The Impact of Labour Market Partial Reforms on Workers Productivity: The Italian Case

Transcription

1 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 The Impact o Labour Market Partial Reorms on Workers Productivity: The Italian Case Andrea Beccarini * Whilems Universität Münster (Germany) & LUISS (Italy) and University o L aquila (Italy) Abstract According to a widespread point o view, lexibility in the labour market is considered a means to ight unemployment and to increase labour productivity. But i lexibility is introduced into the system at the margin (only or new contracts) then its eect may be ambiguous and even perverse as Blanchard and Landier (2002) argued. This work ollows this strand and demonstrates that partial reorms do not necessarily stimulate diligence o workers. I there is a large amount o protected workers, lexible workers may perceive that diligence does not worth. Estimates with Italian data are presented and conirm this theoretical argumentation. Keywords: Partial labour market reorms, eiciency wage, threshold model JEL Classiication: J24, J4, K3. Introduction Many European countries have undertaken labour market reorms in the past decades. These reorms aim at increasing lexibility in the labour market; however, they generally operate at the margin in the sense that they regard new contracts (or an overview o these reorms see OECD, 999, or European Commission, 2005). For example, the Italian labour market has experienced two important reorms in 997 (the so called Legge Treu) and in 2003 (the so called Legge Biagi). By these reorms the labour market has become more lexible, giving the single employers and the employees the possibility o bargaining almost all eatures o the new job s contracts, in particular, the wage and the duration. The share o new contracts which exploits the new laws was ast increasing. So, many researches have been devoted to ind the reason why irms want temporary contracts: Bronstein (99) inds that European employers have three main reasons why they want a temporary job: temporary substitutions o permanent workers, unexpected demand luctuations, screening or permanent positions. According to the analysis o Portugal and Varejao (2003) screening is the single most important reason why irms use this type o contracts. So, i temporary works are set or lexibility reasons (substitutions and demand luctuations but not or screening) then these works do not provide access to permanent jobs meanwhile only screening reasons imply a conversion into permanent jobs. Atkinson et all (996), Abraham (988) and Houseman (2000) provide statistics about the reason why irms declare to preer a temporary position. The reasons o screening are 20%, 6% and 43% o the total, respectively. For the Italian case, Ichino et all (2005) estimate that the probability o inding a permanent job increases o 4% ater a temporary job. However, asymmetric inormation problems may arise, in the sense that workers do not necessarily know the true reason o the choice o a

2 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 2 temporary contract, hence they may perceive the requency o the screening reason higher than the one declared by irms. 2 Firms may cheat about the true reason o the vacancy because workers, during the screening job s period may be stimulated to behave eiciently, (see Shapiro-Stiglitz, 984): since jobs are temporary, workers know that the possibility o a job s contract renewal (or conversion to a permanent job) depends on their diligence. 3 However, workers may anticipate this behaviour, hence they must orm a probability o a renewal which depends on several actors that this work aims to investigate. So, on the one side, these reorms, removing rigidity in the market, should increase labour productivity o lexible workers. Nickell (2003) shows the beneits in terms o employment rom removing such kind o rigidities, in act, these workers are not necessarily unionised, 4 hence they cannot get mark-ups by the monopoly bargain and duration contracts can be also bargained, or all these acts, the market clears. However, Bertola (990) argued that the long-run employment is not aected by the degree o lexibility in the labour market meanwhile lexibility may increase the variability o employment and eiciency o irms. On the other side, many research works cast many doubts on their positive eect on eiciency or on welare. Blanchard and Landier (2002) highlighted the perverse eects o such a partial reorms since their main eect would be an increase in the turnover o ixed contracts and indicate the true reason why o such a kind o reorms on the opportunism o the politicians. Boeri and Garibaldi (2007) ollowed this strand and argued that these reorms cause a all in the productivity o labour; Bentolila and Dolado (994) also argued that the main eect o these reorms is to create a duality in the labour market. In this study, the same conclusions are reached but rom a dierent perspective: dierently rom Boeri and Garibaldi (2007) who consider productivity as a stochastic exogenous variable, here labour productivity is endogenous and it is determined by the optimising behaviour o workers; more precisely one may veriy whether the reorms in question have indeed stimulate labour productivity as the Eiciency Wage theory predicts. In this paper, it is argued that during recessions, irms must evaluate to dismiss temporary contracts independently rom what reason why irms wanted temporary contracts beore the recession; consequently, it is argued that temporary workers may or may not be stimulated by the temporary eature o the contract. In act, even i the possibility o a renewal o the contract may work as diligence stimulus, it seems also plausible that the chance o having again a job depends on the structure o job s contracts the irms owns other than the business cycles: a recession, together with a large percentage o the permanent workers with respect to the all workers may induce temporary workers not to increase their productivity because the recession will hit them independently o their productivity since (one may assume) the permanent (unionised) workers cannot be ired in the short run; moreover, marginal productivity o temporary workers also depend on the total number o workers which includes the number o unionised workers. In other words, when recessions occur, irms may desire to dismiss some workers; those that will be dismissed would be the most unproductive unless they are unionised hence protected 5 in the short run; i the share o these workers is high on the total personnel, then only some temporary workers will suer the recession. So, the probability o a renewal o the contract largely depends on the business cycles and on the ratio between the number o temporary contracts and the total number o job s contracts. 2. How Much Should the Optimising Temporary Worker Be Diligent?

3 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 3 Consider the proit unction o a representative irm where p is the product price, y is the output, b is the business cycle parameter, is the number o temporary workers, l is the number o unionised workers, w and w are their wages respectively; α and β are parameters capturing the eort o l temporary and unionised workers respectively. 6 Π = py w w l () l with 7 y = Y (, l; α, β ) and p = P[ Y (, l; α, β ); b]. In the short run, irms can only choose the number meanwhile l remains ixed. So, the irms maximises their proit by choosing appropriately: Π = P Y Y + PY y w = 0; (2) Y P Y + P) = w. (3) ( y Speciying a production unction it can be possible, in principle, to solve the last equation or. For example, i: 8 y = Y (, l; α, β ) = Y ( α + βl) (4) then: Y Y ( α + βl) = α with q = α + βl; (5) q rom eq. (3): Y Y ( α + βl) w = α = ; (6) q ( P Y + P) y w α + βl = V (7) α( P yy + P) Y (.) where V[.] is the inverse unction o. q w β = V l α α( P yy + P) α Now deine n as the number o workers in the labour orce; hence n and the rate o employment when a temporary job is n. Suppose that n can be interpreted as the probability o having a temporary job or the n people, next time. Considering eq. (8) this probability is now: (8) n w βl = V α( P yy + P) (9)

4 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 4 Assuming a competitive market or the irms ( P = 0) the above equation can be rearranged as: y n w βl = V (0) αp The above probability is generally a non-monotonic unction o P, l, and α ; or Y(.) hence V(.) having the usual properties it positively depends on P and α and negatively on l. For example, i y = Y ( α + βl) = log( α + βl + ) then it becomes: n w = αp βl + = P w βl + = n P βl ; () w n The above equation implies that during recessions (b is low) P is low, hence workers may increase their eort α in order to compensate the adverse product market condition; the same situation occurs when l is high. However increasing α has a cost (sacriice); hence workers choose it so as to balance its marginal cost with its marginal beneit. Suppose that workers maximise the ollowing utility unction: Argmax U c, α ) + δe[ U ( )] (2) c 0,α ( 0 c where U(.) is concave with respect to c0 and convex with respect to α and: 9 high low E[ U ( c)] = U ( c ) + U ( c ) (3) n n From eq. () speciy / n = pr(. The irst order conditions or α imply: U ( c0, pr( + δ U ( c higk pr( ) U ( c low ) = 0; (4) or U ( c, pr( = higk { δ [ U ( c ) U ( c )]} 0 low (5) Eq. (5) states that the level o diligence α should be chosen such that the marginal decrease o utility in period 0, should equal the discounted increase o the utility (in period ) due to the contract s renewal (i. e. when c is high) times the increase o the probability / n = pr( o the contract s renewal. Holding the second actor o the right-hand side o eq. (5), the eort the worker is willing to make depends on its impact on the probability o the contract s renewal.

5 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 5 pr( Since pr ( α ) = / n is not a monotonic unction o α, l and P then may be also small or some values o P and l. It implies that the value α can be also small. For example, i the irm hiring lexible workers experiences large losses (P is very low) then increasing α does not suiciently increase the probability o having next period a job; hence maximising utility or workers means choosing a low level o α, that is behaving unproductive. More importantly, the impact o α also depends on l, the number o unionised workers. Larger is l lower is this impact, hence lower is the choice o α. Intuitively, marginal productivity o lexible workers also depends on the total number o workers in the irm because output is a concave unction o the sum o workers, having weighted each worker by his/her speciic productivity α, β ; hence higher is l lower must be the productivity o lexible workers who, at the same time, are on the margin in the sense that the marginal adjustments o irms in order to equal the marginal productivity o labour to the real wage are made by hiring or iring the lexible workers. 3. Evidence Suppose in the economy there are a large number o identical representative irms so that aggregate values can be considered. The empirical evidence can be ound in two manners. Firstly, one can veriy whether eq. () is consistent with the considered data, that is whether the rate o employment o lexible workers depends on the right-hand side variables o eq. (). Secondly, one can veriy whether, when the ratio between the number o unionised workers and the labour orce increases, the eort o lexible workers (measured by α ) decreases, as predicted by eq. (5). Eq. () can be estimated by regressing the rate o lexible employment /n on a constant, on the real wage divided by labour orce, on the rate o unionised employment and on the inverse o the employment level. The relevant coeicient is the one associated to the rate o unionised employment and it measures the ratio between β and α. To veriy the relative change on α with respect to β caused by the increase o the rate o unionised employment, a Threshold model can be used, see Granger and Terasvirta (993). By means o a Threshold model (TM) the sample is split in two clusters, the irst one including the data when l/n is below a given threshold and the second one when l/n is above the threshold. Given these two subsamples, it is possible to estimate two values o the ratio between β and α on or each cluster. The threshold value is ound by minimising the sum o the two residual sums o squared, one or each subsample. So, the sub-samples are, respectively, characterized by lower and higher values o l/n, with respect to the threshold level. The threshold rate can be ranged between the minimum and the maximum observed l/n. 0 Quarterly data or Italy are considered, ranging between 996: and 2006:3; they are provided by Eurostat. Call tempr the ratio between the number o ixed-term contracts stock on the number o total contracts stock. The percentage o lexible workers on employment (/n) has been obtained by multiplying tempr times the employment rate. The percentage o permanent workers on employment (l/n) has been obtained by multiplying one minus tempr times the employment rate. 2 Also, deine the series rcl as the real labour cost times labour orce. According to eq. () the relevant variables are: - the rate o lexible workers on the labour orce, /n; - the rate o permanent workers on the labour orce, l/n;

6 Beccarini, International Journal o Applied Economics, 6(2), September 2009, the inverse o the product between the average real wage o lexible workers times the labour orce, /rcl; - the inverse o the labour orce, /l. Beore implementing the above mentioned analysis, the stationarity properties o the series must be veriied. By a graphical inspection, the series /n, /rcl and /l are suspected either to have a deterministic or a stochastic trend over time. Hence the null hypothesis o one unit root is compared to the alternative o deterministic trend. Table shows the ADF tests 3 or each series with the relevant alternative hypothesis and the decision met. According to the table results the series /n, /rcl, are stationary ater one dierentiation. /l is stationary ater having removed a deterministic trend and the series l/n can be considered also stationary. Hence the relevant regression is shown in table 2. The regression results do not show any signs o misspeciication. 4 Coeicients have the expected sign and the one o l/n is signiicant at the usual levels o signiicance. It is important now to veriy whether this ratio changes i l/n changes. As already said, by means o the threshold model (TM), one can veriy whether the coeicient o l/n estimating the ratio between β and α changes; more particularly whether it increases when l/n is high implying a decrease in α (when l/n) as the theoretical part predicts. The threshold model splits the sample in two clusters. The sample o 42 observations is split in one with 24 observations such that or all o them l/n is smaller than the threshold value and the other sample, with 8 observations such that l/n is larger than the threshold value. 5 The coeicient 6 o the irst sub-sample is 0.04 (0.26); or the second one it is 0.85 (0.25). Since in absolute value, 0.85 is larger than 0.32 it means that α has decreased with respect to β when the value o l/n increases, that is what the theoretical model predicts. 4. Conclusions Partial labour market reorms that are reorms at the margin have been implemented throughout in Europe despite many criticisms like Blanchard and Landier (2002) or Boeri and Garibaldi (2007). This work ollows this strand and it is argued that labour productivity o workers does not necessarily increase with temporary jobs i temporary workers cannot aect the probability o a contract renewal. This is the case when the irm business cycle is low and when there is a relatively large number o protected workers aside the temporary ones. In acts, during recessions irms may desire to reduce the personnel, however they cannot dismiss some permanent workers without some iring costs; so, the separation o temporary workers may be realistic even i their productivity is higher than permanent workers. This is exactly the situation occurred through the partial reorms. Evidence or the Italian case is provided: the probability o contract renewal negatively depends on the share o protected workers on the labour orce and the labour productivity o temporary workers decreases when this share increases. Endnotes

7 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 7 * address: pbec@tin.it.. For Italy, rom these reorms on, the unemployment rate has been alling but so the labour productivity and, consequently, the GDP growth seems to be unaected. So, these reorms making the labour market competitive, have dramatically reduced the voluntary unemployment but their eect on labour productivity is not so apparent as it has been alling, in the considered period. 2. There is an adverse selection problem that is irms may promise a renewal i the workers behave diligently also when jobs are demanded or other reasons. 3. In principle, temporary jobs may be rolled over orever. 4. Nor they have unemployment beneits. 5. By large iring costs. 6. Lower case letters indicate variables and upper case letters indicate unctions. 7. The production unction Y(.) and the demand curve P(.) have the usual properties. 8. O course the unction Y(.) is now partly speciied. 9. The utility unction U(.) has also the other usual properties. E[.] is the expected operator based on the probability /n. 0. More particularly, or each sample observation, i the value o l/n is below the speciied threshold value, then the observation belongs, say, to the irst sub-sample; i it is above the threshold values, the observation belongs to the other sub-sample. A routine is constructed so as to speciy all possible threshold values o l/n and to estimate the relevant parameters. The minimum value o the threshold l/n equals the lowest value o the (entire) sample plus a positive constant k (otherwise the sub-sample with values lower than the threshold would collapse); the maximum value equals the highest value o the sample less k. A step is speciied in order to obtain the array o all possible threshold values; or each o these values the TM is estimated and parameters values stored.. Both in decimal terms. 2. Both in decimal terms. 3. The variables lags are selected according to the Schwartz-Bayesian criterion. 4. Usual tests are perormed such as: serial correlation LM test, White Heteroschedasticity test, RESET test, Normality test, CUSUM and CUSUMQ tests. 5. The threshold value o l/n minimising the sum o squared is The mean o observations o the irst sub-sample is 0.8, o the second sub-sample is Standard errors are put in brackets.

8 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 8 Reerences Abraham, K. G Flexible Staing Arrangements end Employers Short-term Adjustment Strategies, in Hart R.A. (ed), Employment, Unemployment and Labour Utilization, Unwin Hyman. Atkinson, J., J. Rick, S. Morris, and M. Williams Temporary Work and the Labour Market, Report N.3, The Institute or Employment Studies, Brighton. Bentolila, S. and J. Dolado Labour Flexibility and Wages: Lessons rom Spain, Economic Policy, 8, Bertola, G Job Security Employment and Wages, European Economic Review, 34, Blanchard, O. and A. Landier The Perverse Eects o Partial Labour Market Reorm: Fixed- Term Contracts in France, The Economic Journal, 2, Boeri, T. and P. Garibaldi Two Tier Reorms o Employment Protection Legislation. A Honeymoon Eects, The Economic Journal, 52, Bronstein, A. S. 99. Temporary Work in Western Europe: Threat or Complement to Permanent Employment? International Labour Review, 3, European Commission Tracking labour Market Reorms in the EU Member States: An Overview o Reorms in 2004 based on the LABREF Database, Economic Papers no Granger, C. W. J. and T. Terasvirta Modelling Non-linear Economic Relationships. Oxord University Press, Oxord. Houseman, S. N Why Employers Use Flexible Staing Arrangements: Evidence rom an Establishment Survey, Upjohn Institute Working Paper N.0-67 Ichino, A., F. Mealli, and T. Nannicini Temporary Work Agencies in Italy: A Springboard Toward Permanent Employment? Giornale degli Economisti e Annali di Economia, -27. Nickell, S. J Labour Market Institutions and Unemployment in OECD Countries, in DICE- REPORT, Journal or Institutional Comparisons, 2, OECD. (999) Employment Protection and Labour Market Perormance, Employment Outlook, Paris. Portugal, P. and J. Varejao Why do Firms Use Fixed-Terms Contracts? Banco de Portugal, WP Shapiro, C. and J. E. Stiglitz Equilibrium Unemployment as a Worker Discipline Device, American Economic Review, 74,

9 Beccarini, International Journal o Applied Economics, 6(2), September 2009, -9 9 Table. ADF Tests Variable augmented Dickey-Fuller test alternative hypothesis decision /n deterministic trend H0 /rcl deterministic trend H0 /l deterministic trend H l/n stationarity H Table 2. Regression Results Dependent Variable: d(/n) Sample(adjusted): 996:2 2006:3 Included observations: 42 ater adjusting endpoints Variable Coeicient Std. Error t-statistic Prob. c (/rcl)-(/rcl(-)) (l) l/n R-squared 0.44 Prob(F-statistic) Adjusted R-squared 0.40 Durbin-Watson stat.85