31E00700 Labor Economics: Lecture 4 - Aalto-econ.fi

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Marshall's Rules of Derived DemandLong-run elasticity of labor demand ELR ELR /ELR δLR w /w w wELRMarshall's Rules: Labor Demand is more elastic (δLR larger) whenelasticity of substitution is greater(see above)elasticity of demand for rm's output is greater(higher wages lead to higher prices elastic demand means large cut inoutput rms cut employment heavily)the labor's share in total costs of production is greater(if production is labor intensive, even a small increase in wagessubstantially increases marginal cost of production)the supply elasticity of other factors of production is greater(more expensive to substitute workers to capital if the price of capital isvery responsive to demand for capital)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Factor Demand with Many InputsMany di erent inputsSkilled and unskilled labor, old and young workers, old andnew machines.31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Factor Demand with Many InputsMany di erent inputsSkilled and unskilled labor, old and young workers, old andnew machines.Cross-elasticity of factor demand xi /xi w j /w ji.e. the percent change in the demand for input i given a percentage change inthe price of input jSubstitutes: cross-elasticity positiveComplements: cross-elasticity negative31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?The Demand Curve for a Factor of Production isA ected by the Prices of Other InputsPrice ofinput iPrice ofinput iD0D1D0Employment ofinput iD1Employment ofinput iThe labor demand curve for input i shifts up if the price of a substitutable inputrises (left panel) and down if the price of a complement rises (right panel)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Work-SharingMany people believe that the amount of work is xedIf so, unemployment follows from not sharing this work equally decreasing working hours will help the unemployed31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Work-SharingMany people believe that the amount of work is xedIf so, unemployment follows from not sharing this work equally decreasing working hours will help the unemployedHunt (1999) examines the issue empirically by focusing on areduction of standard hours in GermanyIdenti cation problem: Many things can happen simultaneouslywith the working hours reform (e.g. business cycle)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Work-SharingMany people believe that the amount of work is xedIf so, unemployment follows from not sharing this work equally decreasing working hours will help the unemployedHunt (1999) examines the issue empirically by focusing on areduction of standard hours in GermanyIdenti cation problem: Many things can happen simultaneouslywith the working hours reform (e.g. business cycle)Solution: Exploit the fact that cuts in regulard hours weremade on industy-by-industry basis31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Work-SharingMany people believe that the amount of work is xedIf so, unemployment follows from not sharing this work equally decreasing working hours will help the unemployedHunt (1999) examines the issue empirically by focusing on areduction of standard hours in GermanyIdenti cation problem: Many things can happen simultaneouslywith the working hours reform (e.g. business cycle)Solution: Exploit the fact that cuts in regulard hours weremade on industy-by-industry basisTakeaway: (a) hours ,(b) wages ,(c) statisticallyinsigni cant estimates for employment e ect31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?BackgroundGerman unions negotiate industry level minimum wages,regular hours, vacations.(agreements also bind non-union members)After a IG Metall strike in 1984 regular hours for metalworkersreduced from 40 to 38.5 starting from 1985Other unions followed(and IG Metall pushed metalworkers hours to 35 by 1995)Hours cuts accompanied with full wage compenstation (not clear what this means)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

ModelFirm's problemmax g (h, N , K ) whN fN pw (h hs ) N rK{h,N ,K }where g ( ) is the production funtion, h hours per worker, N number of workers, Kcapital, w wages, f a xed-cost of hiring a worker, p overtime premium, hs standardhours (anything above this is overtime) and r rental rate of capital.

ModelFirm's problemmax g (h, N , K ) whN fN pw (h hs ) N rK{h,N ,K }where g ( ) is the production funtion, h hours per worker, N number of workers, Kcapital, w wages, f a xed-cost of hiring a worker, p overtime premium, hs standardhours (anything above this is overtime) and r rental rate of capital.Marginal costs of an extra worker and hourMCNMCH wh f pw (h hs )(1 p ) wh

ModelFirm's problemmax g (h, N , K ) whN fN pw (h hs ) N rK{h,N ,K }where g ( ) is the production funtion, h hours per worker, N number of workers, Kcapital, w wages, f a xed-cost of hiring a worker, p overtime premium, hs standardhours (anything above this is overtime) and r rental rate of capital.Marginal costs of an extra worker and hourMCNMCH wh f pw (h hs )(1 p ) wh The impact of an reduction in12hs(Fig. 1)h hs : N (substitution from workers to hours, scale e ect,substitution from labor to capital) h hs : N (substitution from hours to workers, scale e ect,substitution from capital to workers)

Identi cationThree estimation equations (Sect VI)yln yitln yitln it β2 hsit β3 hsit 1 ui vt it β2 hsit β3 hsit 1 ui vt β1 t it β2 hsit β3 hsit 1 ui vt β1 t β4 wit 1 itwhere yit is employment or hours at industry i at period t , , hs is standardhours hours at this industry at period t , ui is a set of industry-level xed-e ect,vt is a set of time e ects, β1 t captures industry-speci c trends and wit 1 is theitnominal bargained wage in the previous quarterIdentifying assumptions (dif-in-dif )timing of standard hours reduction independent of time-speci c deviations from common time-e ects (after conditioning ontime-invariant industry-e ects) caused by the treatmentunobservablesSpec. as , but allowing for industry-speci c trendsSpec. : as , but also conditioning on industry-level bargained wages(probably should not do this as wages likely a ected by standard hours)

Identi cationThree estimation equations (Sect VI)yln yitln yitln it β2 hsit β3 hsit 1 ui vt it β2 hsit β3 hsit 1 ui vt β1 t it(1)(2) β2 hsit β3 hsit 1 ui vt β1 t β4 wit 1 it(3)where yit is employment or hours at industry i at period t , , hs is standardhours hours at this industry at period t , ui is a set of industry-level xed-e ect,vt is a set of time e ects, β1 t captures industry-speci c trends and wit 1 is thenominal bargained wage in the previous quarteritIdentifying assumptions (dif-in-dif )timingSpec. 1:of standard hours reduction independent oftime-speci c unobservables deviations from common time-e ectsconditioning on time-invariant industry-e ects) caused by the treatment(afterSpec. 2: as 1, but allowing for industry-speci c trendsSpec. 3: as 2, but also conditioning on industry-level bargainedwages (probably should not do this as wages likely a ected by standard hours)

Impact on Employment (Table V)Col 1, last rowpoint estimate: one hour decrease in standard hours decreasesemployment by 3.8 percenthowever, standard error is 3.6 percent no statistically signi cante ectCols 2 and 3: similar resultsIV is used here to correct for attenuation bias due to measurementerror (not selection)note that measurement error always leads to a bias towards zerobut this can be corrected if one has two independent measures ofthe same thing (one measure used as an instrument for the other)Cols 4 and 5: weighting by industry size reduces the estimatesnot clear why to weight (treatment is at industry-level)the di erence between unweighted and weighted results simply tellsyou that less was going on in larger industries. and the di erence between the results from alternativespeci cations are not statistically signi cant anyways

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Impact on Employment (Table VI)Point estimates on col 1 2 suggest that one hour reduction instandard hours decreases employment by 5 6 percent amongmen (statistically signi cant)Hunt: this is too large to be plausibleAs we discussed in the rst lecture, underpowered studiesproduce either insigni cant or very large estimates (see Gelman,Weakliem (2009): Of Beauty, Sex and Power, American Scientist 97)Some point estimates for women (col 5 6) suggest positiveemployment e ects, but these are not signi cant31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Impact on Actual Hours (Table VII)Benchmarksone hour reduction from 40 hours is 2.5 percent acoe cient of 0.025 for standard hours implies one-for-onereduction in actual hoursa coe cient of zero would imply no e ect on actual hours (i.e.full shifting to overtime)Results (cols 1 3): 0.024 (s.e. 0.003)Estimates for men smaller implying some shifting to overtime(consistent with employment results)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Impact on Actual Hours (Tables I and II)These estimates based on individual-level panel datay actual hours or overtime dummies, D standard hoursCols 1 3:overtime);β 0 implies no e ect on actual hours (full shiftingβ 1 complete pass-through to actual hourstoCol 1: Cross-sectional variationCol 2: Within-individual variationCol 3: Within-individual variation IV to correct formeasurement error in reported standard hoursCol 4:β 0 implies an increased likelihoodβ 0 implies no e ectto work overtime ashsdecreases,Col 5:hsβ 0decreases,implies an increased likelihood to work undertime asβ 0implies no e ectTakeaway: One hour reduction in standard hours reduces actualhours by 0.88 1 hours31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?Impact on Wages (Table III)Here no claims for causalityWages and hours jointly determinedFuture reductions in standard hours may be anticipated inbargaining processStill informative about correlation between standard hours andwagesResults: One hour reduction in standard hours associated with2 2.4 percent increase in hourly wage monthly pay remainsroughly constant (for a person not working overtime)31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?ConclusionsHoursmost/all reduction in standard hours transfered to actual hoursWageshourly wages increase when standard hours decrease31E00700 Labor Economics: Lecture 4Matti Sarvimäki

IntroductionShort-RunLong-RunDoes Work-Sharing Work?ConclusionsHoursmost/all reduction in standard hours transfered to actual hoursWageshourly wages increase when standard hours decreaseEmploymentThis study does not have enough data to really tell whathappensNevertheless, a reasonable working hypothesis is thatwork-sharing does not increase employmentAdditional evidence: Crépon, Kramarz (2002, JPE) concludethat an one-hour reduction of the workweek decreasesemployment by 2 4%31E00700 Labor Economics: Lecture 4Matti Sarvimäki

31E00700 Labor Economics: Lecture 4 Matti Sarvimäki 8 Nov 2012. IntroductionShrt-RunoLong-RunDoes Wrk-Shaoring Wrk?o First Part of the Course: Outline 1 Supply of labor 1 static labor supply: basics 2 static labor supply: bene ts and taxes 3 intertemporal labor supply 2 File Size: 1MB