Changed wage formation in a changing world?

Transcription

1 Changed wage formation in a changing world? BY KENT FRIBERG AND EVA UDDÉN SONNEGÅRD Employed in the Monetary Policy Department. During the 199s the conditions for the wage formation process changed. A new route was prescribed for stilisation policy, fiscal policy was tightened, the budget was consolidated and the krona was allowed to float. This article is an attempt based on a relatively simple wage formation model to answer the question of whether the wage formation process has changed. This question is important for the Riksbank s assessment of future developments in wages and prices. The result indicates that it is mainly the low inflation expectations and the low demand that explain the lower rate of wage increase. The low inflation expectations are probly due to the change in stilisation policy. It is often claimed in the economic policy It is often claimed in the economic debate that wage formation has changed, policy debate that wage formation particularly during the latter half of the has changed, particularly during the 199s. The 199 round of wage settlements latter half of the 199s. is used as an example of a new wage formation model, where co-ordination has been combined with local wage setting. Assessments of wage trends are important in the Riksbank s monetary policy work, as monetary policy is based on forecasts of inflation trends one to two years ahead. A change in wage formation therefore affects the Riksbank s assessment of future wage trends and price trends. Ever since the beginning of the 199s, the collective wage negotiations have been pursued in an economy with new stilisation policy conditions, where fiscal The authors would like to thank Claes Berg, Villy Bergström, Lars Heikensten, Per Jansson, Hans Lindblad, Staffan Viotti and Anders Vredin for their valule views. We would also like to thank the Riksbank s scientific advisors Peter Englund, Torsten Persson and Lars E.O. Svensson, as well as other participants at a seminar held at the Riksbank for their valule comments on an earlier version of this study. We would also like to thank Christina Eurén at the National Mediation Office for providing us with data.

2 policy is aimed at budget consolidation and monetary policy is aimed at price stility. At the same time, the negotiations have gradually changed in nature. As early as the beginning of the 19s, the traditional wage formation model was left behind and negotiations began to be held at industrial sector level. According to what is known as the Calmfors-Driffill hypothesis (the C-D hypothesis), however, negotiations at industrial sector level provide the least beneficial macroeconomic results. The C-D hypothesis assumes that settlements at both central level and local level will give lower real wages and higher employment than settlements at industrial sector level. 1 The cost crisis at the beginning of the 199s led to the appointment of a mediator in the collective wage negotiations to get to grips with the increases in wage costs. The Rehnberg Agreement came to mark the beginning of a period with significantly lower nominal wage rises than in previous years. The two most recent rounds of wage negotiations showed a greater element of local wage formation, while the degree of co-ordination increased through the 1997 Manufacturing Agreement. According to the C-D hypothesis, the institutional changes may have subdued wage rises to some extent. Nominal wage increases have also been lower during the 199s than in previous decades. At the same time, real wages have increased more than expected as the inflation rate fell below the expectations of the social partners (see Figure 1). Real wages increased by approximately 3 per cent on average during 1995, which can be compared with an annual rise in real wages of approximately 1 per cent between This article makes an attempt on the This article makes an attempt on basis of a relatively simple wage formation the basis of a relatively simple wage model to answer the question of whether formation model to answer the wage formation has changed during the question of whether wage formation 199s. Previous empirical studies have shown has changed. that the wage drift, that is to say, wage trends over and ove the collective settlements, can be relatively well explained by the demand pressure on the lour market and inflation expectations within the framework of an expectations-augmented Phillips curve. Here a corresponding study is made both of total wage increases and the wage drift in manufacturing during the period The result from the simpler model used indicates that it is mainly the low inflation expectations and the low demand that explain the lower rate of wage increase during the past decade and that there has been no structural change in the actual wage formation process. 1 See Calmfors & Driffill (19). 3

3 5 Figure 1. Actual and expected rate of increase in real wages Per cent Expected real wage trend employer Expected real wage trend employee Actual real wage trend Note. Expected real wage increases refer to expectations two years earlier and are calculated as expected nominal wage increase minus expected CPI inflation. Wage increase refers to the percentage wage increase calculated over all sectors of the economy excluding social security contributions but including wage drift. Sources: Prospera Research AB and Statistics Sweden. First we give an outline of macroeconomic developments and the negotiation systems used during the past three decades. Then comes a structured discussion on inflationary trends on the basis of the sub-components of the wage drift, followed by a run-through of the wage drift in different sectors of the economy. Subsequently we report the results from the empirical study, and finally there is a summary of the study and discussion of our conclusions. Wage developments since the 197s When looked at from an historical perspective, the wage negotiation systems have Wage formation since the 197s can be roughly divided into three periods. changed as macroeconomic policy has changed. When looking back over developments since the 197s, wage formation can be roughly divided into three periods: (1) the traditional central wage formation model , () a period with more decentralised wage formation and (3) wage formation in accordance with the Rehnberg Agreement and under new stilisation policy conditions The period was marked by central settlements between the Swedish Trade Union Confederation (LO) and the Swedish Employers Confeder-

4 ation (SAF) with elements of widespread investment in low-wage earners. During this pe- The period was marked by central wage settlements between riod the full employment target was given top LO and SAF with elements of priority in economic policy. This meant that widespread investment in low-wage high nominal wage increases were accommodated partly through an expansion of the pub- earners. lic sector. The reconciliation policy in the wake of the first oil crisis forced the economy into a cost crisis and the krona was devalued both in 197 and twice in At the end of the period, the krona was devalued on two further occasions by a total of per cent (191 and 19). Wages increased by an annual total of approximately 1 per cent in the economy throughout this entire period (see Figure ). The purpose of the first four devaluations mentioned was to re-estlish competitiveness in the Swedish export industry. On the other hand, the devaluation in 19 known as the super devaluation can be described as an aggressive devaluation to safeguard the competitive advantage of the export industry for a couple of years ahead. Real wages were adjusted downwards in that inflation rose; the inflation rate peaked was at an all-time high in 19, 13.9 per cent. The social partners had thus learnt that The years can be the fixed exchange rate policy did not set any described as an in-between period, upper limit for the wage costs, although the with alternating central negotiations 19 devaluation came to be the last of its and negotiations at sector level or kind. In 193 the Swedish metalworkers individual trade union level within union did not participate in the central negotiations, but instead reached a separate LO s and SAF s fields. agreement with the Swedish Engineering Employers Association and andoned the traditional central wage formation model. The years can be described as an in-between period, with alternating central negotiations and negotiations at sector level or individual trade union level within LO s and SAF s fields. During this period the credit market was deregulated, which can be regarded as the starting signal for an expansive economic policy. Classical overheating tendencies with widespread increases in prices and wages finally led to a new cost crisis starting in The inflation rate increased on average by approximately 7 per cent a year between 193 and 199. Wages increased by approximately per cent a year during the period, which meant that real wages only increased by approximately 1 per cent. The cost crisis, together with the international recession, led to open unem- See Jonung (1999). 5

5 Figure. Wage growth in Sweden Per cent Wage drift Total wage rise Sources: The National Institute of Economic Research and Statistics Sweden. ployment quadrupling between 199 and During the period 1993, the 1993, to more than per cent. The government appointed a mediator in the 1991 wage central wage negotiations took place under new stilisation policy negotiations in order to bring wage costs conditions. under control. The Rehnberg Agreement brought much lower wage growth than the previous years. During the years 1993 the central wage negotiations were carried out under new stilisation policy conditions. The exchange rate was floating and devaluations were no longer an option to counteract deteriorations in competitiveness in the event of excessively high wage rises. Instead, monetary policy was aimed at price stility. The Riksbank announced a price stility target that involves CPI inflation being per cent with a deviation interval of ±1 percentage point. The rate of wage growth was halved compared with the previous period; wages increased by approximately per cent on average during Real wages doubled and increased by approximately per cent on average during the period. When the 1995 wage negotiations began, When the 1995 wage negotiations inflation expectations had not yet come began, inflation expectations had not down to the current levels and monetary policy had not won complete credibility (see yet come down to the current levels and monetary policy had not won Figure 3). Attempts to co-ordinate the wage complete credibility. negotiations failed and conflicts arose. How-

6 3.5 Figure 3. Inflation expectations two years ahead of social partners Per cent Inflation expectations, employers Inflation expectations, employees Source: Prospera Research AB. ever, the element of local negotiation increased within certain settlement areas and the local partners were le to freely dispose of the wage increases that had been achieved in the central negotiations. The largest changes were achieved for salaried workers in the private sector, where some trade unions signed only local settlements. Wages increased by almost 5 per cent a year throughout the entire economy during Inflation expectations had fallen to below per cent for two years ahead when the 199 wage negotiations began. The co-ordination between the social partners succeeded in these negotiations. An agreement that included regulations on negotiation procedures, known as the Manufacturing Agreement, was signed for the manufacturing industry at the beginning of This Manufacturing Agreement was then followed by an agreement on negotiation procedures for salaried workers. Unlike the 1995 wage negotiations, LO s recommendation for co-ordinated wage demands was adopted by all member unions. On the employers side, there was consultation between the various employer organisations during the wage negotiations. The trend towards a more local wage formation process continued, mainly in the salaried workers field. Some of the trade unions signed general settlements where the greater part of wage formation was distributed locally. In other trade unions, for instance in the IT field, the settlements did not contain any central wage rises at all. The 199 wage negotiations were also distinguished by the fact that many sectors managed to reach agreement without 7

7 even threatening conflict measures. 3 The result was modest wage increases of approximately 3.5 per cent on average for the years 199. A survey of the negotiating systems indicates A survey of the negotiating systems a gradual shift towards a greater local element in wage formation. At the same time, indicates a gradual shift towards a greater local element in wage the degree of co-ordination increased in the formation. most recent wage rounds, particularly with regard to the manufacturing industry. The new Mediation Institute and the estlishment of new agreements on negotiation procedures can also be said to have an influence in this direction. This means that the most favourle positions according to the C-D hypothesis are utilised at the same time; the local situation, where wages are set according to local demand and supply, and the central situation, where the social partners take their socio-economic responsibility. However, it is far too early to determine whether the 199 wage rounds are the beginning of a new period in wage formation. This articles focuses instead on whether wage formation has undergone structural changes since the beginning of the 199s. What constitutes wage drift? In economies where wage settlements are Wage drift has to be calculated signed for large groups of wage earners, the afterwards, as it is by definition the wage rises consist of both agreed wage increases and what is known as wage drift. difference between the negotiated and actual wage increases. When wage settlements have been signed for a particular period of time often two to three years ahead it is possible to calculate the agreed wage rises for a whole industry, sector or for the entire economy. However, the wage drift must be calculated afterwards, as it is by definition the difference between the negotiated and actual wage increases. When making an assessment of wage growth over a forecast period, where the settlements are largely complete; it is therefore very useful to be le to use a suitle model to explain wage drift. The two wage concepts fulfil partially different functions in the wage formation process. In the wage negotiations, for instance, fair distribution of income may take on great importance, while the wage drift lies outside the control of the negotiating parties. However, the parties can make an estimate of the expected wage drift when they calculate a payroll expense framework prior to the negotiations. There are two main explanations behind the occurrence of wage drift. The 3 See SOU (199).

8 first is that wage drift is determined by market forces. When wage settlements are signed There are two main explanations behind the occurrence of wage drift. for large groups of wage earners within many The first is that wage drift is different industries and companies, the negotiated wage increases may deviate from the determined by market forces. The wage adjustment that takes place for wage increases determined by supply and the individual over and ove the demand within a certain industry, a certain agreed wage rise can be termed a region, or for a particular individual. It may real wage drift. also be the case that the supply and demand conditions change during the period that the settlement is in force, which is increasingly often a period of three years. In addition, high negotiated wage increases in one part of the lour market can have spread effects to other parts of the lour market, as a result of compensatory wage increases. The wage adjustment that takes place for the individual over and ove the negotiated wage rise regardless of the reason can be termed a real wage drift. The second explanation is statistical and The second explanation is statistical due to aggregation problems. The lour and due to aggregation problems. This force within a company or industry changes type of wage drift is called over time, while statistics only report average structural and has probly gained wage increases for the lour market as a greater significance during the 199s. whole at a certain point in time. This type of wage drift is called structural and has probly gained greater significance during the 199s. During periods when there are reductions in the work force and it is the low-wage earners that are made redundant, the average wage increases compared with previous years, merely because of a change in the work force. When the economy then improves, the average wage may increase if companies mainly recruit well-educated workers. An evaluation of the latest round of wage negotiations shows that the structural wage drift in the manufacturing industry was positive between 199 and 1999; approximately. percentage points of the average wage increase of 3. per cent was due to changes in the work force. Empirical wage formation studies of Swedish date in recent years have been aimed at explaining the total wage increases with various models, while the wage drift in itself and the relationship between negotiated wage increases and wage drift is a relatively unexplored field. 5 Earlier studies prior to the beginning of the 199s See Carling et al. (). 5 See, for example, Johansson et al. (1999) and Forslund & Kolm (). The first-mentioned study covers wage formation in the mining and manufacturing industry during the period The last-mentioned study covers how lour market policy measures have affected wage formation in the Swedish business sector during the period

9 show that the wage drift in industry can be explained relatively well by expected price in- The wage drift in itself and the relationship between negotiated creases and pressure of demand on the lour wage increases and wage drift is a market on an expectations-augmented Phillips curve. However, attempts to also allow relatively unexplored field. negotiated wage increases to determine the wage drift have given mixed results. A couple of earlier studies show that higher negotiated wage increases lead to lower wage drift and vice versa. For instance, the widespread investment in low-wage earners during the 197s up to 193 contributed to a higher wage drift, although the effect was slight. 7 Wage trends can be inflationary if the From a monetary policy perspective, increase in wage costs exceeds the increase in it is the total wage increase or productivity, provided that the profit and rather the unit lour cost and not wage percentages are out of balance in the the wage drift that is important for long term. The size of the impact on inflation depends on the number of companies the development of inflation. that can immediately adapt their prices and how long the period of adjustment is until the economy returns to a long-term balance. From a monetary policy perspective, therefore, it is the total wage growth or rather the unit lour cost and not the wage drift that is important for the development of inflation. However, it can be assumed that the structural wage drift does not usually affect infla- However, it can be assumed that the structural wage drift does not usually tion, as it is normally matched by changes in affect inflation, as it is normally productivity. If newly recruited workers have matched by changes in productivity. a low productivity level at the beginning of their employment, a low wage is motivated, and this results in negative wage drift. On the other hand, a high wage at recruitment can be motivated by a high level of productivity. The positive wage drift that is then registered in the statistics, is not inflationary. However, there are examples where the structural wage drift can affect inflationary trends. A high structural wage drift can be due to the economy undergoing a major transformation with large changes in the lour force or to severe fluctuations in the economy. This could mean lower inflationary pressure when the transformation has been implemented, but the risk of inflation impulses arising can also increase during the adjustment, compared with the situation if economic trends had been stle. See Hibbs & Locking (199); Holmlund & Skedinger (199). 7 See Tson Söderström & Uddén-Jondal (195). See Sveriges Riksbank () estimates of a price formation model based on a new-keynesian Phillips curve theory. 5

10 The mechanisms affecting the real wage Wage drift can contribute to a more drift are more interesting. It is not possible to effective distribution of resources, draw any clear-cut conclusions on the connection with inflationary trends in this context. while it is the level of the negotiated wage increases that gives rise to The wage drift can contribute, for instance, to inflationary pressure. a more effective distribution of resources, while it is the level of the negotiated wage increases that can give rise to inflationary pressure. When consideration for fair distribution of income is an important element in the wage negotiations, it is proble that the negotiated wage increases will not be followed by a corresponding increase in productivity. Then the wage drift can ensure that the development of relative wages corresponds to the relative supply and demand situation. For instance, consideration for fair distribution of income led to extensive investments in low-wage earners by LO and SAF during the period , while the actual wage outcome provided a less compressed wage structure than was stipulated in the central wage settlements. This can be interpreted as the wage drift to some extent counteracting the ambition for a fairer distribution of income. 9, 1 However, the high wage increases combined with the accommodating economic policy led to the creation of a devaluation pressure in the competitive sector and the creation of inflationary pressure in the domestic sector. Wage drift is thus not a homogeneous Wage drift is not a homogeneous concept and therefore has no clear-cut effect concept and therefore has no clearcut effect on inflation. on inflation. However, there is still good reason to study wage drift separately from the agreed wage increases. The most obvious reason is that the negotiated wage increases are known, sometimes for several years ahead, while the final actual wages are determined by wage drift. A good assessment of wage drift is therefore necessary to make a good assessment of the total wage trend. The wage drift in different sectors The wage drift has changed in nature both over time and between different wage negotiation areas. 11 Since the advent of settlements at industry or trade union level, the wage drift only covers adjustments at company and individual level. During the period , when the economy became increasingly overheated, wage drift took on greater significance than in earlier years. This applied in particular to 9 As is appears difficult to get wages to move downwards, the wage drift did not completely counteract the ambitions for a fairer distribution of income. 1 See Hibbs & Locking (199). 11 See Eurén (199). 51

11 Figure. Wage growth in manufacturing Per cent c 75c 7 77c 7 79 c 1c 3c f 5c c 7 f 9c 9 91* 9* 93f 9 95f ** 99** Total wage increase Wage drift Note. c = central negotiations; f = trade union level negotiations; * = the Rehnberg Agreement; ** = the Manufacturing Agreement. Source: The National Institute of Economic Research. manufacturing, where wage drift comprised almost per cent of the total wage growth (see Figure ). The wage drift s percentage remained considerle during the greater part of the 199s and comprised approximately 3 per cent towards Figure 5. Wage growth and lour shortages for construction workers Per cent Total wage increase (left-hand scale) Wage drift (left-hand scale) Lour shortages (right-hand scale) Note. There is a break in the series of total wage increases and wage drift in 199. Sources: The National Institute of Economic Research and Statistics Sweden. 5

12 the end of the decade. Both the wage drift and During the period , when the total wage growth covary relatively well with the demand pressure in industry. 1 the economy became increasingly overheated, wage drift took on The negotiated wage increases in the greater significance than in earlier construction industry have traditionally been years. This applied in particular to small, while the wage drift has to a large the manufacturing industry. degree reflected market conditions (see Figure 5). During the end of the 199s, however, The negotiated wage increases in the the negotiated wage increases have had construction industry have greater significance for the actual wage traditionally been small. increases, as the demand for construction workers has been slight. There has only been a slight wage drift There has only been a slight wage in the central government and local government sector, and this has been mainly due to drift in the central government and local government sector, and this has structural factors. However, the lour shortage that is becoming increasingly evident been mainly due to structural factors. may cause greater market-determined wage drift in these sectors too (see Figure ). The possibility of a large wage drift was Figure. Wage growth in central government and local government Per cent Total wage increase, central government Total wage increase, local government Wage drift, central government Note. Break in the series in 199. Sources: The National Institute of Economic Research and Statistics Sweden. Wage drift, local government 1 See the section Empirical estimations of wage formation in manufacturing. 53

13 1 1 Figure 7. Wage growth for some trade unions between 199 and 1999 Per cent 1 1 The Hotel and Restaurant Workers Union The Commercial Employees Union The Municipal Workers Union The Metalworkers Union The Salaried Employees Union The Swedish Association of Health Professionals Negotiated wage increases Actual increases Wage drift Source: The Hotel and Restaurant Workers Union. used as an argument by the Swedish Association of Health Professionals to conclude a long-term settlement with low negotiated wage increases in Spring. Between 199 and 1999, the wage drift for nurses increased by almost as much as the negotiated wage increases (see Figure 7) Figure. Wage growth in commerce for salaried employees in the commercial sector Per cent Total wage increase Wage drift Note. Break in the series in 199. Sources: The National Institute of Economic Research and Statistics Sweden. 5

14 The private services sector has taken on In the private services sector, the greater importance during the 199s; approximately per cent of those employed element of local wage formation is extensive with regard to salaried worked in this sector in 1999, compared with employees, which can reduce the 35 per cent ten years earlier. The element of need for wage adjustment. local wage formation is considerle on the salaried employees side, which can reduce the need for wage adjustment over and ove the settlements. This can be one explanation for the low wage drift in the commercial sector in recent years, while the low demand has also had significance (see Figure ). In the hotel and restaurant industry, the turnover in personnel has been extremely high; it was over 7 per cent between 199 and The result has been a negative structural wage drift as new recruits have often been given a lower wage than those who have left their posts. (see Figure 7). 13 The survey of wage growth in different The trend of low wage increases sectors indicates that something may have continued throughout the entire happened in wage formation in recent years. decade, with the exception of the The Rehnberg Agreement , years 199 and which led to low nominal wage growth, was the beginning of a period with low wage growth during the entire previous decade, with the exception of the years 199 and Both negotiated wage Tle 1. Average total wage growth, negotiated wage growth and wage drift during the period 1995 to Per cent Sector Total wage Negotiated wage Wage drift Percentage growth growth wage drift Private sector Manufacturing Central government Local government Total economy Private sector Manufacturing Central government Local government Total economy Note. Total wage growth according to Statistics Sweden and the Riksbank. The figures for the total wage growth can differ from those produced by the National Institute of Economic Research, as they are aggregated using a different weight system. Negotiated wage growth according to the National Institute of Economic Research s calculations. 13 See Jacobsson (). 55

15 increases and wage drift declined in all sectors during 199 compared with , although the decline was slight within the central government and local government sector (see Tle 1). Empirical estimations of wage formation in manufacturing Wage drift is thus not a homogeneous concept with clear-cut inflation effects. It probly has less significance on inflationary trends when the structural part is relatively large compared to when the real individual wage drift is relatively large. However, it is still the negotiated wage increases and wage drift taken together, in relation to the development in productivity, that affects inflationary pressure. On the other hand, an analysis of the mechanisms of wage formation should ideally be based on a division of wage drift into its components. When the 199 wage negotiation rounds in industry were evaluated, the wage increases were adjusted with regard to the changed composition of the work force, in terms of both age and sex. It was thus possible to conclude that the real wage drift between 199 and 1999, when calculated afterwards, fitted into the payroll expense scope that when the negotiations began was judged to be availle with regard to the expected trends in inflation and productivity. 1 By using wage statistics based on individuals, Our purpose is to examine using a it is possible to make a better analysis of relatively simple wage formation wage formation than when aggregated wage model whether the wage formation data is used. However, the availility of data relationships have changed during on individuals is limited and the published the 199s. wage statistics are aggregated in one form or another. In our study, therefore, we analyse wage formation on the basis of the availle macrodata. We do not analyse here how changes in wage costs affect inflation; our purpose is instead to examine using a relatively simple wage formation model whether the wage formation relationships in manufacturing have changed during the 199s. The survey of wage growth ove indicates that the wage formation process may have changed during the 199s. The institutional changes in the wage negotiations and the changeover in stilisation policy can be expected to have two effects on wage formation. Firstly, the increased elements of local wage formation may have meant that it is the total wage increases that are best explained by market con- 1 See Carling et al. (). 5

16 ditions, while the wage drift is largely a statistical concept that has had its day in the fore- The institutional changes in the wage negotiations and the casting context. Secondly, the new stilisation policy regime may have meant that de- changeover in stilisation policy could be expected to have an effect mand no longer has the same effect on wage on wage formation. formation as before, or, expressed in a different way, that the inclination of the Phillips curve has changed. It is also possible that the effects of expectations of price rises on wage increases have changed if, for instance, the money illusion has earlier had greater significance for the social partners or whether there has been a change in other factors that affect the sluggish movement of wages. The first question can be illustrated by making empirical estimations of both the total wage growth and the wage drift in manufacturing and the second can be illustrated by testing for any structural change during the 199s. Many empirical studies made on wage If manufacturing leads the way in formation refer to the manufacturing sector. wage formation, estimations of wage This is partly because there are long historical series availle here, but also because growth in this sector can give an indication of future wage manufacturing can be seen as leading the development in the total economy. way in wage formation, as it has often been the first to sign wage settlements. According to the EFO and FOS models, the competitive sector which corresponds to approximately one-third of the lour market can be assumed to lead the way in wage formation. Manufacturing comprises an important part here, although the industrial composition in the K sector has changed over time. 15 If manufacturing leads the way in wage formation, estimations of wage growth in this sector can give an indication of future wage development in the total economy. We have also chosen to focus on wage formation in manufacturing. The theoretical starting point for the estimations is an expectations-augmented Phillips curve, where wage drift and total wage increases are determined by demand on the lour market and by expectations of price rises. Some equations also contain expectations regarding the development of productivity and the profit situation respectively. When estlishing the scope for wage increases, it is possible that the social partners use an intellectual framework based on the EFO model. Profit 15 The EFO model was introduced in Edgren et al. (197). The FOS model, which is a slightly modified version of the EFO model, was introduced in Faxén et al. (19). The competitive sector, or K sector as it is described in the original EFO model, was composed of forestry, fishing, mining, parts of the food industry, the metalworking and engineering industry, other industry apart from beverages and tobacco and transportation road. LO, for instance, appears to base its calculations on the EFO model, where wage increases are determined at least a little way ahead by developments in prices and productivity in the K sector. See Calmfors & Lundberg (197). See also the 1997 Manufacturing Agreement. 57

17 The theoretical starting point for the estimations is an expectationsaugmented Phillips curve, where wage drift and total wage increases are determined by demand on the lour market and by inflation expectations. expectations can be seen as an alternative measure for gauging productivity developments in the near future. The social partners could reasonly make an assessment of how profit shares are developing to bring out a desired growth in capital and improvements in productivity. 17 The estimated equations are: (1) WD t = α + α 1 LD t + α pt e + α 3 Qe + α t (π/y)e+ ε t 1t and () W t = β + β 1 LD t + β pt e + β 3 Qe + β t (π/y)e+ ε t t, where the wage drift WD t and the total wage increases W t for manufacturing blue-collar workers respectively are explained by the demand situation on the lour market LD t, inflation expectations pt e, expectations of productivity growth Qt eand expectations of profits (π/y)e t. The error terms are assumed to have the cus- Tle. The variles and their sources. The number of unfilled vacancies refers to 1 s of vacancies. Other variles are measured as a percentage Varile Dependent variles Rate of total wage increase for manufacturing blue-collar workers Rate of wage drift for industrial workers Lour market situation Percentage of openly unemployed members of manufacturing unemployment benefit fund Percentage of firms stating a shortage of manufacturing blue-collar workers Number of unfilled vacancies in manufacturing Inflation expectations Time-lagged CPI inflation Inflation expectations according to surveys Other variles The firms profits as a percentage of GDP at factor price, total economy Growth in lour productivity in manufacturing Source National Institute of Economic Research National Institute of Economic Research National Lour Market Board National Institute of Economic Research, Business Tendency Surveys National Lour Market Board Statistics Sweden Statistics Sweden, Household purchasing plans (HIP) Statistics Sweden, National Accounts Statistics Sweden, National Accounts 17 The econometric model is relatively simple. However, this model could be interpreted as an error correction model, where productivity growth and inflation expectations reflect the short-term relationships. Unemployment as a measure of the demand situation and profit shares then provide long-term relationships. 5

18 tomary properties and are denoted ε 1t and ε t respectively. Various demand measures and two alternative price expectation assumptions were used in the estimations, which were made using ordinary least squares, OLS (see Tle ). The expectations of productivity growth and profit shares respectively were assumed to be adaptive, i.e. they are determined by the outcome one period earlier. To answer the question of whether a structural change occurred during the 199s, breakpoint tests were made. Dummy variles were also introduced into the model to pick up any changes in wage formation during the 199s. 1 Households inflation expectations have The adaptive inflation expectations the highest correlation of all variles with change down to a lower gear during both wage drift and total wage increases (see the latter half of the 199s covaries Tle 3). However, this series from Statistics with the slowdown in the rate of Sweden s HIP surveys does not being until wage increase and therefore proved to be too short to provide robust estimates. As an alternative, it was assumed that the inflation expectations were adaptive and indicated by the inflation rate of the previous period. The adaptive inflation expectations change down to a lower level during the latter half of the 199s covaries with the slowdown in the rate of wage increase (see Figure 9). Of the alternative measures of demand, both wage drift and total wage increases covary to a relatively high degree with the shortage of professional workers, but the correlation is slightly higher between unfilled vacancies and total wage increases (see Figure 1). The most important estimation results are reported in Tles and 5. The estimated coefficients show how wage drift and total wage increases respectively are affected by a marginal change in the explanatory variles. The estimates show that around 5 per cent of both wage drift and total wage growth can be explained Tle 3. Correlations between dependent and explanatory variles Rate of total wage growth Rate of wage drift Rate of negotiated wage increases..3 Percentage of firms stating a shortage of lour..9 Unfilled vacancies..59 Percentage of open unemployed members of manufacturing unemployment benefit funds Inflation expectations according to HIP.7.75 CPI inflation (time-lagged) Growth in lour productivity (time-lagged).7.3 The firms' profits as a percentage of GDP at factor price (time-lagged) A dummy varile can be introduced in the estimations to control for a specific event or a structural change during the estimation period that is not explained by the model. Normally, the time of the event or the period of the structural change respectively are given the value 1 and other times in the series are given the value. 59

19 Figure 9. Wage growth in manufacturing and CPI inflation in the previous year Per cent Total wage increases Wage drift CPI inflation in the previous year Sources: The National Institute of Economic Research and Statistics Sweden. by the demand pressure on the lour market and by inflation expectations (the equations 1b, 1c, b and c). When the demand pressure is measured as the number of unfilled vacancies, a slightly better result is achieved for the total wage increases than when shortage figures are used (equation b). The result shows that 1 1 Figure 1. Wage growth in manufacturing and shortages of manufacturing bluecollar workers Per cent Total wage increases (left-hand scale) Wage drift (left-hand scale) Shortages (right-hand scale) Source: The National Institute of Economic Research.

20 wage drift increases by almost. percentage The estimates show that around - points when the number of vacancies increases by 1 or when inflation expectations in- 5 per cent of both wage drift and total wage growth can be explained crease by 1 percentage point (equation 1b). by the demand pressure on the The impact on the total wage increases is approximately twice as high; they increase by al- lour market and by inflation expectations. most. percentage points with corresponding changes (equation b). This means that both the agreed wage increases and wage drift are affected by roughly the same amount. per cent each if these changes occur. The result can be compared with estimates made by Eurén using quarterly data up to the end of The estimated coefficients in her study are well in line with the results in equations (1b) and (b). Tle. OLS estimation result. The dependent varile is the rate of wage drift for manufacturing blue-collar workers Varile (1a) (1b) (1c) (1d) (1e) (1f) Estimation period Constant **.5..9** (1.3) (.51) (1.1) (1.5) (.3) (1.3) Unemployment.5 (.53) Unfilled vacancies.19***.*** (.5) (5.) Lour shortage.73***.9***.*** (7.15) (.) (3.33) Inflation expectations according to HIP.3** (.7) according to CPI ( 1).***.3***.197**.9***.15* (3.1) (3.) (.7) (.7) (1.77) Profit share ( 1).*.**.15 (1.99) (.9) (1.7) Lour.13.5 productivity ( 1) (1.3) (.3) AGREEMENT (1.) REHNBERG-.37 MONETARYPOL (.3) D.W Adjusted R # Observations Note. * (**) *** indicates significance at 1, 5 and 1 per cent levels. Absolute figures of t-value in brackets. 19 See Eurén (1991). According to equation (1b), a change in the number of unfilled vacancies of 1 involves a change in the wage drift of.19 percentage points. The corresponding effect in Eurén s estimates is..17 percentage points. According to equation (b), the rate of wage increase rises by.3 percentage points if inflation expectations change by 1 percentage point. The corresponding estimation by Eurén gives an effect of.37 percentage points. 1

21 Tle 5. OLS estimation result. The dependent varile is the rate of total wage growth for manufacturing blue-collar workers Varile (a) (b) (c) (d) (e) (f) Estimation period Constant 3.7**.*** *** 3.75*** 1. (.9) (.1) (1.5) (.9) (3.) (1.5) Unemployment.95 (.1) Unfilled.35***.***.35*** vacancies (.) (.99) (.13) Shortage, industrial.11***.113*** workers (5.) (5.) Inflation expectations according to HIP.7*** (3.95) according to CPI ( 1).3***.***.***.1***.5** (.5) (.5) (.) (.5) (.) Profit share ( 1).59***.***.9** (3.3) (3.1) (.3) Lour.7.95 productivity ( 1) (.59) (.) AGREEMENT ** (.) REHNBERG- 1.3 MONETARYPOL. (.9) D.W Adjusted R # Observations Note. * (**) *** indicates significance at 1, 5 and 1 per cent levels. Absolute figures of t-value in brackets. According to theory, the coefficient before inflation expectations is lower than 1, which In many empirical studies, the coefficient before inflation means that it is the rate of increase of real expectations is lower than 1, which wages that is important in the wage formation means that the rate of increase of process. However, in many empirical studies, real wages declines as inflation rises. the coefficient is lower than 1, which means that the rate of increase of real wages declines as inflation rises. 1 The same result was obtained in our estimations. This can be due to several factors. The time overlap of various wage settlements and the lengths of settlement periods mean that wages move slowly. An alternative explanation is that wage earners experience some measure of money illusion and that the rise in nominal wages also has significance in the wage negotiations. When the model is augmented to include adaptive profit expectations, the 1 See Calmfors & Lundberg (197).

22 degree of explanation increases to around 7 Expectations of profit shares are per cent for wage drift and around per clearly significant and raise the cent for total wage increases (equations 1e degree of explanation considerly. and d respectively). The adaptive productivity expectations are not significant. This The result indicates that wages are adapted in the long term to a certain could be due, for instance, to productivity level of the wage share. data being of poor quality or to the expectations picture looking different. However, expectations of profit shares are clearly significant and indicate that wages are adapted in the long term to a certain level of the wage share. The impact from inflation expectations will now be higher, particularly for total wage increases. A change of one percentage point in the adaptive inflation expectations means that the total wage increase changes by.. percentage points (equations d and e). This can be interpreted as wage increases in the short term not being adapted to inflation expectations, which means that the growth rate for real wages declines (equation b). When a longrun term is introduced via profit shares, the adaptation becomes almost complete, that is to say, the growth rate for real wages remains unchanged in the long run (in any case for equation e). The reported results indicate that the One way of investigating whether a specified model explains wage formation relatively well. However, the stilisation policy structural change has occurred is to use dummy variles in the regime during the 199s and the institutional equations. changes in the wage negotiations could have affected the wage formation process. If this is the case, the effects on wage formation may have been miscalculated. One way of investigating whether a structural change has occurred is to use dummy variles in the equations. Two dummies were used; one for the period when extremely high wage settlements were signed on the lour market (AGREEMENT757) and one for the period when the Rehnberg Agreement was signed and there was a change in monetary policy (REHNBERG-MONETARYPOL). The only dummy that affected the result was the one for , and this was regarding the total wage increases (equations 1f and f). However, the introduction of the dummy meant that the coefficient before inflation expectations fell back to., i.e. not even half of an increase in inflation expectations has an impact on wage increases. The relatively simple model thus cannot clearly illustrate how great an impact inflation expectations have on wage formation. The dummy for a change in wage formation at the beginning of the 199s had no significance. Another way of testing for any structural changes is to use 3

23 Tle. Result of Chow s breakpoint test Equation (1a) 11.1 (.).95 (.7).5 (.1).3 (.11) (1b). (.) 1. (.3) 1.53 (.3) 1.5 (.31) (1c) 1.39 (.7) 1.55 (.3) 1.7 (.1).35 (.1) (1d) 3.59 (.) 1.3 (.).1 (.1) 1.57 (.) (1e) 1.3 (.3).9 (.51) 1.15 (.37) 1.1 (.3) (a).53 (.7).5 (.).39 (.7).35 (.79) (b) 1. (.31).9 (.5).53 (.7).1 (.9) (c). (.) 1.1 (.17) 1. (.17) 1. (.17) (d).5 (.). (.9).3 (.7).35 (.7) (e).97 (.) 1.3 (.3) 1. (.) 1.5 (.1) (1b)* 3.99 (.3).75 (.) 7.3 (.). (.) (1c)* 1. (.31) 5.1 (.1.39 (.1).5 (.1) (b)*.7 (.) 5. (.1) 5. (.1). (.) (c)* 3.5 (.) 5.7 (.1) 5.1 (.1).3 (.) Note. F-statistics and p-values (in brackets) are stated in the tle. The null hypothesis (H ) in the test is that no structural change has occurred at the actual point in time. A high value for the F-statistics means that the null hypothesis can be rejected. Bold text means that the null hypothesis can be rejected at the five per cent level. *) In the models the measure used for inflation expectations is excluded as an explanatory varile. Chow s breakpoint est. This test indicates a The dummy for a change in wage structural shift in the mid-197s for the total formation at the beginning of the wage increase. However, the test result does 199s had no significance. The not support the hypothesis that a structural conclusion that can be drawn from change occurred at the beginning of the this is that the stilisation policy 199s. It is also possible to draw the conclusion here that the stilisation policy change change is reflected in the low inflation expectations. is reflected in the low inflation expectation. If inflation expectations are excluded from the equations, the Chow test sure enough also indicates that a structural change has occurred (see Tle ). To summarise, the empirical estimations indicate that there has been no fundamental change in wage formation and that the connections reported in earlier studies have not changed apprecily during the 199s. A reasonle explanation for this result is that the change in monetary policy is caught up in the model as inflation expectations have declined slightly at the same time as the demand for lour remained low during the greater part of the 199s. Summarising discussion During the 199s the conditions for the wage formation process were changed. The accommodating economic policy of the 197s and 19s with its ensuing cost crises and devaluations, as well as the huge fall in international demand at