Globalization, Jobs, and Welfare: The Roles of Social Protection and Redistribution 1

Transcription

1 Globalization, Jobs, and Welfare: The Roles of Soial Protetion and Redistribution Priya Ranjan University of California - Irvine pranjan@ui.edu Current Draft Deember, 04 Abstrat This paper studies the welfare and poliy impliations of globalization when risk averse workers fae the risk of unemployment. If the jobs performed by domesti workers an be easily substituted by imports, then globalization redues wages and inreases unemployment. In this situation, in the absene of any government intervention globalization not only redues the welfare of workers but ould redue soial welfare as well. Both unemployment bene ts and severane payments an protet workers against labor inome risk, but the latter enhanes welfare more if job destrution is the soure of unemployment. When optimal redistribution and soial protetion poliies are in plae, globalization neessarily improves soial welfare. Keywords: o shoring, unemployment, endogenous job destrution, severane payments, unemployment bene ts JEL Codes: F6, F66, F68 I would like thank the seminar paripants at the Universities of Calgary, Munih, Linz, UC-Irvine, Paris Shool of Eonomis, GSIS, Geneva, and Gabriel Felbermayr, Kangoh Lee, and Dalia Marin for useful omments.

2 Introdution While eonomists have devoted a lot of attention to the impat of various aspets of globalization on wage and inome inequality, the poliymakers and the publi at large have been more onerned with the impliations of globalization for jobs. However, there has been a reent surge in the researh on the impliations of globalization for jobs. The empirial literature using datasets from various ountries and industries nds mixed results. Dutt, Mitra, and Ranjan (009) nd trade liberalization to be assoiated with lower unemployment at longer intervals in a ross-ountry study, however, there is a spike in unemployment in the immediate aftermath of trade liberalization. A reent study by Autor, Dorn, and Hanson (03) nds that the inreased ompetition from Chinese imports has inreased unemployment in the loal U.S. labor markets and explains about one quarter of the ontemporaneous aggregate deline in the U.S. manufaturing employment. Monarh, Park, and Sivadasan (04) nd a deline in employment for o shoring rms. Wright (04) nds that o shoring has di erential e ets on the employment of workers with di erent skills, however, the overall e et seems to be positive. Gorg (0) provides a survey of the empirial literature on o shoring and unemployment and nds a diverse set of results: o shoring a ets employment adversely in some industries/ountries and positively in others. Given the possibility of globalization inreasing unemployment, at least in the short to medium run, a serious disussion of poliies related to this issue is warranted whih is the subjet of this paper. We onstrut a theoretial model with risk averse workers whih is a key departure from the standard models of globalization and labor market. A single good is produed using domesti labor and o shored/imported inputs with a onstant elastiity of substitution prodution funtion. While all workers are ex ante idential, the math spei produtivity is random, and it is not worthwhile for rms to keep very low produtivity mathes. Wage determination follows the ompetitive searh tradition of Moen (997), and Aemoglu and Shimer (999) where rms post a wage to attrat workers. The advantage of this framework is that the deentralized outome is e ient when workers are risk neutral and therefore, any ine ieny that arises is solely due to risk aversion. In this set up, it is shown that the impat of o shoring on the labor market and welfare ruially depends on the elastiity of substitution between domesti labor and o shored inputs. If there is su ient omplementarity between domesti labor and o shored inputs, then o shoring improves the welfare of workers by lowering unemployment and inreasing wages and inreases soial welfare (measured as the sum of welfares of workers and pro t owners) as well. On the other hand, if o shored inputs an be easily substituted for

3 domesti labor then workers are adversely a eted by o shoring: unemployment inreases and wages derease. In the latter ase, there is an inrease in inequality in the distribution of inome sine non-wage inome rises and wages fall. More importantly, if workers are su iently risk averse, then o shoring not only redues the welfare of workers but redues soial welfare as well. Therefore, in the absene of any instruments for redistribution or soial protetion, there would be a ase for restriting o shoring to inrease soial welfare. The potential welfare loss from o shoring is a onsequene of the risk aversion of workers. If instead, workers are risk neutral then irrespetive of the elastiity of substitution between domesti labor and o shored inputs, o shoring always inreases soial welfare. Moving to poliy issues, we show that the risk aversion of workers auses prodution ine ieny independent of distributional onerns. That is, the deentralized output in the eonomy is lower than what would happen if a soial planner were maximizing output. We explore the role of some ommonly used soial protetion programs in restoring e ieny in the deentralized ase. 3 In partiular, we study the roles of unemployment insurane (UI) and employment protetion (EP) legislation. While the role of UI as an instrument of soial protetion is relatively well known, it is less lear how some elements of EP programs an at as an instrument of soial protetion. Employment protetion refers to a host of mandatory restritions pertaining to the separation of workers from rms. The two key elements of employment protetion are severane payments (SP) whih is a transfer from rms to workers and an administrative ost borne by employers whih does not arue to employees diretly. Given the widespread use of severane payments, a serious disussion of this poliy is warranted. 4 Our theoretial predition that o shoring an inrease unemployment in some industries and redue them in others is onsistent with the diverse empirial ndings summarized in Gorg (0). A more diret evidene is provided in Harrison and MMillan (0). Using data on the U.S. multinationals, they nd that when the tasks performed by the subsidiary of a multinational are omplementary to the tasks performed at home, o shoring leads to more job reation in the United States; however, o shoring auses job losses when the tasks performed in the subsidiary are substitutes for the tasks performed at home. 3 While soial protetion refers to safety nets of various kinds, in this paper we restrit it to mean soial insurane programs that enable individuals to negotiate labor market risk. The main reason for the existene of suh programs in market eonomies is that the market for private insurane against inome risk is missing for various reasons. 4 In a ross-ountry study of severane payments, Holzmann et al. (0) nd that out of 83 ountries for whih information is available, 5 have mandated severane payments shemes (8 perent), 8 have quasi-mandated shemes through omprehensive olletive agreements, and only 3 (7 perent) have neither. 3

4 We show that both UI and SP an restore prodution e ieny in the deentralized ase. 5 That is, by proteting workers against the risk of unemployment, both UI and SP make the eonomy prodution e ient. The e ient level of SP fully insures workers against the risk of unemployment while the e ient level of UI provides inomplete insurane. A onsequene is that the worker welfare and soial welfare are higher with e ient SP than with e ient UI. An administrative ost of ring (whih is not a transfer to workers), on the other hand, exaerbates the existing ine ieny and does not provide insurane to workers. What this suggests is that not all omponents of employment protetion have the same e ieny and welfare e ets, an insight that may be relevant for empirial work. Empirial work on the subjet lumps together all elements of employment protetion in onstruting an aggregate index of employment protetion. Moving from e ieny to soial welfare, while the soial protetion measures like EP and UI inrease output and the welfare of workers, they do not guarantee that o shoring will improve soial welfare. We show numerially that soial protetion alone an onvert the negative welfare impliations of o shoring into a positive one, however, it may not be enough in all ases. The reason is that in the presene of risk averse agents any inequality in the distribution of inome needs to be addressed through redistribution. In the absene of suh redistribution, o shoring an ause soial welfare losses even if e ient soial protetion poliies are in plae. It is shown that when e ient soial protetion and redistribution poliies are in plae, then o shoring neessarily inreases welfare. The baseline model disussed above abstrats from mathing fritions to fous on job destrution whih reates a role for severane payments. Sine mathing fritions are an integral part of the unemployment story, we extend the model to inorporate mathing fritions. Now the adjustment in response to o shoring takes plae through both less job reation and greater job destrution. In partiular, when o shored inputs an easily substitute for domesti labor, o shoring inreases unemployment by inreasing job destrution as well as reduing job reation. The latter happens through a redution in the market tightness. Again, o shoring redues worker welfare if the elastiity of substitution between workers and o shored/imported inputs is high. As well, soial welfare dereases if the degree of risk aversion is high. Looking at poliies, again the deentralized outome is prodution ine ient due to the risk aversion of 5 The di erene between the two in our stati framework is in terms of funding. While SP is either paid diretly by rms or indiretly through a tax on ring, UI is naned either through a tax on workers or a payroll tax on rms. 4

5 workers and e ieny an be restored using soial protetion poliies. One di erene from the baseline model is that sine severane payments (SP) are given at the time of separation, they annot be used to insure workers who are unemployed beause they fail to math. Unemployment insurane (UI) an be used to insure unmathed workers as well as red workers. Therefore, either UI alone or a ombination of UI and SP an be used to ahieve e ieny in the deentralized setting. Consistent with the welfare results earlier, worker welfare and soial welfare are higher with a poliy that ombines SP with UI than UI alone. That is, SP an omplement UI when unemployment arises due to a ombination of job destrution and mathing fritions.. Related Literature Many papers studying the labor market impliations of globalization in eonomies with searh fritions arry out omparative stati exerises with respet to labor market poliies suh as unemployment bene ts, hiring and ring osts et. 6 A ommon approah in these papers is to lump these labor market interventions together with searh fritions and to onlude that the impliations of these interventions are similar to that of an inrease in searh fritions. This equivalene arises beause workers are risk neutral in these papers. An important ontribution of our paper is to show that the welfare impliations of these poliy interventions are very di erent from an inrease in searh fritions when workers are risk averse. By ignoring risk aversion these papers miss out on the insurane role that these interventions play in proteting workers against the risk of unemployment in both losed and open eonomies. The paper most losely related to our work is Keushnigg and Ribi (009), whih to the best of our knowledge is the only paper to study the poliy impliations of globalization in a model with unemployment and risk averse workers. Our model di ers from their model in several respets. While they assume domesti labor and o shored inputs to be perfet substitutes, we work with a CES prodution funtion whih allows us to study ases when o shored inputs are omplementary to domesti labor as in the seminal paper by Grossman and Rossi-Hansberg (008) where this raises the possibility of wages inreasing for workers whose jobs are o shored. In fat, we get a uto value of the elastiity of substitution parameter suh that if the elastiity of substitution is higher than the uto 6 e.g. Moore and Ranjan (005), Helpman and Itskhoki (00), Egger and Etzel (0), Felbermayr, Larh and Lehthaler (03). 5

6 then the workers are hurt by o shoring, but gain otherwise. Additionally, while wages are determined through Nash bargaining in their set up, rms post wages in our framework. A onsequene is that the distortion in our framework arises solely due to the risk aversion of workers even in the presene of searh fritions 7. This allows us to fous on poliy issues arising from risk aversion. Also, while in Keushnigg and Ribi (009) unemployment arises solely beause some workers are unmathed, in our baseline model unemployment arises solely from job destrution while in the extension unemployment arises due to both mathing fritions and endogenous job destrution. Additionally, while Keushnigg and Ribi (009) fous on unemployment bene ts, we study severane payments and unemployment bene ts as alternative ways to provide soial protetion, and in this sense the two papers are omplementary. We show that if unemployment arises solely due to job destrution then severane payments an be a superior tool for insuring workers than unemployment bene ts. When unemployment arises due to both job destrution and mathing fritions, a poliy that ombines severane payments and unemployment bene ts an be superior to unemployment bene ts only. While most of the reent papers on labor market impliations of globalization use models with risk neutral workers thereby obviating the need for soial protetion, there is an older literature in international trade dealing with risk averse agents. For example, Newbery and Stiglitz (984) onstrut a model with risk averse agents where trade an be Pareto inferior to autarky. Dixit and Rob (994) show how trade may be inferior to autarky in the presene of missing insurane markets when individuals are risk averse. Due to missing insurane markets, the deentralized solution di ers from the planner s problem and hene trade an be inferior to autarky or even a tari equilibrium an be inferior to autarky. This is similar in spirit to our result desribed earlier that when domesti labor is a good substitute for o shored inputs, o shoring an redue soial welfare. However, these papers do not deal with the labor market risk arising from unemployment. Among other related papers, Brander and Spener (994), Feenstra and Lewis (994), and Davidson and Matusz (006) study various poliies to ompensate the workers who lose from trade. However, workers are risk neutral in these papers. Closer to our approah is the paper by Breher and Chaudhuri (994) whih examines the issue of Pareto superiority of free trade over autarky through Dixit-Norman 7 With Nash bargaining in the presene of searh fritions and large rms, as in Keushnigg and Ribi (009), there are two distortions even with risk neutral workers when large rms hire many workers: searh externalities and the "overhiring e et" identi ed by Stole and Zwiebel (996). This makes the poliy analysis more ompliated in suh a setting. 6

7 ompensation shemes when there is unemployment in the eonomy aused by e ieny wage onsiderations and unemployed workers get unemployment ompensation. In this setting, workers who beome unemployed due to trade an be fully ompensated for their losses only if unemployment bene ts beome equal to the wages. However, in this ase, no e ort will be undertaken by any worker, and hene output will beome zero. Therefore, fully ompensating workers who lose their jobs is not feasible. Even though this paper has unemployment as well as unemployment ompensation, workers are risk neutral and hene the insurane motive for unemployment bene ts is not present. As far as the related work on soial protetion is onerned, while muh work in labor/maro eonomis fouses on the administrative ost aspet of employment protetion, Pissarides (00) and Blanhard and Tirole (008) highlight the potential role of severane payments in providing insurane. Our stati model of endogenous job destrution with large rms employing multiple inputs an be viewed as a generalization of the one-worker- rm model of endogenous job destrution in Blanhard and Tirole (008). The large rm model with heterogeneous math spei produtivity of workers is also similar to Helpman, Itskhoki and Redding (00). In their model rms have to sreen the mathed workers after bearing a ost to nd out if the produtivity of workers is above a uto. Workers below the uto are not hired. Given rm heterogeneity, more produtive rms sreen more whih leads to di erent rms having workers with di erent average produtivities resulting in di erent wages. This set up allows them to study the impliations of globalization for wage inequality. Sine our fous is on the employment e ets of globalization with risk averse workers, we reate a simpler framework with homogeneous rms where the math spei produtivities are revealed to rms ostlessly as in Blanhard and Tirole (008). To summarize, the key ontributions of this paper are the following. In the absene of any government intervention, the deentralized equilibrium is ine ient from the point of view of both prodution and welfare. In this setting, globalization an redue worker welfare as well as soial welfare by inreasing unemployment and redistributing inome from workers to pro t owners. Labor market interventions like severane payments or unemployment insurane inrease unemployment but make the eonomy prodution-e ient (maximize the value of output), and in ombination with redistribution an ensure that globalization is soial welfare improving. Finally, severane payments are superior to unemployment bene ts when job destrution is the sole soure of unemployment, and a ombination of severane payments and unemployment bene ts is superior to unemployment bene ts alone when unemployment 7

8 is aused by both job destrution and mathing fritions. In the next setion we present the baseline model without searh fritions. Setion 3 studies the impliations of o shoring for labor market and welfare and onduts the poliy analysis. Setion 4 presents the extension with searh fritions. Setion 5 provides a disussion of robustness issues. Setion 6 provides onluding remarks. All the derivations are gathered in appendix A and the proofs of lemmas and propositions in appendix B. The Model The prodution funtion is given by Z = A((L e ) + M ) ; 0 < < ; () where L e is the domesti labor in e ieny units and M denotes foreign produed inputs. aptures the elastiity of substitution between domesti labor and foreign produed inputs and aptures the diminishing returns. Diminishing returns an arise either due to limited span of ontrol as in Luas (978) or due to the presene of some spei fator in xed supply. 8 It is also assumed that there is a ontinuum of domesti rms of unit mass so there is no distintion between a rm level variable and an eonomy level variable. 9 Workers are idential ex ante but their math spei produtivity, ; is random. Without loss of generality, assume that is drawn from a uniform distribution over [0; ]. This is a standard distributional assumption in the literature on endogenous job destrution (e.g. Mortensen and Pissarides (994)). In the benhmark model we assume the mathing to be fritionless and later we extend the model to allow for mathing fritions. One the math spei produtivity of a worker is revealed, the rm an deide whether to retain the worker or re them. Firing ould be ostly due to mandated severane payments or administrative burden. If rms use a uto rule whereby they retain workers 8 If = ; then domesti labor and o shored inputs beome gross omplements, therefore, one annot disuss the ase of gross substitution whih is the ase when domesti workers ould lose from o shoring. < allows us to disuss both the ases of gross substitution and gross omplementarity. 9 As disussed in the "Disussions" setion later, the impliations of allowing for free entry whih makes the mass of rms endogenous is similar to the ase of =. 8

9 with produtivity above and re others, then the average produtivity of retained workers is + : If they hire L h workers then they retain ( )L h of them, and hene the amount of labor in e ieny units that is used in prodution is L e = L h = + L; () where L is the number of workers retained by the rm. Therefore, the prodution funtion () an be written as Z = A + L + M : (3) The above implies that rms fae a quantity-quality trade-o in the hiring of workers. To produe a given level of output, they an go for higher quality and lower quantity or vie-versa. Sine ring is ostly, higher quality omes at a higher ost. The total number of workers in the eonomy is denoted by L: Denote the aggregate pro t of rms by : The pro t is distributed among N agents whih ould be the owners of the spei fator used in prodution. Eah owner gets a share given by = N : (4) All agents are risk averse with the utility funtion given by 0 U(x); U 0 > 0; U 00 < 0 (5) where x is their inome. Sine all workers are mathed in the baseline model and some are retained while others are red, the inome of workers when they are retained is x = w; where w is the wage, while the inome when they are red is x = z where z is the value of leisure/home prodution. For pro t owners, x =. Firms post wages and ring rates to attrat workers. Denote the wage rate posted by rm-i by w i and the uto produtivity by i (same as ring rate given the uniform distribution of ): Workers diret their appliations to the rm whose (w i ; i ) pair gives them the highest expeted utility. Suppose 0 While we are assuming rms to be risk neutral (they simply maximize pro ts), the pro t reipients are assumed to be risk averse as are the workers. Making the reipients of pro ts risk neutral won t hange any results beause there is no unertainty in their inome. 9

10 W is the highest utility that a worker an expet from a job at another rm. Now, in order to attrat workers, (w i ; i ) must satisfy ( i )U(w i ) + i U(z) W: (6) E etively, for any ring rate that the rm posts, (6) determines the wage that the rm has to o er. If a rm wants to raise the average produtivity of its workfore by being more seletive (higher i ) then it will have to o er higher wages. The main advantage of using wage posting is that, as shown later, the deentralized equilibrium is e ient (orresponds to the planner s solution) when workers are risk neutral. Therefore, any ine ieny in the model arises due to the risk aversion of workers. This allows us to fous on the poliy issues arising from risk aversion. Even though looking at (6) one gets the impression that rms an hoose di erent pairs of (w; ) to satisfy (6), it an be shown from the rm s maximization exerise that all rms end up posting the same wage rate. Therefore, in the analysis below we drop the rm subsript i: Denote the per unit prie of the imported/o shored input by : Now, rms perform the following pro t maximization exerise. subjet to the onstraint Max fz wl Mg L;M;w; ( )U(w) + U(z) W: (7) In writing the rst order onditions for the above maximization exerise and throughout the paper, we use the following ompat notation: Notation : z L = + 0 L + M ; z L + M A : Note that this way of modeling labor market is similar in spirit to the ompetitive searh framework of Moen (997) and Aemoglu and Shimer (999) where rms post wages and workers diret their searh. The di erene is that in the ompetitive searh framework rms post wages, whih for a given W determines the length of the queue, q i; and onsequently how fast the vaany is lled. That is, a rm is hoosing a pair (w i; q i) to ensure that the worker gets a utility of W; while in our framework the rm hooses (w i; i) to ensure that the worker gets a utility of W: This an be aomplished by noting that the wage rate an be expressed as a funtion of W and in the rm s maximization exerise. Sine eah rm takes W as given, it ends up hoosing the same ; whih implies the same wage rate. 0

11 Using % to denote the Lagrangian multiplier on the onstraint in (7), the rst order onditions for the above maximization are given by + L : Az L L = w (8) M : Az L M = (9) w : L + %( )U 0 (w) = 0 (0) A + : z L L = %(U(w) U(z)) () Intuitively, the l.h.s of (8) is the marginal produt of an additional retained worker while the r.h.s is the ost of a retained worker. Similarly, the l.h.s of () is the bene t of a higher, whih for a given L results in higher average produtivity of these workers. The r.h.s is the ost of a higher resulting from the higher wages to satisfy the wage onstraint beause when the probability of getting red is higher it must be o set by a higher wage. This ost is related to the risk aversion of workers. The greater the risk aversion, the greater the ost in terms of meeting the reservation wage of workers. Sine all workers are mathed, the number employed simply equals the number not red and therefore, the aggregate labor market equilibrium ondition is given by L = L( ): () The 5 equations (8)-(), and () determine w; L; M; ;and %: It is shown in the appendix that using (8)-() and () we an obtain the following two key equations in w and whih are useful for proving the existene of equilibrium as well as omparative statis. w = + ; (3) + ( A + where we use the following ompat notation: ) Notation : U(w) U(z) U 0 ; w (w) L = w( ); (4) When workers are risk neutral, the existene and uniqueness of an interior equilibrium with (0; ) and w > z is easily established in the appendix. When workers are risk averse, the possibility of a

12 orner solution ( = 0) with full employment, but w > z; exists. This ase an be ruled out by a su ient ondition on parameters (high L or low A) suh that at full employment the marginal produt of labor is less than z: This yields the following result. Proposition : A unique equilibrium always exists with the equilibrium wage, w; neessarily exeeding z: Additionally, under a su ient ondition on parameters (high L or low A); there always exists an interior equilibrium with positive unemployment: (0; ): Next we derive the expressions for pro ts and soial welfare whih are useful in deriving some key results in the paper later. Using () the expression for aggregate pro ts in equilibrium is given by = A L + M A The measure of welfare of workers is W whih an be written as w( )L M: (5) W = ( )U(w) + U(z): (6) Soial welfare is given simply by the sum of welfares of workers and pro t owners: SW = NU() + LW: (7) Before studying the impliations of o shoring for welfare, it is useful to understand the distortions aused by risk aversion in our model. Below we show that risk aversion makes the deentralized eonomy prodution-ine ient and any inequality in the distribution of inome between workers and pro t owners makes soial welfare sub-optimal. To see the prodution ine ieny resulting from risk aversion in the deentralized ase, we ompare it to a planner s problem where the planner is interested in maximizing aggregate output.. Planner s problem The planner an hoose a uto produtivity, ; o shored input, M; and employment L to maximize the following. Z M + z(l L): (8) The planner reognizes that higher leads to higher unemployment, that is L = ( )L; and therefore, the planner maximizes Z P M + z L; (9)

13 where Z P A L + M A : (0) It is shown in the appendix that the e ient level of is given by the solution to the following equation. A + z L = z: () It is proved in the appendix that the equation above has a unique solution whih we all e where e (0; ):.. Comparison of deentralized equilibrium with the planner s problem Case of Risk Neutral Workers The following lemma is easily veri ed in the ase of risk neutral workers, that is when the utility funtion is of the form: U(x) = ax + b where a > 0 and b are onstants. Lemma : When workers are risk neutral the deentralized equilibrium is prodution-e ient. That is, when workers are risk neutral, the deentralized equilibrium unemployment rate and output are same as one obtained by a soial planner interested in maximizing output. Therefore, when workers are risk neutral there are no distortions in the model eonomy from the point of view of prodution e ieny. The results parallel the e ieny of deentralized equilibrium in a ompetitive searh framework as in Moen (997). Similar to Moen (997), wage posting by rms delivers an e ient outome in the deentralized ase. Later when we inorporate searh fritions in the model, it is still the ase that the deentralized outome is e ient when workers are risk neutral. Case of risk averse workers It is shown in the appendix that when workers are risk averse, the in the deentralized equilibrium is given by the solution to the following equation. A + z 0 L = z 0 ; () where z 0 w : Denote the solution to the above equation by r : Comparing () whih gives us the e ient level of with () giving us the deentralized equilibrium value of ; we obtain the following result. 3

14 Lemma : When workers are risk averse, the deentralized equilibrium level of is ine iently low ( r < e ). This is similar to the result of Aemoglu and Shimer (999) that the deentralized equilibrium level of unemployment is too low when workers are risk averse. While they work with single-worker- rms and the soure of unemployment in their framework is searh fritions, here we obtain this result in a large rm model with endogenous job destrution. What happens is that risk averse workers prefer a lower unemployment rate and are willing to aept lower wages to keep the unemployment rate low ( r < e ) : Lemmas and learly establish that the deentralized outome is prodution-ine ient due to the risk aversion of workers. When we talk about soial welfare de ned in (7), the onavity of the utility funtion of agents implies that inequality in the distribution of inome is another distortion whih will be important for results on soial welfare. Having identi ed the key distortions in the model, we turn to the impat of globalization on unemployment and welfare. 3 Globalization, Unemployment and Welfare As mentioned earlier, globalization in the model is aptured by a redution in the prie of o shored/imported input, M: The following proposition is proved on the impat of globalization in a deentralized equilibrium. Proposition : A redution in the ost of o shoring inreases wages and redues unemployment if < ; leaves them unhanged if =, and redues wages and inreases unemployment if > : Intuitively, a derease in has two e ets on the demand for domesti labor. Sine o shored inputs are heaper now, rms substitute away from domesti labor. However, there is a produtivity e et arising from the inreased usage of o shored inputs. That is, the inreased usage of o shored inputs inreases the marginal produt of domesti labor. For > the substitution e et dominates, and hene the demand for domesti labor dereases (domesti labor and o shored inputs are gross substitutes). As rms redue their demand for domesti labor, the expeted reward of labor, W; dereases. This derease in W allows rms to raise : More mehanially, at the aggregate level the 4

15 amount of labor employed in e ieny units is L e = ( ) L: Therefore, the only way the amount of labor employed in e ieny units an derease is through an inrease in : For < the produtivity e et dominates (domesti labor and o shored inputs are gross omplements) leading to an inrease in the demand for domesti labor resulting in lower unemployment and higher wages. The expressions for the impat of o shoring on the welfare of workers and soial welfare (derived in the appendix) are given by dw d NU 0 () d d + LdW d = U 0 (w) ( ) dw d = (U 0 (w) U 0 ())L d d ( ) dw d d d (3) U 0 ()M (4) Before disussing the welfare impliations of o shoring for the ase of risk averse agents, it is useful to note the results for the ase of risk neutral agents: U(x) = ax + b. The following result is easily veri ed from (3), (4), and proposition. Proposition 3: When agents are risk neutral, o shoring inreases workers welfare if < ; leaves it unhanged if = ; and redues it otherwise. However, o shoring always inreases soial welfare. Lemma veri ed the e ieny of the deentralized equilibrium with risk neutral workers. Sine there is no di erene between aggregate output and soial welfare when agents are risk neutral, it is not surprising that o shoring, whih is like a positive produtivity shok, is welfare improving for the eonomy as a whole. Going bak to the ase of risk averse agents, note from proposition that there are two relevant ases to disuss. Case : < ) dw d d < 0 and d > 0: In this ase, o shoring unambiguously inreases the welfare of workers. Additionally, soial welfare inreases as well if U 0 (w) > U 0 (): That is, as long as the marginal utility of inome for workers is higher than for pro t owners (we assume this to be the ase throughout), soial welfare inreases with globalization. O shoring inreases pro ts diretly through a deline in the prie of o shored inputs whih is a soure of welfare gain (last term in (4)). In addition, note from (3), (4) that the term giving the hange in the welfare of workers (( ) dw d d d ) has the opposite e et on pro ts and, therefore, its net impat on soial welfare depends on the respetive marginal utilities of workers and pro t owners. Under the assumption of U 0 (w) > U 0 () soial welfare hanges in the same diretion as 5

16 the welfare of workers. Therefore, in ase soial welfare inreases unambiguously. Case : > ) dw d d > 0; d d < 0; and d < 0: In this ase, the welfare of workers learly dereases. Sine workers are poorer than pro t owners (U 0 (w) > U 0 ()), soial welfare dereases due to a derease in the welfare of workers. The diret e et of o shoring on pro ts is always positive, and therefore, the net impat of o shoring on soial welfare is ambiguous. Note that workers are hurt by both a derease in wage and an inrease in unemployment. An inrease in unemployment inreases the risk they have to bear ausing welfare losses in exess of what would obtain in the presene of insurane markets. The greater the risk aversion the larger the deline in the welfare of workers and hene the greater the possibility of soial welfare losses from globalization. Numerial simulations reveal that when the degree of risk aversion is high, soial welfare dereases as the ost of o shoring dereases. Figures and provide numerial examples. Both gures are based on a CRRA utility funtion of the type U(x) = x gures = 4 and = =3 so that we are in the > where is the oe ient of risk aversion. In both ase, and = :5 (low risk aversion) in gure and = 3 (high risk aversion) in gure : In both ases as the ost of o shoring dereases unemployment ( ) inreases ( gures a and a) and wages derease ( gures b and b) and onsequently the welfare of workers dereases ( gures and ). The di erene is in soial welfare. While in panel d soial welfare inreases when the degree of risk aversion is low, in panel d soial welfare dereases with a higher degree of risk aversion. Sine wages derease and pro ts inrease, the inequality in the distribution of inome as measured by pro ts to wage inome also rises. The results derived above assuming U 0 (w) > U 0 () are summarized in the proposition below. Proposition 4: When < ; o shoring redues unemployment and inreases wages, thereby, inreasing the welfare of workers as well as soial welfare. When > ; not only does the welfare of workers derease but soial welfare an derease as well. In the latter ase, there is an inrease in inequality in the distribution of inome as well sine pro ts rise and wages derease. It follows from proposition 4 that there may be a ase for reating obstales to o shoring if no other poliy interventions are available. Sine o shoring dereases the welfare of workers and possibly soial welfare when > ; our disussion of various poliies below fouses on this ase. 6

17 4 Poliy Analysis It was shown earlier that the risk aversion of workers reates distortions whih give rise to the possibility of globalization ausing soial welfare losses. As seen in proposition 3, with risk neutral agents (obviating the need for an insurane market or redistribution), globalization is soial welfare improving. Below we disuss two types of poliies. One, labor market interventions that help workers negotiate labor market risk and two, redistribution. We study three ommon labor market poliies- severane payments, unemployment insurane, and ring taxes whih are not transfers to workers- and analyze their potential to restore produtione ieny in the eonomy and analyze the impat of o shoring in the presene of these poliies. 4. Deentralized equilibrium with alternative poliies The rst poliy we disuss is a ring tax, f t ; by the government whih is not a transfer to workers. This an be thought of as the administrative burden imposed on rms with the aim of reduing ring. The seond poliy we disuss is mandated severane payments (SP), f w. This is a transfer from the rm to the red worker. Finally, we disuss unemployment insurane (UI) given to red workers. In the publi nane literature the funding of UI takes many alternative forms: a lump sum tax on all workers; a tax on only employed workers; or a payroll tax on rms. The results in all ases are qualitatively similar and we hoose to disuss only the ase where the tax is on employed workers (same as in Keushnigg and Ribi (009)). Denote the unemployment bene ts by b: This is naned by a tax, ; on employed workers, therefore, the balaned budget ondition is given by b = ( ): Note that if UI is naned by a tax imposed on red workers, then in our urrent framework it is equivalent to the mandated severane payments. Therefore, the key di erene between SP and UI in the baseline model is in terms of naning. While the former is either paid diretly by rms to red workers or funded by a ring tax olleted by the government, the latter is funded through one of the three alternative ways disussed above. 3 3 While the U.S. does not have a mandated SP, the ontribution of the employers towards funding UI is experiene rated whih essentially means that it is related to the number of workers they re. That is, the funding of UI in the U.S. makes it similar to a severane payment program. 7

18 Below we develop a uni ed framework that nests all 3 poliies and then disuss eah in turn. Our goal is to see if prodution-e ieny an be restored using these poliies. The equilibrium with poliies is solved using a two stage game where the planner hooses the poliy in the rst stage and then rms maximize their pro ts taking the poliies as given. With the above poliies in plae the rms perform the following maximization exerise in the seond stage. Z wl subjet to the onstraint Max L;M;w; (f w + f t ) L M ( )U(w ) + U(b + f w + z) W: (5) The rst order onditions for the above maximization exerise are derived in the appendix where we derive the following ondition haraterizing the equilibrium hoie of : A + w p (f t + f w ) ; L = w p (f t + f w ) ; (6) where p U(w ) U(b+fw+z) U 0 (w ) : Below we disuss eah of the three poliies in turn. 4.. Administrative ost of ring Setting b = = f w = 0 in (6) obtain A + w p f t L = w p f t ; (7) where p = U(w) U(z) U 0 (w) in this ase: Comparing (7) to (), note that ring taxes lead to e ient if p = w z f t : The onavity of U( ) implies that p > w z (sine w exeeds z); therefore, the e ient level of f t is haraterized by w z f t > w z or f t < 0: That is, e ieny requires a negative level of administrative burden of ring. Sine the administrative burden of ring an at most be redued to zero, it annot help ahieve e ieny beause we have already seen earlier that when f t = 0 the deentralized outome is ine ient. Intuitively, sine is too low in the absene of any intervention, a poliy restoring e ieny must raise : Inreasing the administrative burden of ring (inrease in f t ) ends up reduing whih makes the existing distortion worse. 4 4 To see how f t > 0 lowers below the e ient level, note that > w z implies that z > w and hene 8

19 The result above suggests that a ring subsidy by the government may ahieve e ieny. Suppose we think of f t < 0 as a monetary ring subsidy and ignore the issue of raising money for a ring subsidy. Can suh a ring subsidy restore e ieny? The answer turns out to be no. We verify numerially that the equilibrium is non-monotoni in f t : That is, as the level of ring subsidy inreases rms initially re more workers but after a point (well before the e ient level of ) they start ring less. Therefore, the e ient level of is not attained by a ring subsidy. The rough intuition is that a ring subsidy doesn t address the underlying distortion arising from the risk aversion of workers. While we are fousing on e ieny, it is worth noting that inreased administrative burden of ring does sueed in lowering, and hene redues unemployment. Therefore, if the goal of poliy is to simply redue unemployment, then in our set up an inrease in f t is able to ahieve this goal. 4.. Severane payments To obtain the expression for the equilibrium level of with severane payments, use b = = f t = 0 in (6) and obtain A + w p f w L = w p f w ; (8) where p = U(w) U(fw+z) U 0 (w) : Comparing (8) with () note that severane payments lead to e ient hoie of if p = w f w z. Sine U 00 ( ) < 0; the only solution to p = w f w z is f w = w z; that is, a severane payment that provides full insurane restores e ieny. How would the government hoose suh a f w? For any f w hosen by the government the orresponding deentralized equilibrium is (f w ) and w(f w ) where x(f w ) is the equilibrium value of x for a given f w : The government solves f w = w(f w ) z to get the e ient level of f w : Therefore, a severane payment that results in full insurane delivers the e ient level of in the model. While we have disussed severane payments as a government mandated poliy, it is worth pointing out that rms will have an inentive to provide severane payments to risk averse workers voluntarily. It is easy to verify from the model that if rms ould o er severane payments, they would do so z > w f t: Following the same reasoning as in the proof of lemma, one an verify that the that solves (7) is lower than the that solves (). 9

20 and the equilibrium level of severane payments will orrespond to the e ient level disussed above. Essentially, rms would o er a ontrat with wages and a level of severane payments that fully insures workers. However, there may be reasons why rms are unwilling or unable to o er severane payments. One possible reason is wage rigidity. Note that in order for rms to o er insurane through severane payments, they should have the ability to redue the wages of employed workers. However, wage rigidity may prevent them from doing so. Alternatively, in real world severane payments rely on a long term ontrat whereby workers aept a lower wage in return for a promise to get severane payments when they are red. Now, ontratual fritions an reate problems with this kind of ontrat. Modeling these issues is beyond the sope of this paper, but they suggest why there may be a role for mandated severane payments Unemployment insurane To obtain the expression for with unemployment insurane, set f t = f w = 0 in (64) and obtain A + w p L = w p ; (9) where p = U(w ) U(b+z) U 0 (w ) and the balaned budget ondition implies = b: Again, omparing (9) with () note that a level of unemployment bene ts, b; leads to e ient if p = w z: The e ient level of b an be found as follows. For eah b there is an equilibrium w(b) and p (b): The planner solves for b suh that w(b) p (b) = z: 5 It an also be veri ed that the e ient level of unemployment bene ts does not imply full insurane. Full insurane implies p = 0; while e ieny requires p = w z: The two an be satis ed together only if w = z and b = = 0; whih annot be true in any equilibrium (see proposition ). Thus, both severane payments and unemployment bene ts an be used to ahieve e ieny, however, while the former provides full insurane to workers, the latter doesn t. This has impliations for welfare whih is summarized in the proposition below and proved in the appendix. Proposition 5: The e ient levels of severane payments and unemployment bene ts yield the same levels of output, unemployment, and pro ts, however, the welfare of workers as well as soial 5 It was mentioned earlier that unemployment bene ts an be naned alternatively using a payroll tax on rms or a lump sum tax on all workers. The outome (output, unemployment, pro ts, welfare) with the e ient level of unemployment insurane in either of these ases orresponds exatly to the ase disussed in the text. 0

21 welfare is higher with e ient severane payments than with e ient unemployment insurane. Intuitively, sine the missing market for insurane is the key obstale preventing rms from hoosing e ient ; by providing insurane, both unemployment bene ts and severane payments allow rms to hoose their e iently. The di erene between the two in terms of welfare impliations arises from the fat that e ient level of severane payments involves full insurane while the e ient level of unemployment insurane involves inomplete insurane. Figure 3 provides a numerial example of the omparison between e ient severane payments and e ient unemployment insurane when the CRRA risk aversion parameter = 3 and >. The red line depits the ase of e ient severane payments, the blak line depits the ase of e ient unemployment bene ts, and the green line depits the ase of no poliy intervention. Figure 3a shows that the wage is higher in the ase of e ient unemployment insurane ompared to the ase of severane payments whih in turn is higher than the no intervention ase. In gure 3b the vertial axis is the ratio of onsumption in the unemployed state to the onsumption in the employed state. It shows that e ient severane payment provides full insurane but e ient unemployment bene ts provide inomplete insurane. The omplete insurane with e ient severane payments is also re eted in a higher worker welfare in gure 3 and a higher soial welfare in gure 3d ompared to the e ient unemployment bene ts. As well, e ient unemployment bene ts yield higher worker and soial welfare than the no intervention ase. It is worth re-iterating that both worker welfare and soial welfare are higher with e ient poliies (severane payments and unemployment bene ts) than without intervention at all levels of o shoring ost (the red and blak lines in gures 3 and 3d lie well above the green line). This important result, driven by the risk aversion of workers, is in ontrast to several studies mentioned in the introdution (see footnote 6) whih lump these poliies together with searh fritions and onlude that their impliations is similar to an inrease in searh fritions, whih is to redue welfare. 6 The point is that models with risk neutral workers miss out on the insurane role of these poliies in both losed and open eonomies. In general even with e ient levels of SP or UI in plae, o shoring an redue welfare as an be seen from gure 3. However, there is a range of risk aversion parameter for whih the presene of e ient poliies turns the impat of o shoring on welfare from negative to positive. Figure 4 provides 6 Later we show that our results on the welfare impliations of severane payments and unemployment bene ts are robust to the inlusion of searh fritions.

22 an example when = :5 and the poliy is severane payments: Figures 4a and 4b plot worker welfare and soial welfare with respet to the o shoring ost in the absene of e ient soial insurane poliies and show that both derease as the o shoring ost dereases. When e ient soial insurane poliies are in plae, gure 4 shows that worker welfare still dereases, however, gure 4d shows that soial welfare inreases. 7 We summarize the result below. Proposition 6: For some parameter values, while o shoring redues soial welfare in the absene of soial insurane poliies, it inreases soial welfare with e ient soial insurane poliies in plae. More generally, despite the presene of e ient soial insurane, o shoring an redue soial welfare beause from the point of view of soial welfare, there are two distortions in the model: lak of insurane and inequality in the distribution of inome. The latter an be addressed using redistribution, whih is what we turn to next. 4. From E ieny to Welfare 4.. Welfare maximization by the planner Our earlier analysis of the planner s problem foused on output maximization beause we wanted to talk about prodution-e ient poliies. Now we look at the planner s problem when the planner is interested in maximizing soial welfare given by the sum of welfares of workers and pro t owners as de ned in (7). We assume that the planner provides a transfer b 0 to unemployed workers, w 0 to employed workers, and y to pro t owners or the owners of the spei fator and performs the following maximization exerise. Max U(b 0 + z) + ( ;M;b 0 ;w;y )U(w 0 ) L + NU(y) subjet to the onstraint L(b 0 + z) + ( )Lw 0 + Ny Z P M + Lz; where Z P is the output de ned in (0). It is veri ed (see appendix) from the above maximization exerise that there is no trade-o between equity and e ieny. That is, the level of from the above maximization is given exatly by the 7 The same result is obtained with e ient unemployment insurane, however, as expeted, welfare (both worker and soial) is higher with e ient severane payments than with e ient unemployment insurane.

23 ondition (). Therefore, the planner simply maximizes net output Z P M + Lz and then redistributes it among workers and owners of spei fators to equalize their marginal utilities by hoosing b 0 ; w 0 ; and y suh that U 0 (b 0 + z) = U 0 (w 0 ) = U 0 (y): We all this the rst-best ase. How an this outome be ahieved in a deentralized equilibrium? Below we show that it an be deentralized using mandated severane payments and a redistributive transfer. 4.. Welfare Maximization with severane payments In a deentralized equilibrium the planner does not hoose the wage rate, w or : However, the planner an mandate severane payments f w and a redistributive transfer s where s > 0 implies a transfer from pro t owners (owners of spei fators) to workers. Unlike the planner s welfare maximization disussed earlier where the transfers to workers were unonstrained, now we are onstraining the transfers to employed and unemployed workers to be idential. The rms and workers take f w and s as given. Therefore, the deentralized equilibrium an be solved as a two stage problem where in the rst stage the planner hooses f w and s and then rms hoose w, L; and in the seond stage. The planner hooses f w and s to maximize the following in the rst stage. Max ( U(f w + z + s) + ( )U(w + s)) L + NU (y) ; f w;s where y = Z P M f w L ( )Lw sl : N In the seond stage the rms hoose ; L; and w to do the following maximization. subjet to Max Z M wl L;w; f w L U(f w + z + s) + ( )U(w + s) W: In addition, the equilibrium ondition L = ( )L must be satis ed. Reall that the soial planner makes a transfer of b 0 to unemployed workers and w 0 to employed workers in the planner s welfare maximizing solution in the rst-best ase disussed earlier. It is proved in the appendix that there exists a pair of f w and s that repliates the rst-best outome derived in the previous sub-setion. This gives us the following important result. 3