income policy agreement with the employer when productivity changes approximately the same over time with respect to different labour components. On the other hand, in the opposite situation we naturally end up with a union-level agreement.
If the changes in relative wages affect employees to move from one enterprise to the other and their incentives to give effort, the situation becomes more complicated.
It may be advantageous or disadvantageous for firms to continue with centralised income policy agreements because the wage level of one of the labour components is higher or lower than in the union-level agreement. On the other hand, if centralised income policy agreement is very stringent so that the wages of both components increase only a little (i.e. at most to point T in figure 4), a centralised income policy agreement would be more advantageous to firms than a union-level agreement. Based on this we can decide that
Outcome 4. The continuation of centralised income policy agreements would always be better for one of the unions, the “weaker” one. For enterprises the continuation of centralised income policy agreements may be advantageous or disadvantageous depending on how stringent it is.
The main conclusions from this analysis are that the coordination produced by income policies can have positive effects on the economy. The impacts of centralised income policy agreements must nevertheless be evaluated over time, to which direction we proceeded in the analysis above.
Impact of centralised income policy agreements on wage formation in firms
Labour market agreements provide a framework for minimum wage scales and increases in contract wages. In practice, most employees work for wages above the minimum wage. Nevertheless the stipulations for minimum wages play a significant role in the labour market.
Let us now briefly address the relationship between wage formation in centralised income policy agreements and firm-level agreements. Let us assume that a firm has N workers to which it would like to pay wages of Wi at a certain time. The employees are ordered according to rising productivity and wage levels. The desired wage is determined on the basis of productivity and the firm’s view of incentives stemming from wage differentials. Centralised income policy agreements determine the level of low wages based on the changes in minimum wage scales. This has, however, an impact on the entire wage distribution, as discussed in Alho (2002b). In a labour market equilibrium the relative wage for more efficient workers (Wi/WMIN) falls when the lower wages are raised. The intuitive explanation for this is that the rise in the minimum wage reduces the demand for low productivity labour, which decreases the productivity also of more efficient workers and thus their wages. This is because factors of production are typically cooperative between themselves, i.e. increasing the amount of one factor of production increases the productivity of other factors of production. This situation is discussed further in Alho (2002b).
This narrowing of the wage dispersion will not necessarily take place in practice since in order to keep sufficient incentives stemming from wage formation, a firm has to keep the wage distribution unchanged, i.e. an increase in the minimum wages will prompt pay hikes across the entire distribution of wages.
On the one hand, firms have a certain tolerance against a rise in total costs, enabling them to manage in the pressures of market competition. The trade unions, on the other hand, seek to negotiate as high a general pay hike as possible in line with this wage norm. The trade unions also seek to have minimum contract wages increased more than the general pay hike, because wages have in practice drifted above the previous level of the contract wages. Thereby, there will be a further push to the firms to drift wages even more upward. We obtain the following result:
Outcome 5. Firms are often in a pressure, caused by collective wage bargaining, to raise wages more than the market equilibrium would require and what their competitiveness will bear if they want to obtain their optimal internal wage distribution.
Uusitalo (2002) presents evidence that the relative distribution of wages in Finland has become more dispersed when there is a union-level agreement while the wage dispersion has decreased under a centralised income policy agreement.
A centralised income policy agreement can thus be detrimental from the standpoint of a firm, but this holds also for union-level agreements. A practical solution for these problems includes new pay schemes, namely performance-related pay schemes, such as profit sharing, and some leeway in deciding about wages at the firm level stipulated in connection with comprehensive settlements. Wage formation under centralised income policy agreements and enterprise-level agreements as well as their relationship to economic growth and the allocation of resources will be addressed later in this project, see also Alho (2002b).
Empirical estimate of macroeconomic effects of income policies
As mentioned in the introduction, it is surprising how little the significance of income policy has been analysed by labour market researchers in Finland, even though it has played such a key role in the labour market negotiations and economic policy. In the following, we will seek to present such an analysis.[28] The aim is first to estimate how centralised income policy agreements dampen the rise in real labour costs and then to assess its macroeconomic significance.
In studies comparing international practices in different countries, it has been found that coordination practiced in wage negotiations between both employers and employees reduces unemployment significantly, see e.g. Nickell (1998).
The rise in wages has recently been moderate in Finland. The rise in contract wages has stabilised after the recession at 3 per cent per annum, as indicated by figure 5. The rise in total earnings including wage drift has been less than 5 per cent per annum.
(Illustrasjon mangler)
The point of departure in the following is to utilise an indicator of the collective agreements that has been compiled in ETLA by Ruutu (1997) and more recently by Marjanen (2002). Their main purpose is to depict the nature of collective agreements, i.e. their degree of centralisation and coverage. These indicators are presented in figure 6.
(Illustrasjon mangler)
Ruutu’s (1997) index is called CIPA. It measures how centralised the labour market agreements have been, i.e. how many of the employees have participated in the agreement. Marjanen’s (2002) indexes are called CIPACOV and CIPABIND. The former depicts the national coverage of the agreement while the latter depicts how well the participants have adhered to the centralised agreement. Ruutu’s index is in principle continuous and scaled between 0 (completely union-level) and 1 (centralised income policy agreement). Marjanen’s indexes are subjective estimates of the degree of adherence and they obtain a discrete value from zero (poor) to three (very good).[29] Figure 6 nevertheless shows that the indexes are fairly similar in shape and in the practical estimations, explained below, they work in much the same manner, so it is not necessary to report the results separately in this respect.
First the income policy index was entered as an input into the wage formation equation of Alho’s (2002a) labour market model (see below). This proved to be an insignificant additional factor in this connection. This may be because even without this information the model is able to depict the general kind of effects of an agreement calling for a modest pay hike of a certain size even though it does not take into consideration the precise nature of the agreement. There is no explanatory power left over for the information describing the type of agreement.
Let us then form an aggregate model depicting the significance of the wage agreement with the following three equations:
wcontr = p + prod + C(1)CIPA (9a)
w = wcontr + C(2)(wcontr-p-prod)-1 (9b)
p = C(3)(w+payroll-prod) + (1-C(3))pm , (9c)
where wcontr is the rise in the contract wage index (%), p is the rise in the GDP deflator (%), prod is the rise in productivity of labour (in the total economy, i.e. GDP/hours worked, %) and CIPA is the indicator of the nature of the labour market agreement explained above. In the second equation w is the rise in the wage level (earnings index) in the total economy (%) and the subscript –1 denotes a lag of one year. In the third equation payroll is the change in employers’ social security payments as a percentage of total wage bill, %-points), and pm is the change in import prices (%). The terms C(1), C(2) and C(3) are unknown parameters that are estimated from the data.
Equation (9a) depicts the formation of contract wages. Their rise is assumed to depend on the norm of the economy’s wage increase, i.e. sum of productivity and inflation, as well as depend on the nature of the agreement, i.e. whether or not it is a centralised income policy agreement or a union-level agreement. In the following we do not make a distinction between different centralised income policy agreements, which have not been uniform, see figures 5 and 6, but try to reach a view of their average impact. The point of departure is that a union-level settlement is pegged to the norm of the economy’s wage increases (so that the CIPA index gets a value of zero) and the centralised income policy agreement means a more modest increase (coefficient C(1) is negative). The second equation depicts how the earnings rise directly in line with the increase in contract wages, but there is also an adjustment factor whereby, if the wage increases are less than wage norm, then in practice wage drift ensures that at least part of the moderateness of the rises in contract wages is eliminated by labour market forces. Thus we can assume that also the coefficient C(2) is negative. The third equation depicts the ordinary formation of inflation via cost pressures. When the equations are solved with respect to the endogenous variables w-p and p, we obtain the following solution over the long run.
w - p = prod + C(1)(1+C(2))CIPA (10a)
p = (Illustrasjon mangler)(10b)
This gives us the long-term impact of centralised income policy agreements on real wages if the coefficient C(1) differs from zero and the coefficient C(2) differs from minus unity. The model was estimated by two-stage least squares (TSLS) using as instruments productivity, the change in import prices, lagged inflation and lagged increase in wages and the CIPA indicator. The estimation results for 1976-99 are as follows:
|
Parameter
|
Estimate
|
t-value
|
|
C(1)
|
-3.97
|
-4.86
|
|
C(2)
|
-0.50
|
-7.39
|
|
C(3)
|
0.76
|
5.61
|
According to the results, a centralised income policy agreement dampens the rise in contract wages very strongly. The impact implied by the coefficient C(1) would be to decrease the rise in wages by 4 percentage points on average. Half of this wage moderation is nevertheless eliminated by market powers as indicated by the coefficient C(2).
Thus the effect is that the centralised income policy agreement dampens the rise in real wage costs by about two percentage points over the long run.[30]
There is one structural feature in the model that should be considered in more detail. In Finland, as elsewhere, the share of labour income out of national income, has fallen sharply in the 1990s. Various explanations for this have been offered. For example, Ripatti and Vilmunen (2001) suggest that this phenomenon stems from changes in technological development so that the efficiency of capital has risen and so-called price-cost margins have increased. In the above model, the shrinking of the share of labour income in national income is attributable to centralised income policy agreements, but only to a small portion.[31] Thus, the model does not capture the change in income distribution in a satisfactory manner. If a time trend and its square is included to describe the changed price-cost margin, the CIPA indicator becomes insignificant. If the model is again modified so that the rise in contract wages is a part of the extra income available for paying wages (i.e. p + prod), then the coefficient for the CIPA indicator now becomes insignificant again. This form is not very sensible since the dampening of the rise in wage costs is then taken into consideration already in this average coefficient.
The model was also estimated recursively so that equations (9a) and (9b) were combined as one wage equation, which was then estimated over rolling 15-year time spans. The key parameter C(1)(1-C(2)) was estimated to bring the following result, see figure 7.
We again observe that centralised income policy agreements have had a dampening effect on the rise in real wage costs, which differs to some extent statistically significantly from zero. We can also see that the impact has been decreasing over time. It is nevertheless still large enough that a centralised income policy agreement dampens, over the long run, the rise in real wage costs by about 1.5 percentage points compared to the case where the agreement would be a completely union-level one. This discrepancy stems from the fact that the estimate of the coefficient is multiplied by the difference of the CIPA index (centralised = 1, union level = 0).
(Illustrasjon mangler)
The impact is somewhat smaller in this estimation than in table 1, albeit for a different estimation period. When the basic model (9a)-(9c) is estimated for the short period of 1984-99, the parameter C(1) now obtains the value –3.4 and C(2) was still near half (-0.48), so in this basic framework the impact of centralised income policy agreements would have decreased in recent years compared to the entire period. As the estimate of parameter C(3) in the model is as high as 0.75, the inflation dampening effect of the centralised income policy agreement would have been unrealistically high. If we use the more ordinary value of 0.6 for the parameter C(3) in cost pressure calculations, we find that the inflation dampening effect of centralised income policy agreements would be 3 percentage points, which is also very much.
As a final stage, we will seek to estimate the macroeconomic impact of centralised income policy agreements using a labour market model constructed in the study Alho (2002a). It consists of two equations: 1) labour demand or price setting and 2) wage setting. The labour demand equation is written to depict the change in the unemployment rate. It is explained by real wage costs relative to productivity and GDP growth. Real wage formation is affected by the unemployment rate, labour income’s share of GDP as an error correction mechanism as well as GDP growth and the tax wedge. In the long run, the model depicts the so-called equilibrium rate of unemployment and labour income’s share of national income. The model is closed so that the rate of GDP growth does not have a long-term impact on equilibrium rate of unemployment, although it does have an impact on the unemployment rate in the short run. The demand effect of GDP growth gradually ceases to affect the unemployment rate after four years due to the reaction in wage formation. The long-term relations of the model are depicted in figure 8, for details see Alho (2002a). The exogenous required return on capital (real interest rate) determines how much of income firms are able to allocate to wages over the long run. The wage demands of trade unions and the outcome of wage negotiations are affected by the level of unemployment as well as other factors. The interaction of these phenomena determines equilibrium employment and the unemployment rate in this framework.
When assessing the impact of centralised income policy agreements, we must take a stand on whether such a settlement means a more modest increase in labour costs over the long run than the alternative of a union-level agreement. We answered affirmatively to this question above when the coefficient C(2) differed significantly from minus unity. But the question is deeper than this. We have to ask whether resorting to a comprehensive and collective income policy agreement one year spawns such pressure in the labour market that the next negotiating round inevitably ends up with union-level settlement, which in turn leads to an acceleration of the rise in labour costs. Centralised income policy agreements would thus not have long-term dampening effects.
Technically this question means taking a stand on the time path of an exogenous factor and its autocorrelation. Another interpretation is that, as described above, the shift from a centralised agreement to a union-level agreement is a consequence of structural developments in the labour market, i.e. diverging productivity trends, so that this shift cannot be blamed on the centralised agreement per se.
For this reason in the following we evaluated two cases:
1) Centralised income policy agreements have a temporary impact, given by model (9), in the medium run on real wage demands, not a long-term impact, but the effect is neutralized after a certain spell of time.
2) We assume that a modest income policy is practiced permanently so that wage demands are moderated in a way that, with a given rate of unemployment, the share of labour income within national income is permanently reduced by one percentage point (this is because the rise in real wages slows down by two percentage points and the share of labour income by about half), see the dotted line in figure 8.
For the first case the labour market model gave the following result, see figure 9.
In this alternative, it is simply assumed that wage formation takes the form of the wage equation with respect to real wages from year 3 onward, while initially thanks to the centralised agreements, wage increases were lower than the alternative in years 1 and 2, as given by model (9). We see that, as a result of centralised income policy agreements, unemployment decreases by about one percentage point over the medium term. However, the effect is not permanent and will disappear over the long run.
(Illustrasjon mangler)
The result of the second case is displayed in figure 10. In this case, the unemployment rate again falls over the medium term by about a percentage point, while in the long run, the downward impact on the unemployment rate would eventually stabilise at about half a percentage point.
In any case, we can see that there is not a significant difference between these two cases in the short and medium term regarding trends in the unemployment rate.[32]
(Illustrasjon mangler)
Furthermore, centralised income policy agreements do not have long-term effects on the labour share of national income, but rather it remains unchanged. This is illustrated by figure 8, where the wage setting curve shifts to the left as a result of centralised income policy agreements, but the ability to pay-curve remains unchanged. A wider range of policy simulations is presented in the study Alho (2002a). So we get,
Outcome 6. The medium-run impact of centralised income policy agreements has been, on average, a reduction in the unemployment rate by one percentage point as compared to union-level agreements. They raise the profitability of firms in the short run so that the share of non-labour income in GDP rises by one percentage point, but this effect will disappear in the long run.
The connections between fiscal policy and income policy, such as curbing of inflation and employment work so that both of these avenues lead to a more expansive policy and short-term economic growth similar to that in the alternative, union-level agreements. These impacts have not been incorporated into the analysis, however, for the time being.
The above-presented simulations have been made by looking at the change in the equilibrium of the model. Let us assume that the economy is not in equilibrium, but rather, for example, unemployment is higher than its equilibrium level. Then a policy that speeds up the adjustment to the equilibrium has an effect on the economy that is visible also in the long run. Thus centralised income policy agreements can in principle have a permanent impact on employment trends also in case 1.
Is the above-presented view of the significance of centralised income policy agreements too positive? This may be the case to some extent, because we have omitted differences in circumstances under which labour market settlements are made. In a way, we assume that every year there is a possibility to negotiate a moderate centralised income policy agreement or its alternative a union-level agreement, which is technically the case, but in practice the type of agreement is dictated partly by economic and labour market conditions, which means that these two types of agreements are not necessarily genuine alternatives for each another. These considerations would lead us to expand the model in order to explain which kind of agreement would be made in different situations. The analysis above is the answer we obtain when a standard economic policy analysis is carried out.[33]
Conclusions
We have sought to provide a theoretical framework for income policy and derive quantitative estimates of its impact. This is, of course, in many ways a preliminary and incomplete framework for income policy and its position in the Finnish labour market. It is our intension to expand the framework in various ways in later stages of the project.
What is the answer to the question of whether it is worthwhile to continue income policy in Finland? Moderate income policy has many good qualities, but also problems owing to its inflexibility. It can be difficult to reconcile these conflicts between different goals. Above we presented quantitative estimates of centralised income policy agreements. We find that they play a role in macroeconomic stabilization and therefore it is worthwhile keeping them as one tool that can be used if necessary. When assessing the results, it is good to keep in mind that there are different kinds of centralised income policy agreements in terms of costs. More detailed analysis of this phenomenon will be left for future research.
Acknowledgements
This is part of the research project ”Rules of the game in the labour market: Industrial relations, the bargaining system and income policies in the 2000s”, carried out in cooperation by ETLA, The Research Institute of the Finnish Economy, and the Labour Institute for Economic Research, funded by The Finnish Work Environment Fund. A previous version of this paper was presented in June 2002 at the XIX Economists’ Summer Seminar in Jyväskylä and in September 2002 at the conference organized by Fafo in Oslo. I wish to thank the participants of these workshops as well as Antti Kauhanen, Jaakko Kiander, Jukka Lassila, Jussi Mustonen, Hannu Piekkola, and Kari Sihtola for useful comments and inspiration. The responsibility for the paper nevertheless lies solely with the author.
References
Alho, K. (2002a): ”The Equilibrium Rate of Unemployment and Policies to Lower It: The Case of Finland”, ETLA Discussion Paper, No. 839.
Alho, K. (2002b): ”Tuottavuus, suhteelliset palkat ja sopimustoiminta” (Productivity, Relative Wages and the Bargaining System, in Finnish), presentation at the seminar
organised by the project team, Helsinki 9.1.2002.
Aronsson, T.M., K.-G. Löfgren and T. Sjögren (2002): “Wage Setting and Tax Progressivity in Dynamic General Equilibrium”, Oxford Economic Papers, Vol. 54, No. 3, July, 490-504.
Calmfors, L. and J. Driffill (1988): ”Bargaining Structure and Macroeconomic Performance”, Economic Policy, 6, April, 13-61.
Holm, P. and A. Romppanen (1999): “Vuosien 1995 ja 1997 tulopoliittisten sopimusten Työllisyysvaikutuksista” (On the Effects on Employment by the Income Policy
Agreements of 1995 and 1997 (in Finnish), Government Institute for Economic Research, Discussion Papers, No. 202.
Koskela, E. and R. Uusitalo. (2002): ”The Unintended Convergence: How the Finnish Unemployment Reached the European Level”, CESIfo Conference ‘Unemployment in Europe: Reasons and Remedies’, December.
Marjanen, R. (2002): “Palkkaratkaisujen sisältö ja toteutuminen tulopolitiikan aikakaudella” (The Content and Realisation of Wage Settlements in the Period of Income Policy, in Finnish), ETLA series B, No. 188.
Nickell, S. (1998): “Unemployment: Questions and Some Answers”, The Economic Journal, Vol. 108, No. 448, May, 802-816.
Ripatti, A. and J. Vilmunen (2001): “Declining Labour Share – Evidence of a Change in Underlying Production Technology?“, Bank of Finland, Discussion Papers, No. 10.
Ruutu, J. (1997): ”Suomalainen työehtosopimusjärjestelmä, palkat ja inflaatio” (The Finnish System of Wage Agreements, Wage Rises and Inflation, in Finnish), ETLA Discussion Paper, No. 611.
Uusitalo, R. (2002): ”Changes in the Finnish Wage Structure: Will Demand and Supply Do?” The Scandinavian Journal of Economics, Vol. 104, No. 1, 69-85.
[25] This means that the wage depends positively on productivity. In fact, the elasticity can be somewhat greater than one, see equation (3). This differs from the so-called monopoly union model, where θ = 0, and the trade union alone decides about the wage, where at the optimum εL < -1, in order that wage W would be greater than unemployment compensation b, see e.g. Aronsson et al. (2002). The negotiation model presented above is in many respects more realistic, because in practice the long-run elasticity of labour demand with respect to real wages is usually at most unity.[]
26 The lower the location of union 1’s indifference curve, the greater the benefit to union 1. Correspondingly, the benefit to union 2 grows as it shifts along its reaction curve to the left.[]
27 In figure 2 this is not the case. But if the indifference curves are “flat”, it can be that the indifference curve of union 1 going through point R’ lies below point R, and the corresponding indifference curve of union 2 lies to the left of point R, and the situation mentioned above emerges. It is also possible that the unions would like jointly to proceed to higher wages than those in R’, if their bargaining power is higher in CIPA as compared to separate negotiations. This enlargement is omitted in the following.[]
28 One exception is Holm and Romppanen (1999).[]
29 In this connection Ruutu’s index has simply been extended after 1996 using Marjanen’s CIPACOV index.[]
30 The goodness of fit of the contract wage equation is fairly low (only 13 %), while the fit of the earnings index is fairly high (88 %). We also included the share of social security contributions in the wage equation. It obtained a negative coefficient, which is in line with expectations, i.e. raising social security contributions decreases the ability to pay wages by the firms, but the coefficient was not significant and the coefficient for the CIPA indicator became insignificant. We also tested the value added tax and social security contributions in the price equation. They were not significant.[]
31 This share averaged 54 per cent in the 1980s and 51.8 per cent in the 1990s, but the reduction indicated by the model was only 0.25 percentage points.[]
32 It is of course possible that income policy will ”self-destruct” as mentioned above, i.e. lead to pay hikes greater than the wage norm in years 3 and 4. This alternative leads at first to practically as great a decrease in the unemployment rate as in figure 9, but it would return to its initial level already in year 4.[]
33 There is an independent simultaneous use of the same indicator of CIPA as here to analyse nominal wage moderation by income policies in Finland, see Koskela and Uusitalo (2002).