Insider forces, asymmetries, and outsider ineffectiveness: empirical evidence for Norwegian industries, 1966-87.
How important is industry specific performance in shaping Norwegian industry pay? This is the main question addressed in the present paper which uses annual data for 117 industries for the time period 1966-87. The results imply that increased factor income per hour worked by 10% will increase industry wages by 2%. This finding is not in accordance with the results in Holmlund and Zetterberg (1991) who report small and insignificant insider effects for all the Nordic countries.
It may be argued that insider forces are less important in countries with centralized wage setting as compared with economies where wages are determined at the industry or firm level. Indeed, with complete centralization, there is little scope for industry wage response to industry specific performance. Although Norway is usually characterized as an economy with a highly centralized and coordinated wage setting system, we should keep in mind that wage bargaining takes place at both the national and the firm level. Since wage drift has contributed to more than 50% of total wage increases during our sample period, local wage bargaining implies considerable scope for firm or sector specific rentsharing.
The importance of insider forces in wage determination may vary both across industries and over time. First, institutional evidence reported in Holden (1989) and Rodseth and Holden (1990) indicate that firm level bargaining has become more important in Norway during our sample period as the share of wage drift increases. If centralization really matters, we should expect the effect of insider variables to increase over time. This issue is investigated by estimating wage equations for different sub sample periods.
Second, insider effects may be more important in expanding industries as compared to declining industries, and similarly, wages may be more responsive to insider variables in good times than in bad times. Such asymmetric insider effects imply downward wage rigidity and will tend to put more pressure on employment when times are bad. The empirical evidence on asymmetric insider effects is rather mixed, see Brunello and Wadhwani (1989), Nickell and Wadhwani (1990) and Layard et al. (1991, p. 209) concluding that 'this area, especially, deserves further research'. Hence, another issue will be to test asymmetric insider effects in the Norwegian wage formation process.
In the literature, Norway is usually counted among the countries with high real wage responsiveness to unemployment, and a high degree of wage flexibility is identified as a key mechanism behind Norway's favourable unemployment performance. This view is contradicted by the results in Johansen (1995) who shows that the Norwegian wage curve becomes flat for all but very low levels of unemployment. A final issue will be to test the robustness of this result using our disaggregate data set.
The plan of the paper is as follows. Section 2 summarizes some features of the data. Section 3 presents the theoretical background, while empirical specification of the dynamic wage equation will be discussed in Section 4. Empirical results are reported in Section 5, while concluding comments are in Section 6.
2. Inter-industry wage dispersion and industry performance.
Solidarity wage policy, social norms, and the notions of fair relative wages may generate a stable wage structure despite the fact that local wage bargaining implies considerable scope for industry specific rent sharing. Holmlund and Zetterberg (1991, p. 1012-13) argue that social norms are especially important in a two-tiered bargaining system if the local unions respect centrally formulated guidelines for wage policy.
Figure 1 shows that the Norwegian inter-industry wage dispersion is very small and remains stable during our sample period, while Table 1 reveals that the industry wage structure is stable as the correlation of log hourly wages between different years is very high. Industries with high wages in 1962 are likely to remain high wage industries also at the end of our sample period. From Fig. 1 we see that the dispersion in value added per hour worked is much larger than the inter-industry wage dispersion. In particular, the dispersion in value added per hour worked fluctuates more and increases during the 1980s.
The persistent wage structure, especially taken together with the evidence on relative industry performance, favours the hypothesis that relative wages are unaffected by industry specific variables. However, the evidence on relative wages reported in Elgsaether and Johansen (1993) reveals that deviations from long-run equilibria are rather usual and seem to last for a relatively long time. Finally, unit root tests for panel data indicate a rather weak tendency for relative wages to revert to equilibrium levels after an innovation. When the first difference of relative wages is regressed against lagged levels, including industry specific fixed effects, we obtain
[Delta][(w[c.sub.i] - wa).sub.t] = -0.13[(w[c.sub.i] - wa).sub.t - 1] + [[Alpha].sub.i], [Tau] = -15.01 (1)
TABLE 1 Descriptive statistics
Correlation between 1987 Mean Values and selected years
wc va wc va
1987 4.76 4.98 1.00 1.00
1980 4.07 4.31 0.96 0.85
1975 3.58 3.82 0.93 0.81
1970 2.88 3.19 0.85 0.76
1965 2.41 2.73 0.80 0.73
1962 2.45 2.18 0.79 0.62
wc = log wage costs per hours worked, va = log value added per hours worked.
where the 'Dickey-Fuller' test for unit root, [Tau], is slighty below the 5% critical value equal to -14, see Levin and Lin (1992). On the other hand, there is a strong tendency for the wage share to revert to 'normal levels', and we firmly reject the null that the wage share is random walk since
[Delta][(w[c.sub.i] - v[a.sub.i]).sub.t] = -0.40[(w[c.sub.i] - v[a.sub.i]).sub.t - 1] + [[Alpha].sub.i], [Tau] = -26.23 (2)
The result that the wage share is stationary may reflect causality from wage costs to prices, not necessarily that industry wages are affected by the industries' ability to pay. Hence, more detailed dynamic modelling, treating industry specific variables as endogenously determined, is in order. But first we discuss the theoretical background of the wage equations to be estimated below.
3. Theoretical background.
The key issue is to investigate the importance of insider forces in the Norwegian wage formation process. The bargaining model presented in Nickell and Wadhwani (1990), Nickell and Kong (1992) and Layard et al. (1991, Ch.2) seems well suited as a theoretical framework for such an analysis. It is assumed that wages are determined by bargaining between the union and the firm. The firm is assumed to set prices and employment in order to maximise profits, and faces a downward sloping demand schedule which is affected by random short-run demand shifts. The firm and the union bargain over wages prior to the revelation of actual demand, while the firm determines prices and employment after demand is realized. Hence, employment levels, prices and profits will be unknown when wage bargaining takes place.
The union is concerned with the welfare of the representative worker employed by the firm when bargaining takes place. If layoffs are made by random draw, the probability for a representative insider to stay with the firm decreases with the real product wage and the number of insiders. The welfare of the representative worker staying with the firm is assumed to depend on the consumer real wage as well as relative wages.(1) A laid-off insider will be unemployed for a period receiving unemployment benefits. Thereafter, she obtains another job and receives the expected alternative wage. It is assumed that finding another job increases her welfare, while higher unemployment will increase the expected time spent unemployed. Hence, the expected utility for an insider, conditional on becoming unemployed, increases with the expected alternative wage and the benefit level, and decreases with unemployment and the comparison wage.
Finally, we assume that the outcome from the wage bargain is given by the solution of an asymmetric Nash bargaining problem. We assume that strike is the industrial action that can be used. For simplicity, we further assume that the workers' utility during a strike depends on the outside income, while the fallback profits are zero. Given these assumptions we can derive the following log-linear approximation to the nominal wage equation
[Mathematical Expression Omitted]
where w[c.sub.it] is wage costs per hour, v[a.sub.it] value added per hour and nit employment.(2) The insider weight, [[Mu].sub.1], is the main parameter of interest, and is interpreted as the long-run elasticity with respect to industry prices and productivity. Higher expected employment relative to the number of insiders will reduce the probability for an insider being laid off and hence push up wages. This effect is represented by the insider hysteresis term, [Mathematical Expression Omitted]. The outside wage, w[a.sub.t], represents the effects of both the alternative and the comparison wage, and we make no attempt to identify partial comparison and alternative wage mechanisms. The replacement ratio is given by [b.sub.t], while [U.sub.t] is the aggregate unemployment rate. The final term represents a wedge effect, and contains consumer prices, p[c.sub.t], income taxes, t[i.sub.t] and the payroll tax rate, t[p.sub.t]. The static wage eq.(3) can be regarded as the long-run solution to the dynamic counterpart discussed below, and defines nominal industry wages as a convex combination of sector specific variables, outside labour market variables and a wedge term. Finally, it should be noted that the wedge drops out if the workers and the firm are risk neutral. In such a case, wage costs are unaffected by taxes and consumer prices, and higher taxes are entirely borne by labour.
4. Empirical specification
The study uses a panel of annual time series data from the National Account Statistics for 117 Norwegian industries for the time period 1966-87. These industries include mining, manufacturing, and part of private services. In all, data cover 79% of private sector employment. The wage variable is wage costs per hour worked. Since wage negotiation can be considered as a bargain over the employers' and workers' shares of factor income, the insider variable, va, is implemented empirically by factor income per hour worked. The outside wage is defined as the average wage in all other industries. This short-cut implies that all industries are considered as equally relevant to the workers in any one industry. An alternative interpretation is that the workers are mainly concerned with their position relative to the average wage level.
Higher unemployment reduces the bargained wage by increasing the expected cost of job loss. Blanchflower and Oswald (1990) and Nickell (1987) provide arguments in favour of a non-linear relationship between wages and unemployment, and results from time series analysis like Nymoen (1989), Stolen (1993), and Johansen (1995) imply that the Norwegian wage curve is convex. We assume an inverted square specification corresponding to the preferred equation in Johansen (1995). To test the robustness of the results obtained using aggregate time series data, we also estimate a log specification.
Expectation inertia and long-term contracts are arguments for complicated dynamics, and the results in Nymoen (1989) and Johansen (1995) indicate very sluggish adjustments of Norwegian wages. Since most centrally negotiated wage agreements have been struck for a 2-year period, it seems appropriate to include two lags for the explanatory variables in the dynamic model. The replacement ratio is excluded since Elgsaether and Johansen (1993) report a negatively signed and insignificant effect. The wage equation is expanded with changes in normal hours, [Delta]h, to control for compensation effects of reduced working time, and a binary variable, d79, for the wage-price freeze 1978-9. Finally, industry specific fixed effects, [[Alpha].sub.i], are included to control for wage differences due to time-invariant unobserved variables. Hence, the general dynamic model that acts as the starting point for further simplification goes as follows
[Mathematical Expression Omitted]
where [Epsilon][[center dot].sub.t] are the error terms which properties will be discussed below.
5. Dynamic modelling
Since the fixed effects model (4) includes lagged values of the left-hand side variable, OLS-estimates are biased even if the residuals are white noise. Because firm set prices and unemployment, v[a.sub.it] and [Delta][n.sub.it] will be treated as endogenously determined, while outside wages and the remaining aggregate variables are treated as strictly exogenous. To obtain consistent estimators we make use of the Generalized Method of Moment (GMM) proposed by Arellano and Bond (1991). The model is first-differenced to remove the individual fixed effects. In the absence of second-order serial correlation in the first-differenced residuals, w[c.sub.it - r], v[a.sub.it - r] and [Delta][n.sub.it - r] are valid instruments for r [greater than or equal to] 2. The set of instruments used in the estimation utilize the orthogonality restrictions between the differenced residuals and the second and third lag of industry wages and factor income per hour. Further instruments are employment changes lagged two and three years, the first and second lag of import prices, aggregate manufacturing prices and productivity.
5.1. Full sample estimates of the benchmark model
Table 2 reports one-step robust GMM results for the maintained model as well as parsimonious versions.(3) The Arellano and Bond (1991) [m.sub.2] statistics, testing the null of no second-order correlation in the residuals, are all below critical values. Moreover, both the Sargan (1958) test for instrumental validity and the Sargan difference test look comfortable. Since the latter tests the validity of the orthogonality restrictions between the differenced residuals and w[c.sub.it - 2], insignificant values are evidence against serial correlation.
From eq. I (Table 2) we see that the long-run wedge effect is negatively signed, and so is also the short-run effect of income taxes. The estimate of the insider hysteresis term, [Delta][n.sub.it], is below zero and highly insignificant. Since the maintained equation contains several insignificant estimates, we carry out a [TABULAR DATA FOR TABLE 2 OMITTED] simplification search. The simplified version given by eq. II parsimoniously encompasses the maintained model since a test of the validity of the six restrictions yields [[Chi].sup.2](6) = 7.907 which is well below critical value.(4) When eq. II is expanded with employment changes, the estimate is positively signed but still insignificant. Hence, we find little evidence in favour of insider hysteresis effects in the Norwegian wage formation process.
The estimated level of value added per hour worked, v[a.sub.i], is very sharply determined in all equations reported in Table 2. The long-run insider weight, [[Mu].sub.1], derived from the full sample estimates, is slightly below 0.2, and approximately five times as high as the corresponding estimate for Norway reported in Holmlund and Zetterberg (1991). Thus, the result in Holmlund and Zetterberg (1991) seems not to be robust, and our estimates indicate that insider forces are almost equally important in Norway as in economies with a more decentralized wage setting system.
Using data for Norwegian manufacturing firms, Wulfsberg (1993) reports a significant long-run insider weight close to 0.05. When we compare our results with other studies, we should take into account that some variables considered as insider variables in industry level wage equations are external variables in firm level wage equation. If firm level wages are affected by wages in other firms within the same industry, aggregation from firm to industry level will increase the insider weight. Hence, the difference between our results and those reported by Wulfsberg seems reasonable. The estimated insider weight reported in Table 2 is above firm level estimates for United Kingdom and Spain, cf. Nickell and Wadhwani (1990) and Bentolila and Dolado (1992), respectively. The effect of industry specific prices and productivity is higher than the industry level estimates for Germany reported in Holmlund and Zetterberg (1991), but below the average estimates for British industries in Nickell and Kong (1992) and Lee and Pesaran (1993).
The presence of a permanent relationship between industry pay and industry performance is evidence against competitive forces as well as completely centralized wage setting. The finding that industry pay is affected by industry specific performance reflects a form for rent sharing that is consistent with theories of firm level bargaining. However, we should have in mind that we are unable to control for changes in the composition of skill and human capital which may induce a positive correlation between industry wages and productivity. But it is hard to believe that such bias should be more serious in our study than in similar studies for other countries.
Although insider effects in wage determination basically reflects non-competitive forces, the theoretical foundations for such mechanisms need not be restricted to union model. First, the basic point made by Lindbeck and Snower (1988) is that labour turnover costs create market power which incumbent workers may exploit, or that the firms are willing to pay rent to the insiders in order to avoid hiring, firing, and training costs. Second, the dynamic monopsony or turnover model in Layard et al. (1991, p. 184-5) can generate a wage equation that is almost identical with the one derived from the insider-outsider model in Section 3. Third, Agell and Lundborg (1991, 1992) argue that workers' effort depends positively on their wages relative to wages outside the firm, but also on the functional distribution of income within the firm. Hence, efficiency wage setting implies a positive relationship between industry wages, outside income and industry specific profitability. We will not make any attempt to discriminate between these alternative theories of rent-sharing, but turn to the quantitative importance of the results.
In order to illustrate the importance of the insider effect we calculate mean and maximum values of factor income per hour worked in 1987.(5) The estimated insider weight implies that the effect arising from a hypothetical change in factor income per hour worked from the sample mean to its maximum would increase wages by 67%. This prediction is not entirely out of range as the actual maximum wage level is 47% above the mean value.
The short-run insider effect is also well determined, but very small. Since the estimated partial adjustment coefficient is also low, these results imply very sluggish wage adjustment to price and productivity changes. At this point our results diverge from those obtained in Nickell and Wadhwani (1990) who report short-run estimates well above the long-run insider weight. When interpreting these results we should take into account that product prices have fluctuated strongly for the Norwegian export oriented industries. Hence, if the short-run response is high, these industries would experience very strong nominal wage fluctuations.
The dominating determinant of industry wages is the outside wage. Both the short- and the long-run elasticities are above 0.8 in all versions of the model. The estimates of outside wages may reflect a mixture of alternative wage and comparison wage mechanisms. Within the theoretical bargaining model, alternative wages, unemployment and the replacement ratio enter the wage equation symmetrically, as all these variables affect the expected cost of job loss. Since Elgsaether and Johansen (1993) reports a negatively signed, but insignificant estimate of the replacement ratio, the outside wage effect most likely reflects a comparison wage mechanism. Outside wages may also affect industry wages through efficiency wage mechanisms. Firms may use wages as an instrument to recruit, retain, and motivate their workers, and the ability to recruit, retain, and motivate their workers depends on the wage being paid relatively to the expected outside wage, possibly corrected for employment opportunities. Although wages are formally set by bargaining, Hoel (1989) and Rodseth (1993) show that efficiency wage mechanisms may still be at work.
Higher unemployment reduces industry wages. Since the error correction terms enter significantly, the estimated wage equations imply a long-run relationship between the wage level and the unemployment rate. Hence, the Phillips curve specification is rejected. The estimates of current values and the first lag of the unemployment rate are correctly signed and well determined. Unemployment lagged twice enters with the opposite sign indicating a kind of outsider hysteresis effect which is in accordance with the findings in Johansen (1994). Such outsider ineffectiveness may partly reflect composition effects since higher layoffs immediately increase the pool of short-term unemployment, while the proportion of long-term unemployment increases over time.
The estimated unemployment coefficient, given outside wages, is very small. By making use of the estimates for eq. II we find that increased unemployment ratio from 1-2% implies a long-run wage reduction by 1.3%. But higher unemployment will also reduce the outside wage, and hence affect industry wages indirectly. When symmetric wage spillover effects are taken into account, the estimated long-run unemployment coefficient becomes rather large. To see this we set w[c.sub.i] = wa = wc and solve out for average wages to obtain
wc = va + [0.08U.sup.-2] + const (5)
which implies that a permanent unemployment increase from 1-2% will reduce wages by 6%. Since the estimated wage curve is strongly non-linear, the wage dampening effect of increased unemployment will be low for moderate levels of unemployment. A permanent unemployment increase from 3-4% will only reduce wages by 0.4% even when symmetric spillover effects are taken into account. It is now meaningful to compare the long-run estimates given by eq. (5) with results from aggregate wage equations. Interestingly, the preferred wage model reported in Johansen (1995), using aggregate time series data for the Norwegian manufacturing industry, implies the following long-run wage curve
wc = ya + [0.10U.sup.-2] + const (6)
where we note a striking similarity with eq. (5).
The log specification given by eq. III implies a less convex wage curve. In symmetric equilibrium, increased unemployment ratio from 1-2% will reduce wages by 4% while an increase from 3-4% will reduce wages by approximately 1.7%. However, the results in Table 2 are evidence in favour of strong convexity since eq. II encompasses the log specification. Hence, our results based on panel data confirm previous findings using aggregate time series data. The re-specification of the wage curve does not affect the estimated long-run insider weight, while the short-run estimate and the speed of adjustment parameter increased slightly. Finally, when eq. III is expanded with employment changes the insider hysteresis effect is still insignificant.
The results imply no long-run real wage resistance, and the short-run effect of consumer prices is very small. Higher payroll taxes increases wage costs significantly in the short-run, while in the long-run, payroll taxes are entirely borne by labour. Finally, reduced normal hours induce an immediate positive compensation effect on industry wages. Hence, increased consumer prices and payroll taxes, and reduced normal hours will push actual wage growth above the steady state path. In the long-run, deviations from the equilibrium path, determined by outside wages and own profitability, are eliminated through an error correction mechanism.
TABLE 3 Wage Equations with Time Dummies.
GM M estimates (all variables in first differences). Dependent variable is: [Delta]w[c.sub.i]
1966-87 1972-87 1966-80 V VI VII
[Delta]v[a.sub.it] 0.0339(2.45) 0.039(2.72) 0.025(1.60) w[c.sub.it-1] -0.156(3.57) -0.174(2.49) -0.194(3.27) v[a.sub.it-1] 0.036(4.71) 0.044(4.12) 0.032(2.95) Insider weight 0.234 0.252 0.163
Sargan 102.81(85) 80.43(61) 82.38(57) Sargan diff 22.27(22) 21.28(16) 32.90(15) [m.sub.2] -0.951 -1.49 -0.47
Notes: See notes to Table 2.
5.2. Subsample estimates
Institutional evidence reported in Holden (1989) and Rodseth and Holden (1990) indicate that firm level bargaining have become more important during our sample period as the share of wage drift increases. If centralization really matters, we should expect the effect of insider variables to increase over time. Interestingly, the hypothesis that more decentralized wage setting implies increased scope for rent sharing is supported by the results in Edin and Holmlund (1993).
The sub-sample estimates reported for eq. IV imply a higher long-run insider weight as compared to the estimates obtained using the full sample period. Hence, our results support the hypothesis that more decentralized wage setting will increase the importance of internal variables. Since consumer prices are still excluded, and homogeneity is imposed to the wage models, the long-run effect of the outside wage must decrease. The results also reveal an increased short-run effect of value added per hours worked, while the impact effect of outside wages are below the corresponding full sample estimates
In order to investigate the robustness of the results obtained until now, we estimate the wage equation with all the aggregate variables replaced with time dummies. Results for the full-sample as well as sub-sample estimation are presented in Table 3. We first see that replacing the aggregate variables with time dummies increases the full-sample estimate of the long-run insider weight. Second, both the short- and the long-run estimates of the insider variable are significantly determined for all sample periods. Third, the estimated insider weight based on data for the sub-sample 1972-87 is 50% above the estimate obtained using data for the sub-sample 1966-80. Hence, the alternative approach using time dummies confirms the finding that insider forces matter, and in particular, that the importance of insider forces has increased during our sample period.
5.3. Tests for asymmetric insider effects
Insider effects may be more important in expanding industries as compared to declining ones, and similarly, industry wages may be more responsive to insider variables in good times than in bad times. Such asymmetric insider effects imply downward wage rigidity, and will tend to put more pressure on employment when times are bad. Empirical evidence in favour of asymmetric insider effects in British manufacturing firms can be found in Nickell and Wadhwani (1990) who perform the test by interacting a dummy variable for expected demand with the insider variable. A similar approach is taken by Brunello and Wadhwani (1989) who report more mixed evidence both for Japanese and British firms. Evidence based on a series of surveys reported in Blanchflower (1991, p.492) indicate some asymmetries and downward wage rigidity in the UK as 'workers in expanding plants receive a pay premium; those in contracting plants suffer no pay disadvantage'. The empirical approach taken in the present paper is in particular inspired by the findings by Blanchflower that workers who expected employment to grow received pay premia. If this is also the case in Norwegian industries, the insider weight should be larger in industries and time periods where workers expect (or are experiencing) employment growth. To test this hypothesis we interact lagged employmant changes with the basic variables that enters the benchmark model. Since employment growth is endogenous, all interaction terms are treated as endogenously determined.
When we interact lagged employment changes with the short-run variables and unemployment, these estimates are insignificantly determined and thus excluded. However, as can be seen from Table 4, the estimates of the two error correction terms depend strongly on past employment changes. The estimate of lagged profitability increases significantly in good times, while the response to lagged outside wages decreases even more. Taken together, these results imply more sluggish wage response when times are good than in bad times. This may reflect the sluggish wage responsiveness to prices discussed above, and the more weight is put on the insider variables the more sluggish will be the wage adjustments.
The estimates reported in eq. VIII imply that the long-run insider weight for industries experiencing a 10% employment growth is approximately 75% above the insider weight for industries experiencing an employment reduction of 10%. Finally, we estimate the asymmetric insider model using log unemployment instead of the inverted square. The main results are unaffected by such a re-specification, but the estimated insider weight shows somewhat less variability as compared to the results reported for eq. VIII.
[TABULAR DATA FOR TABLE 4 OMITTED]
6. Concluding comments
Norway is usually considered as an economy with a highly centralized wage setting system. Complete centralization implies that wages are entirely determined by national bargaining, there is little scope for firm or sector specific rent sharing, and we should expect industry performance to be of minor importance in shaping industry pay. The hypothesis that Norwegian industry wages are independent of industry specific performance is, at least partly, contradicted by the results reported in the present paper, which studies wage determination in Norway using data for 117 industries from 1966 to 1987.
The main empirical findings can be summarized as follows. Industry wages are significantly affected by own prices and productivity. The full sample estimates imply a long-run insider weight close to 0.2 which is comparable with estimates reported for economies with a more decentralized wage setting system. The finding that industry wages are significantly affected by industry specific performance is evidence in favour of a hypothesis that local wage bargaining implies some scope for industry or firm specific rent sharing. However, one should be aware that the insider weight may be overstated because of the absence of any control for the skill mix of the workforce. On the other hand, it is hard to believe that the effects of changes in skill composition should be more important for the results in our study than in similar studies for other countries.
The long-run insider weight seems to increase over time. Since the share of wage drift has increased during our sample period, this finding is evidence in favour of a hypothesis that more decentralized wage setting makes insider forces in wage determination more important.
These findings imply a kind of wage flexibility in response to industry specific performance which may be considered as a desirable property of the wage formation process. However, a high insider weight means, ceteris paribus, low responsiveness to outside labour market conditions. Moreover, the paper provides evidence in favour of asymmetric insider effects as wages seem to be most responsive to insider variables in good times and in expanding industries. Such asymmetric insider effects will tend to put more pressure on employment when times are bad.
Industry wages are affected by unemployment. The effect of increased unemployment, given outside wages is small, but when symmetric spillover effects are taken into account, the estimated unemployment coefficient is close to estimates reported in aggregate time series analyses. Our results also support the findings in previous time series studies that the Norwegian wage curve is strongly non-linear and becomes flat for moderate levels of unemployment. On the other hand we found no support for insider hysteresis effects since the estimates of employment changes were insignificantly determined and often incorrectly signed.
Not surprisingly, the outside wage is the dominating determinant of industry wages both in the short- and in the long-run, and favourable industry performance that increase industry wages spills over into higher wages in the rest of the economy. Finally, consumer prices, payroll taxes and normal working time affect wages in the short and medium run, but there is no long-run real wage resistance. Hence, the main long-run determinants of industry wages are the outside wage and the industries' ability to pay.
Valuable comments from Jan Morten Dyrstad, Eilev S. Jansen, seminar participants in Bergen and Uppsala, two anonymous referees, and the editor are gratefully acknowledged. However, the usual disclaimer applies.
1 See Akerlof (1982), Akerlof and Yellen (1990) and Solow (1990) for arguments in favour of the role of social norms and fairness in wage determination.
2 Lower case letters denote log transformed variables except for the tax rates. Expected magnitudes are indicated by superscript e.
3 The GMM estimator is implemented in DPD written in GAUSS, cf Arellano and Bond (1988). We report one-step results since the standard errors generated by the two-step procedure are too low. The two-step results are available from the author upon request.
4 Significant value of [Mathematical Expression Omitted] in eq. III is due to a significant negative estimate of [Delta][t.sub.i]. When [Delta][t.sub.i] is excluded, the remaining restrictions can easily be imposed.
5 These values are 155, 210 and 7, 117, 410 Norwegian kroner, respectively.
AGELL, J. and LUNDBORG, P. (1991). 'Wage Fairness and International Trade Theory and Policy', Seminar Paper 486, Institute for International Economic Studies, Stockholm University.
AGELL, J. and LUNDBORG, P. (1992). 'Fair Wages, Involuntary Unemployment and Tax Policies in the Simple General Equilibrium Model', Journal of Public Economics, 47, 299-320.
AKERLOF, G. A. (1982). 'Labour Contracts as Partial Gift Exchange', Quarterly Journal of Economics, 97, 543-69.
AKERLOF, G. and YELLEN, J. L. (1990). 'The Fair Wage-Effort Hypothesis and Unemployment', Quarterly Journal of Economics, 105, 255-83.
ARELLANO, M. and BOND, S. (1988). 'Dynamic Panel Data Estimation Using DPD - A Guide for Users', Working Paper 88/15, Institute for Fiscal Studies, London.
ARELLANO, M. and BOND, S. (1991). 'Some Tests of Specification for Panel Data: Monte-Carlo Evidence and an Application to Employment Equations', Review of Economic Studies, 58, 277-97.
BLANCHFLOWER, D. G. (1991). 'Fear, Unemployment and Pay Flexibility', Economic Journal, 101, 383-496.
BLANCHFLOWER, D. G. and OSWALD, A. J. (1990). 'The Wage Curve', Scandinavian Journal of Economics, 92, 215-35.
BENTOLILA, S. and DOLADO, J. J. (1992). 'Who are the Insiders? Wage Setting in Spanish Manufacturing Firms', Paper prepared for the European Unemployment Programme Conference, Chelwood Gate, September.
BRUNELLO, G. and WADHWANI, S. (1989). 'The Determinants of Wage Flexibility in Japan: Some Lessons from a Comparison with the UK using Micro-Data', Working Paper No. 1116, Centre for Labour Economics, LSE.
EDIN, P.-A. and HOLMLUND, B. (1993). 'The Swedish Wage Structure: The Rise and Fall of Solidarity Wage Policy?', Working Paper No. 4257, NBER, Cambridge, MA.
ELGSAETHER, A. H. and JOHANSEN, K. (1993). 'Hvor var insiderne? En analyse av lonnsdanningen i norsk industri', Norsk Okonomisk Tidsskrift, 107, 255-76.
FORSLUND, A. (1994). 'Wage Setting at Firm Level - Insider versus Outsider Forces', Oxford Economic Papers, 46, 245-61.
HOEL, M. (1989). 'Efficiency Wages and Local versus Central Wage Bargaining', Economic Letters, 30, 175-9.
HERSDUG, T., KJAER, K. N., and RODSETH, A. (1984). 'Wages, taxes and the utility-maximising trade union: a confrontation with Norwegian data', Memorandum No.12, Department of Economics, University of Oslo.
HOLDEN, S. (1989), 'Wage Drift and Bargaining: Evidence from Norway', Economica, 56, 419-32.
HOLMLUND, B. and ZETTERBERG, J. (1991). 'Insider effects in wage determination: Evidence from five countries', European Economic Review, 35, 1009-34.
JOHANSEN, K. (1994). 'Real Wage Responsiveness, Long-term Unemployment and Duration Effects. Norwegian Evidence 1964-1990', mimeo, Department of Economics, University of Trondheim.
JOHANSEN, K. (1995). 'Norwegian Wage Curves', Oxford Bulletin of Economics and Statistics, 57, 229-47.
LAYARD, R., NICKELL, S., and JACKMAN, R. (1991). Unemployment. Macroeconomic' Performance and the Labour Market, Oxford University Press, Oxford.
LEE, K. C. and PESARAN, M. H. (1993). 'The Role of Sectoral Interaction in Wage Determination in the UK Economy', The Economic Journal, 103, 21-55.
LEVIN, A. and LIN, C.-F. (1992). 'Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties', Discussion Paper 92-23, Department of Economics, University of California, San Diego.
LINDBECK, A. and SNOWER, D. J. (1988). The Insider-Outsider Theory of Employment and Unemployment, MIT Press, Cambridge, MA.
NICKELL, S. J. (1987). 'Why is Wage Inflation in Britain so High?', Oxford Bulletin of Economics and Statistics, 49, 103-28.
NICKELL, S. J. and KONG, P. (1992). 'An Investigation into the Power of Insiders in Wage Determination', European Economic Review, 36, 1573-99.
NICKELL, S. J. and WADHWANI, S. (1990). 'Insider Forces and Wage Determination', Economic Journal, 100, 496-509.
NYMOEN, R. (1989), 'Modelling Wages in the Small Open Economy. An Error correction Model of Norwegian Manufacturing Wages', Oxford Bulletin of Economics and Statistics, 51, 239-58.
RODSETH, A. (1993). 'Efficiency Wages and Local versus Central Bargaining', Oxford Economic Papers, 45, 470-81.
RODSETH, A. and HOLDEN, S. (1990). 'Wage Formation in Norway', in L. Calmfors (ed), Wage Formation and Macroeconomic Policy in the Nordic Countires, Oxford University Press, Oxford.
SARGAN, J. D. (1958). 'The Estimation of Economic Relationships using Instrumental Variables', Econometrica, 26, 191-415.
SOLOW, R. (1990). The Labour Market as a Social Institution, Basil Blackwell, Oxford.
STOLEN, N. M. (1993). Wage Formation and the Macroeconomic Functioning of the Norwegian Labour Market, Ph.D. thesis, University of Oslo.
WULFSBER, F. (1993). 'An Application of Wage Bargaining Models to Norwegian Panel Data', mimeo, Department of Economics, University of Oslo.
Data and sources
W[c.sub.i]: Wage costs per manhour industry i Source: National Account Statistics (NA), Central Bureau of Statistics (CBS)
Wa: Average wage costs per manhour outside industry i Source: NA, CBS
V[a.sub.i]: Factor income per hours worked in industry i Source: NA, CBS
[N.sub.i]: The number of employees in industry i Source: NA, CBS
Pc: The official consumer proce index. Source: Weekly Bulletin of Statistics, CBS
U Aggregate unemployment ratio. Computed as the number of unemployed workers registered by the government employment offices divided by the total number of man-years plus the number of unemployment.
Source: Hersoug, Kjaer and Rodseth (1984), Labour Market Statistics, CBS, and NA, CBS
H: Normal working hours per week in manufacturing. Source: Rodseth and Holden (1990)
Q: Aggregate import price index, manufacturing goods Source: NA, CBS
P: Implicit factor income deflator for total manufacturing Source: NA, CBS
Z: Average labour productivity, total manufacturing Source: NA, CBS
tp: The average payroll tax rate Source: NA, CBS
ti: The average income tax rate
|Printer friendly Cite/link Email Feedback|
|Publication:||Oxford Economic Papers|
|Date:||Jan 1, 1996|
|Previous Article:||Trade unions, efficiency wages, and shirking.|
|Next Article:||The system-wide impact on the recipient region of a regional labour subsidy.|