The consequences of (de)regulation on employment duration and efficiency: an experimental study.

Berninghaus, Siegfried ; Bleich, Sabrina ; Guth, Werner 等

I. INTRODUCTION

When little is known about the future market conditions, offering an employment contract that regulates working conditions for many periods can be dangerous. If market wages drastically decline, one may hire labor more cheaply. Furthermore, an employee may turn out less reliable than expected. Such risks are by no means farfetched. Rather, they relate to crucial aspects of long-term employment. (1) So why is long-term employment still predominant and short-term employment, for example in the form of rented labor, rare, albeit increasing (see Alewell et al. 2007)? An explanation might be the increasing institutional flexibility regarding wage adjustments. Institutional downward flexibility is usually not legally codified but often a consequence of court rulings and thereby based on social norms (2) (see, for instance, Bewley 1995).

In our paper, we want to investigate whether more flexibility in wage adjustment (from none to full flexibility) increases the duration of accepted employment contracts. Our first hypothesis (Hypothesis A) which we wish to test experimentally is that contract duration depends on contract flexibility.

One obvious argument for long-term employment is that offering a short-term contract reveals distrust in the newly hired worker and encourages shirking. But what is wrong with long-term employment resulting from both partners, employer, and employee, repeatedly and mutually opting for rematching? Another reason for long-term employment could be that, in the long run, it allows the interacting parties more easily to overcome efficiency losses due to opportunistic behavior. In our experimental scenario, regulations prevent efficient employment contracts when workers react opportunistically. But do long-term employees care more for efficiency, and therefore invest more than optimal efforts?

Deregulation (3) has been the major policy recommendation for nearly all developed market economies. For the most part, the effects of (de)regulation are predicted by rational choice analyses of labor markets. An interesting exception that takes into account the empirical evidence from behavioral labor economics (Fehr, Gotte, and Zehnder 2009) is the analysis of Agell (1999), in whose tradition we want to provide a behaviorally informed discussion of (de)regulation. Rather than doing this by field data our empirical approach is an experimental one. Our earlier (experimental) findings (Berninghaus, Bleich, and Giath 2008) question the idea that labor market efficiency is higher in deregulated markets. To determine more systematically whether the usual intuition that wage rigidities undermine efficiency is correct, or whether our earlier results are more reliable, we analyze more thoroughly the interdependency of wage flexibility and employment duration. Our second hypothesis claims that long-term contracts are substituted by renewable contracts when contracts become more inflexible (Hypothesis B). Unlike in our previous study, where we varied only the flexibility of fixed wages, in this study, piece rates may also be adapted periodically.

In the experiment of Brown, Falk, and Fehr (2004), where work contracts last for only one period, long-term relationships between employers and employees can only be established if the employer offers a new contract to the same employee in each period. We too allow for voluntary renewal of contracts at the end of each period, but also the possibility of offering contracts which extend for more than one period. Our study is more systematic than previous ones because we do not just compare less with more labor flexibility but distinguish five flexibility treatments which can be at least partially ordered from less to more flexibility. For any kind of monotonicity hypothesis, this can be viewed as a real stress test.

Note that already the one-off interaction of our basic labor market scenario allows for reciprocity (4) since employees choose effort after accepting a more or less favorable employment contract. For long-term contracts, further reciprocation is possible due to future dealings. Although backward induction would predict shirking in finitely repeated gift exchange experiments, behaviorally repeated gift exchange experiments inspire more effort than one-off interactions (Falk, Gachter, and Kovacs 1999; Falk and Gachter 2002; Krnigstein and Villeval 2010). In our experiment, additionally long duration can be efficiency-enhancing via effort-smoothing, which, however, requires some very sophisticated anticipation. We rather expect something like "the longer the better"-heuristics in the sense that mutually agreeable outcomes inspire longer job tenure.

With respect to employment duration, our discussion is related to the (in)completeness of employment contracts (for contract theory, see Bolton and Dewatripont 2005) and thereby to the crowding-out or crowding-in of long-term employment. Will more regulated labor markets crowd out contractual and crowd in voluntary long-term employment, for example in the sense that long-term employment is intended but not contractually codified, or will long-term employment be crowded-out altogether?

Although job tenure is long in most countries, the reasons for this are sometimes manifold. Usually, after a probation period, employment relations cannot easily be terminated by the employer unless there are good reasons, such as misbehavior by the employee, or when this (type of) labor input becomes obsolete. Policy can regulate labor markets by labor law. The latter, however, is more often than not interpreted and implemented by labor courts in surprisingly free ways. In Germany, for instance, most labor courts will require the employer to pay half a monthly salary for each year of employment when the employer wants to fire an employee. In our scenario, employment duration is not legally regulated but rather an endogenous aspect to be determined by the labor parties.

Upward flexibility has never been an issue--an industrial employer is always able to improve the contract terms in favor of the employees. Downward flexibility is often very much restricted: it is either explicitly ruled out by labor law or an evolved attitude of labor courts. At the end of the last century even employers considered it impossible to lower the wages in ongoing employment relationships (see, for instance, Bewley 1995). That this is quite common nowadays is a consequence of serious deregulation of labor markets in most Western countries due to the (fear of the) global economy.

In Section II, we introduce the stochastic, multi-period labor market model which we have experimentally implemented. This model is theoretically analyzed in Section III. Section IV contains the experimental design and the statistical analysis of our results. The major findings concerning our two main hypotheses show a strong tendency to shorter contracts in more regulated markets (Hypothesis A) and more contract renewals in less flexible labor markets (Hypothesis B). Section V contains a detailed discussion of policy implications.

II. THE STOCHASTIC MULTI-PERIOD ENVIRONMENT

In the following, we present the multi-period model of labor contracting, which is suited for experimental testing. In our theoretical framework, we assume money maximizing agents without denying or postulating that participants might exhibit social preferences.

The employers are allowed to offer either one-period or multi-period contracts, which in reality are mostly open-ended. (5)

In every period, t = 1,2, ..., the n ([greater than or equal to] 2) employers i = 1,2, ..., n and workers j = 1,2, ..., n are matched to pairs with the option to establish employment. If they are not matched, there is no employment relation involving this employer and employee in this period. In our view, such momentary take-it-or-leave-it power is rather realistic because usually it is the employer who is aware of the job opportunity. Due to the multi-period interaction, such power is, however, restricted.

The first period t = 1 precludes already existing employment relations. In t = 1, each employer i is randomly matched with one worker j. First, all 2n agents are informed about the randomly determined market wage [w.sup.c.sub.1] in this period (6) without being provided with any clue about future market wages [w.sup.c.sub.t] in periods t > 1. In each pair (i, j), employer i then offers worker j an employment contract

([w.sulp.j.sub.i], [s.sup.j.sub.i], [T.sup.j.sub.i]),

with [w.sup.j.sub.i] ([greater than or equal to] 0) denoting the fixed wage, (7) [s.sup.j.sub.i] [member of] [0, 1] the revenue share for the worker, and [T.sup.j.sub.i] ([greater than or equal to] 1) the employment duration. If worker j accepts, he finally chooses his effort level [e.sup.j.sub.i] ([greater than or equal to] 0). In case of an established employment relation, worker j earns

[U.sub.j] = [w.sup.j.sub.i] + [p.sub.i][s.sup.j.sub.i][e.sup.j.sub.i] - [([c.sub.j]/2)([e.sup.j.sub.i]).sup.2],

where [p.sub.i](>0) is firm i's sale price and [c.sub.j](>0) worker j's effort cost parameter, whereas employer i earns

[[PI].sub.i] = [p.sub.i] (1 - [s.sup.j.sub.i])[e.sup.j.sub.i] - [w.sup.j.sub.i],

that is, effort costs are private costs of workers. If worker j does not accept i's employment offer, he is employed externally at the market wage [w.sup.c.sub.1], whereas i earns nothing in that period.

In periods t > 1, the randomly selected market wage [w.sup.c.sub.t] is made publicly known first without out any clue concerning [w.sup.c.sub.[tau]] for [tau] > t. What may differ from Period 1 is that some pairs (i, j) have already decided to go on with their employment relationship, possibly after adjusting the contract where the flexibility depends on the treatment. In all periods t [greater than or equal to] 1:

* In ongoing relationships (i, j), employer i decides whether and how to adjust the contract, to which employee j can react by his effort choice [e.sup.j.sub.i], whereas

* in newly matched pairs (i, j), the process is the same as in the first period (i offers a contract which j either rejects to be employed externally at the market wage [w.sup.c.sub.t] or accepts and chooses his effort [e.sup.j.sub.i]).

Clearly, there can be at most n ongoing relations (i, j), and each employer i without an employee can be matched with an unemployed worker j. Only newly matched pairs can, furthermore, fail to establish employment, meaning that i earns nothing and j the market wage [w.sup.c.sub.t]. It may appear unrealistic that the employee cannot simply leave her employer in case of a long-term contract. (8) But if the employee can simply run away, this undermines long-term contracts altogether. Furthermore, even when staying with her employer, the employee is by no means defenseless, since she can react to unfair treatment by zero effort.

When an employment relation expires, both partners can opt for rematching. Only when both agree will these two be rematched; otherwise i will be randomly matched with any of the unemployed workers j, just as each j will be randomly matched with an employer i without a worker. (9) Full flexibility can be achieved irrespective of the flexibility treatment by renewed short-term contracts with the same partner. We can thus test the crowding out of long-term employment in two ways: restricting flexibility in adjusting employment contracts reduces either contract length [T.sup.j.sub.i] or voluntary employment duration, that is, by reducing mutual rematching.

Of course, an employer may not exploit the flexibility in adjusting the contract terms and, when the employer exploits downward flexibility, the employee may accept such adjustments to market conditions without reducing efforts. In view of earlier preliminary findings of Berninghaus, Bleich, and Guth (2008), and also the inspiring and illustrative discussion of Irlenbusch (2008) we predict the opposite: flexibility of downward adjustments will be used by employers, and this encourages shirking by employees to an extent that renders such downward adjustments unprofitable and inefficient.

III. THEORETICAL ANALYSIS

The following rational choice analysis (10) assumes common (knowledge) of own payoff maximization and risk neutrality and first discusses the one-period interaction before continuing with repeated interaction (see also Berninghaus, Bleich, and Guth 2008). The one-period interaction analysis explores the strategic problem of an employer and an employee concluding a new contract in the final or the only period of the experiment.

A. One-Shot Interaction

In a given pair (i, j) with employer i and worker j, the optimal effort is

[e.sup.j.sub.i] = [p.sub.i][s.sup.j.sub.i]/[c.sub.j].

If the market wage [w.sup.c] also determined the fixed wages, that is, for [w.sup.j.sub.i] = [w.sup.c], the optimal revenue share would result in [s.sup.j.sub.i] = 1/2. Thus worker j would earn [w.sup.c] + [p.sup.2.sub.i]/[8c.sub.j]. This benchmark behavior could be justified by a collectively negotiated legal minimum wage.

If firms are restricted in t = 1 to [w.sup.J.sub.i] [greater than or equal to] 0, and accordingly in a later round t to [w.sup.j.sub.i] [greater than or equal to] [bar.w] for some given positive N in ongoing employment, the optimal revenue share can be derived by maximizing

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

and [w.sup.j.sub.i] [greater than or equal to] 0, respectively [w.sup.j.sub.i] [greater than or equal to] [bar.w]. If optimality requires

[w.sup.j.sub.i] + [(([p.sub.i][s.sup.j.sub.i]).sup.2]/[2c.sub.j]) = [w.sup.c],

one can substitute [w.sup.j.sub.i] and obtain

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Since [partial derivative][[PI],sub.i]/[partial derivative][s.sup.j.sub.i] > 0 for all 0 [less than or equal to] [s.sup.j.sub.i] < 1, the global optimum would be obtained for [s.sup.j.sub.i] = 1. This fulfills the requirement of [w.sup.j.sub.i] [greater than or equal to] 0 if [w.sup.c] [greater than or equal to] [p.sup.2.sub.i]/[2c.sub.j], which is violated for all possible market wages [w.sup.c] in our experiment. Similarly, one must have [w.sup.c] [greater than or equal to] + [bar.w] + [p.sup.2.sub.i]/[2c.sub.j] for a positive wage [bar.w] due to long-run employment and inflexibility of [w.sup.j.sub.i]. Revenue shares [s.sup.j.sub.i] = 1 would mean that the employer is renting out her firm to the employee for a fixed fee in the sense of [w.sup.j.sub.i] < 0, which we exclude by restricting [w.sup.j.sub.i] to non-negative levels.

When [s.sup.j.sub.i] = 1 is excluded, a boundary solution requires 1/2 < [s.sup.j.sub.i] < 1 since [s.sub.i.sup.J] = 1/2 is optimal for given fixed wages and [s.sup.j.sub.i] = 1 when fixed wages can be varied freely. For [w.sup.c] < [p.sup.2.sub.i]/[2c.sub.j] the binding constraint [w.sup.j.sub.i] = 0 requires

[s.sup.j.sub.i] = ([square root of [2c.sub.j][w.sup.c])/[p.sub.i].

For new contracts in the last period, as well as for myopic players, we expect to observe only one-period offers with a piece rate depending on the market wage of that period and a fixed wage of zero. This combination guarantees the employee the market wage, given an optimal effort choice.

B. Finitely Repeated Interaction

When T is finite and commonly known, the usual backward induction can be applied. The benchmark solution of the one shot-interaction applies in the last round t = T, when a pair is voluntarily or randomly formed in the last round, or when both parties are (known to be) myopic. But this does not provide the usual starting point for a stationary benchmark solution as in repeated games without any structural dependencies across rounds.

To illustrate the possible gains by contractual effort-smoothing, consider the second last round of a finitely repeated interaction with a uniform distribution of market wages on the integers of the interval [[[w.bar].sup.c], [[bar.w].sup.c]] with [[w.bar].sup.c] < [[bar.w].sup.c]. For a newly formed pair (i, j) one option in the second last round is to implement a contract with [T.sup.j.sub.i] = 1 and opt for rematching in the last round. This would guarantee j the payoff [w.sup.c.sub.T-1] in the current round and the expected payoff 0.5. ([[w.bar].sup.c] + [[bar.w].sup.c]) in the last round. To demonstrate efficiency gains due to effort-smoothing we derive the profit-maximizing contract with [T.sup.j.sub.i] = 2, which makes employee j indifferent between accepting a two-period contract, yielding [w.sup.c.sub.T-1] + ([[w.bar].sup.c] + [[bar.w].sup.c])/2, and the one-period option. Guaranteeing j the same payoff requires a two-period contract with [s.sup.+] satisfying

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Offering ([w.sup.j.sub.i] = 0, [s.sup.+], [T.sup.j.sub.i] = 2) would grant employer i all possible gains from contractual effort-smoothing. It thus remains only to show that employer i prefers this contract which, due to j's indifference, is acceptable to employee j. Neglecting the same constant labor costs of employing worker j, contract ([w.sup.jsub.i] = 0, [s.sup.+], [T.sup.j.sub.i] = 2) yields for employer i 2[[p.sub.i](1 - [s.sup.+])([p.sub.i][s.sup.+]/[c.sub.j])] = 2([p.sup.2.sub.i]/[c.sub.j])(1 - [s.sup.+])[s.sup.+], whereas i, in case of the [T.sup.j.sub.i] = 1-option, earns

[p.sup.2.sub.i]/[c.sub.j][(1 - [s.sup.*]([w.sub.T-1]))[s.sup.*]([w.sub.T-1]) + E[(1 - [s.sup.*]([w.sup.C]))[s.sup.*]([w.sup.c])],

where [s.sup.*](x) is the profit-maximizing revenue share with arguments [w.sup.c] [member of] [[[w.bar].sup.c], [[bar.w].sup.c]]. Thus we have to prove

2(1 - [s.sup.+])[s.sup.+] - (1 - [s.sup.*]([w.sub.T-1]))[s.sup.*]([w.sub.T-1]) - E[(1 - [s.sup.*]([w.sup.c]))[s.sup.*]([w.sup.c])] > 0

for all [w.sup.C]. One can easily show that this inequality holds for our particular parameter constellations [w.cup.c] = 13 and [[bar.w].sup.c] = 30 (see Appendix A).

In a similar vein, parties can perceive the last two rounds as just one terminal period and convince themselves that contractual effort-smoothing is even more profitable when extending it over more than just two rounds. Proceeding inductively, this would finally prove that effort-smoothing is optimally achieved by offering the longest possible contract duration. (11)

The previous analysis applies to all situations with no downward flexibility of piece rates which guarantee the employee a certain positive payoff. One of our experimental treatments allows up- and downward changes of fixed wage and piece rate over a long-term contract. As the employer has full flexibility after the employee has accepted the contract, his preferred contract includes [s.sup.j.sub.i] = 1/2 and [w.sup.j.sub.i] = 0. The employee takes this into account when accepting a contract in a treatment with full flexibility. The employer's offer, therefore, has to compensate the employee for giving up the sum of the expected market wages for the duration of the offered contract. (12)

There is experimental evidence (e.g., Anderhub, Gachter, and Konigstein 2002) that parties, especially employees, react optimally to whatever contract they are facing. As shown by Brown, Falk, and Fehr (2004), being together in the long run may, however, encourage the parties to behave less opportunistically and be more efficiency minded instead. For the case at hand, maximizing the total surplus of a given employer i and employee j-pair means investing the effort level [e.sup.+.sub.j] = [p.sub.i]/[c.sub.j] (i.e., efficient production), yielding the surplus of [p.sup.2.sub.i]/[2c.sub.j], which both parties can freely allocate among themselves by an appropriate fixed wage [w.sup.j.sub.i] in the range [w.sup.j.sub.i] [greater than or equal to] 0 and, if necessary, by a revenue share [s.sup.j.sub.i] [member of] [0, 1]. Of course, such a policy would have to rely on trust and reciprocity. In the full flexibility treatment [F.sub.+], for instance, the employee would have to trust in the reciprocity of her employer, whereas in the case of no flexibility at all (Treatment N), this would be reversed. Such efficiency-enhancing cooperation can also be achieved by voluntary and mutual rematching, that is, without any codified obligation (see the corresponding evidence of Brown, Falk, and Fehr 2004).

IV. EXPERIMENTAL PROCEDURES

A. Experimental Setting

The experiment was conducted at the University of Karlsruhe. Subjects were students of various faculties. Treatments differed in flexibility of contracts in ongoing long-term employment due to [T.sup.j.sub.i] > 1. In each session, the ten participants represented a matching group. Members of a matching group interacted for ten rounds and were partitioned into a group of five employers and a group of five employees. (13) Subjects without a given partner were randomly rematched within their matching group after each round. All subjects received an initial endowment of 7.50 Euro at the beginning of the experiment. After reading the instructions (see Appendix B for a translation of the German instructions) they all had to fill out a computerized control questionnaire checking whether the rules were understood. The experiment started when all questions were answered correctly.

In each round, after the employer-employee pairs were formed and the market wage had been announced, the employer proposed a contract which could be accepted or rejected by the employee, except when the pair was already engaged in a long-term contract. After each round, a participant was informed about her current payoff. Once each participant in the matching group had made her decision, the next round started. Participants were able to recall their payoffs of previous rounds at any time on the computer screen.

Employer subjects had to choose fixed wages [w.sup.j.sub.i], revenue shares [s.sup.j.sub.i], and duration of contract [T.sup.j.sub.i]. Employee participants had to fix their effort level [e.sub.j] in each period after accepting the contract. At the beginning of each period, the prevailing market wage [w.sup.c.sub.t], a random number uniformly distributed over a given interval of integers, was announced to all subjects.

The choices were restricted in the following way:

p 10
c 1
[s.sup.j.sub.i] [member of] [0, 1]
[w.sup.j.sub.i] [member of] [0, 60]
[T.sup.j.sub.i] Integer with 1 [less than or equal to]
 [T.sulp.j.sub.i] [less than or equal to]
 "number of remaining rounds"
[w.sup.c] [member of] {13, 14, ..., 30}
[e.sub.j] [member of] [0, 499]

Our five treatments differed in the labor market restrictions that limited wage or revenue share flexibility during contract duration. Starting with Treatment N, exhibiting complete regulation but still allowing effort choices to react to market wages, we ordered the remaining treatments by decreasing degrees of regulation in Table 1. Each session involved one matching group (consisting of ten subjects). Thus we employed 400 participants in total.

An upward arrow means that the respective variable may be increased, a downward arrow that the variable can be decreased during contract duration. One could have distinguished further treatments, such as that with no [w.sup.j.sub.i] but full [s.sup.j.sub.i] flexibility, or full flexibility of one component and only upward flexibility of the other. But in our view, the five treatments here systematically explore and compare the effects of (de)regulation and should suffice to check the robustness of any findings. The accumulated payoffs of each participant were paid out in cash anonymously shortly after the experiment. (14) The average payoff for all treatments is given in the last three columns of Table 1 revealing already that employer participants did not suffer from inflexibility. (15) How rigidities can result from labor law and the evolved attitudes of labor courts has already been discussed in Section I. Our quite complete treatment design is motivated by finding out which rigidities cause the strongest efficiency and employment duration effects. Some treatments, for example N, are unlikely to ever be implemented, but provide an interesting benchmark case, and one more relevant than, for instance, a one-shot control in the sense of a compulsory T = 1-treatment.

Owing to the stochastic future market wages, more rigidity of long-term (T > 1) contracts implies a risk for both parties to be contrasted with the intuition behind "the longer the better" heuristics--or, at a more sophisticated level, with the gains of effort-smoothing. This is why we expect monotonically fewer T > 1contract offers and more voluntary (and mutual) rematching when moving from the bottom to the top treatment in Table 1. Although we randomly assign student participants to the roles (of employers and employees), one can also speculate that when becoming an employee one would prefer a long-term contract with a more or less certain income over a highly volatile wage.

In the following data analysis we will

* compare our experimental findings qualitatively with what (game) theory predicts,

* try to answer our basic research question, whether stricter labor market restrictions crowd out contractual or mutually consented employment duration.

B. Experimental Results

First, we give a descriptive overview of the data set. Table 2 provides the averages of the most important variables. All averages except those for efforts are calculated for the periods in which a new contract was offered. We did this to achieve a better comparison of treatments with different rules of contract adjustment. (16) Effort, however, is averaged over all effort choices, that is, periods with rejected contracts are omitted. The asterisks denote significant differences between treatments according to a Kruskal-Wallis test (at a 5% level of significance). (17)

We find the highest acceptance rate in treatment [I.sub.+], providing maximal insurance against wage reductions without excluding upward wage adjustments. (18) Employers offer the shortest contract duration in the case of no flexibility (N) and the longest in case of full flexibility (F+). (19) Efforts are highest in Treatment N: employees seem to prefer certain wage incomes and reward employers by higher efforts for such certainty. Purely opportunistic employees should invest the lowest effort levels in Treatment N. Revenue shares are all near 0.5, seemingly a focal point for employers. The relative frequencies of the 0.5 piece rate offers increase with the flexibility of the treatment: 37%, 26%, 29%, 36%, and 28%, respectively. These offers are the most frequent ones for all treatments. Actually, the correlation between market wage changes [increment of [w.sup.c.sub.t]] = [w.sup.c.sub.t] - [w.sup.c.sub.t-1] and fixed wage changes [increment of [w.sub.t]] = [w.sup.j.sub.i](t) - [w.sup.j.sub.i](t - 1) is highly significant. (20) Whereas these correlations are highly significant, correlations between market wage and fixed wage respectively piece rates in long-term contracts are poorly correlated. We find a significant correlation between market and fixed wage for treatments [I.sub.+] and [F.sub.+] according to Bravais-Pearson tests. As these are the treatments where both wage components can be adjusted, there seems to be only a limited influence of market wages on employers' wage choices. In these tests as well as in the following statistical tests, we take individual decisions as independent observations unless declared otherwise.

Before analyzing the individual decisions of both contract partners in more detail, we give a short overview of the effects of increasing contract flexibility on adapting relevant contract variables like wages and piece rates. Will increases in piece rates (like in regime [I.sub.+]) be reduced when firms are additionally allowed to decrease piece rates (like in regime [F.sub.+])? Or will we observe an increase in fixed wage reductions when comparing regime [F.sub.+] (where both wage rates and piece rates are allowed to adapt in both directions), for example, with a regime where piece rate adaptations are restricted (like in F or in [I.sub.+])?

According to Figure 1, the stability of contract terms increases with more regulated contract regimes, especially when comparing I, [I.sub.+], and F. But there is only a slight difference in contract terms stability when comparing F and [F.sub.+]. When employers are additionally allowed to reduce fixed wage rates but must keep piece rates fixed (cf., I and F) there is a sharp increase in fixed wage reductions while the percentage of wage increases remains almost the same. When employers are allowed to adapt wages and piece rates (see [I.sub.+]), a significant share of wage increases is substituted by piece rate increases. Comparing regimes F with fixed piece rates and [F.sub.+] with full flexibility of piece rates, we observe a slight increase in wage rate increases (21) while the wage rate reduction in F is partly substituted by piece rate reductions.

Interdependence Between Degree of Regulation and Contract Duration. A completely myopic employer would offer one-period contracts only, whereas a less myopic employer may aim at "contractual effort-smoothing." Let us test when which type of employer prevails. Maximal duration always exceeds significantly the actually offered duration (Table 3), but as we see from the average offered durations not only one-period contracts were offered.

RESULT 1. Participants offered contracts with significantly less than maximal duration, but the average offered contract duration significantly exceeds one period.

Overall, we expected longer contracts in the less-regulated treatments, especially for treatments F and [F.sub.+] where employers can positively and negatively respond to workers' effort choices in a long-term contract. We tested Hypothesis A by applying a Jonckheere-Terpstra Test on increasing duration with increasing flexibility (from N to [F.sub.+]). To test Hypothesis B which claims that in case of less flexibility T > 1 will be substituted by mutual voluntary continuation of the partnership, we applied a nondirectional Kruskal-Wallis One-Way ANOVA on Ranks and isolated by Dunn's Method the treatments causing the difference. The difference is highly significant (H = 51.427; p < .001). To isolate the treatments causing the difference, we compare the treatments pairwise using Dunn's Method. (22) The values of the test variable Q indicate, at the 5% level, significant differences between treatments F, F+ and the remaining treatments. Both tests support our hypotheses:

RESULT 2. Offered contract duration increases with increased contractual flexibility. Treatments F and [F.sub.+] differ significantly from the remaining treatments.

[FIGURE 1 OMITTED]

As an offer does not necessarily lead to a contract, we also tested the duration of concluded contracts for trends over treatments. Again, a Jonckheere-Terpstra test shows that duration of accepted contracts increases with increasing flexibility of wages.

RESULT 3. The duration of accepted contracts increases with increased flexibility of contracts supporting Hypothesis A: more flexibility in adapting contract conditions favors contracts with longer duration.

Acceptance Decisions. Confronted with a contract offer, a rational worker should check whether the expected income of the contract is larger than, or equal to, the expected income from only one-period contracts based on optimal values of piece rate s and effort e. (23) We look at actual and optimal acceptance of contract offers in Table 4, where, as in Table 5 the unit of observation is each acceptance decision, that is, we neglect that several of those come from the same agent. A contract is called "optimally accepted" whenever the worker's income from it is larger than the expected income from one-period contracts over the same number of periods.

Actual acceptance rates significantly differ from optimal acceptance rates in all treatments except F. For treatments with only upward flexibility, contracts are less often than optimally accepted. In treatments with the possibility of lowering the wage(s), more contracts are accepted. Workers do not seem to fear exploitation in treatments allowing employers to reduce payments to employees. To determine whether the acceptance rate of long-term contracts is higher in treatments with a more restricted adaptation of revenue shares, an ANOVA test on ranks was performed, which showed no significant difference in acceptance behavior of workers across treatments regarding long-term contracts.

RESULT 4. Workers' acceptance of contracts is neither optimal nor explicable by fear of flexibility.

To explore the determinants of workers' acceptance decisions in more detail, we ran regressions whose detailed results are presented in Table A1. Revenue share and fixed wage have a positive effect on acceptance. All other determinants (offered duration, a new-contract dummy, and the market wage) deter employees from acceptance. These results additionally confirm Result 4, that contract acceptance does not depend on the treatment.

Effort Choices. After accepting a contract, workers choose effort which is limited to positive values. The optimal effort [e.sup.*] = 10 x s does not depend on the fixed wage. Hence, our next aim is to test the optimality of effort choices given offered piece rates. Table 5 gives an overview of the results. (24) All workers' effort choices were compared pairwise with the optimal effort in the respective period by applying a Sign-Test for each treatment separately. Each individual effort choice is taken as an independent observation. As can be seen from the Pearson Product-Moment correlations in Table 5, the correlation between actual and optimal effort choice is positive and highly significant. The Sign-Tests show that in all treatments except N, there is a significant difference in the distribution of actual and optimal effort, in the sense that workers exert more effort than is optimal. This may be since in Treatment N workers' efforts can be neither rewarded nor punished.

RESULT 5. Workers' effort choices in Treatment N are nearly optimal and there is a highly significant positive correlation between optimal and actual efforts. Otherwise, workers exert more effort than is optimal.

To analyze the determinants of effort choices in more detail, we ran three regressions on selected variables. The results in Table 6 show that both the fixed wage rate and the revenue share have a significantly positive influence on the workers' effort decisions. (25) The first entries in Table 6 represent the [beta]-values of the regression, the numbers in brackets denote the standard deviation, and the term below indicates the p value.

The coefficient for revenue shares is much higher than that for fixed wages, that is, although there is evidence for gift exchange (see Fehr, Kirchsteiger, and Riedl 1993) the reactions to incentives are stronger. Let us confront this finding with the theoretical benchmark solution. To derive the workers' trade-off utility from fixed wages and piece rates according to our theoretical model, we determine the slope of the indifference curves between s and w resulting in [[absolute value of dw/ds] = 10 x e, that is, a piece rate reduction would have to be compensated by a ten times larger fixed wage increase. (26) We conclude--in line with theory--that piece rates are the most important determinant for workers' effort choices. (27)

Furthermore, the regressions reveal some learning. The negative time trend together with our results in Table 5 (higher than optimal efforts in almost all treatments) show that higher than optimal effort levels are observed mainly in the early periods, suggesting that employee participants have concerns about reputation early in the process.

Are Long-term Contracts More Profitable?. To gain from effort-smoothing, (see Section III) firms should prefer long-term contracts. To check the interdependence between contract duration and profits, we computed Pearson-Moment Correlations between the duration of accepted contracts and the resulting profits of employers for all treatments separately. We did not find any significant interdependence and thus conclude:

RESULT 6. Employers do not profit from increased contract duration: employers do not gain from effort-smoothing in long-term contracts.

According to Result 6, effort-smoothing seems a rather far-fetched theoretical possibility which participants either do not recognize or simply discard.

If deregulation in the form of higher contractual flexibility increases efficiency, we should find this in our data. First, we test employers' gains from concluded contracts over treatments. A Jonckheere-Terpstra test on increasing employer gains results in significantly increasing gains, when flexibility is increased. Although we found in Result 3 that contract duration also increases with flexibility, we cannot establish a causal interdependence between contract duration and employer gains (from effort-smoothing). We executed the same test for workers' profits and also for the sum of profits (net joint profits). Net joint profits, which can be seen as an indicator of efficiency, also significantly increase, whereas workers' profits significantly decrease. Thus deregulation increases efficiency, regardless of contract duration. Workers' decreasing payoffs have to be interpreted with caution: they result mainly from their high efforts (see Result 5). In Figure 2, we see that workers' average profits are higher than employers' for all treatments.

[FIGURE 2 OMITTED]

RESULT 7. Employers' gains and net joint profits increase with contractual flexibility. Workers' gains are lower in the case of more fexible contracts.

Are Long-term Contracts Crowded Out by Renewed Contracts? Table 7 shows how employers and employees deal with short-term, long-term, and renewed contracts. The percentages of short- and long-term contracts offered add up to 100%. Renewed contracts may be short term or long term. Offered and accepted contract duration increases with the flexibility of the treatment. (28) The same unambiguous assertion cannot be made for renewed contracts: between 19% and 34% of the contracts are renewed ones. We also had a closer look at participants who opted for a renewed pairing: if only one contract partner opts for rematching, in most cases it is the employer, and this tendency seems to increase with contractual flexibility. Overall, employers opt for another round with the same employee at about 75% of ending contracts, while in Treatment N they do this in about 60% of all cases.

Treatments differ significantly with respect to the percentage of contract offers in renewed pairings and the percentage offering long- and short-term contracts. (29) We find support for our main hypothesis (Hypothesis B), that in more regulated markets long-term contracts are replaced by contract renewals (more short-term and fewer long-term contracts from Treatment N to [F.sub.+]).

RESULT 8. The percentages of contract offers in renewed relationships as well as those of long-term contracts differ significantly across treatments.

The missing trend in the frequency of renewed contracts may be caused by heterogenous durations of renewed contracts. What determines the decision to opt for rematching? Do employers offer a new contract to a worker when the match was successful in the sense of positive profits? And do workers opt for rematching after a positive experience with an employer? We used the [chi square]-test to see how opting for rematching depends on the payoffs of both partners. For employers, this decision always depends on the period's payoff, whereas for workers such dependency exists only in treatments 1 and F.

RESULT 9. Opting for rematching does not correlate with contractual flexibility. Employers' behavior is influenced mainly by the payoff extracted from the match; for employees this is observed only in treatments 1 and F.

V. CONCLUSIONS

In both our theoretical analysis and our experimental scenario, long-term employment is possible in two ways: by offering and accepting longer contract duration or by mutually agreeing to rematch. Thus, although the five treatments cover nearly the full spectrum of no to full flexibility in adapting contract terms, one could always have established long-term employment regardless of how regulated the labor market is.

In the experiment, employers seem to follow a behavioral benchmark instead of acting optimally: fixed wages are significantly positive, piece rates divide the returns from effort equally, and offered contract duration is significantly shorter than would be optimal for effort-smoothing. Employees' acceptance decisions also differ from optimal ones. They accept more contracts than optimal for N to [I.sub.+] and less for F and [F.sub.+]. Effort is higher than optimal: to piece rates of about 50% offered by employer participants, they react with effort levels between 6.5 and 8, whereas optimal efforts range only between 5.2 and 6.2. The offered and accepted contract durations (Results 2 and 3), effort levels (Result 5), and frequencies of opting for rematching (Result 9) differ significantly between treatments.

As employers offer contracts with longer durations in treatments with higher flexibility, our hypothesis regarding the influence of labor market flexibility on contract duration is confirmed (Hypothesis A). Employers' profits increase when regulation decreases. Longer contract duration, however, does not generate increasing gains for employers. Workers' gains are lower when contracts are more flexible, but (Figure 2) they always get more than 50% of the surplus. Opting for rematching by employers increases with a more flexible labor market regime (Hypothesis B). Employees seem to exploit situations with inflexible contracts in which they cannot be punished for low efforts. This leads to low gains for employers, who do not want to be matched with the same worker again (Result 9).

Finally, what can we say about policy implications? Does (de)regulation of labor markets really pay? Do deregulated markets work more efficiently? Unfortunately, our results do not provide an unambiguous answer. The tables above suggest a flexibility ordering < of treatments (30) such as N < I < I+ < F < [F.sub.+], where A < B means that market organization B is more flexible than market organization A.

Using the average effort choices as a measure of market efficiency, we actually find a strictly increasing monotone relationship between market flexibility and market efficiency, if treatment [I.sub.+] is omitted. (31)

Finally, how does (de)regulation affect the average employment duration between employer-employee pairs? Assuming the same flexibility ordering < between our treatments as above and omitting again treatment [I.sub.+] we obtain a hump-shaped curve describing the dependence between market flexibility and average employment duration. Employment duration is highest in treatment I. (32)

ABBREVIATIONS

CU: Currency Units

ECU: Experimental Currency Units

doi: 10.1111/j.1465-7295.2011.00436.x

APPENDIX A: THE EFFICIENCY OF EFFORT-SMOOTHING

To show that [T.sup.j.sub.i] = 2 is preferred by employer i to repeated one-period contracts, we have to check that the following inequality

2(1 - [s.sup.+])[s.sup.+] - (1 - [s.sup.*]([w.sub.T-1]))[s.sup.*]([w.sub.T-1]) - E[(1 - [s.sup.*]([w.sup.c]))[s.sup.*]([w.sup.c])] > 0

holds for any [w.sub.T-1] [member of] {13, ..., 30}.

Let us denote the left-hand side of this inequality by Diff(w). Inserting the parameter values of our experimental design p = 10, c = 1, we obtain the explicit expression Diff(w) = 0.141421[square root of 43 + 2w] - 0.141421 [square root of w] - 0.650804.

The diagram in Figure A1 shows how Diff(*) depends on w.

[FIGURE A1 OMITTED]

[FIGURE A2 OMITTED]

Diff remains positive over the whole range of wages w [member of] {13. ..., 30}, supporting our statement that contractual effort-smoothing pays (in the two-period scenario). By reformulating the problem, we analyze how the profitability advantage of a two-period contract depends on the range of the distribution of market wages. To avoid technicalities, we consider uniform market wage distributions with increasing minimum wage [[w.bar].sup.c] while keeping the maximum wage [[bar.w].sup.c] constant. (33) We vary the minimum wage from 13 to 30. The result is illustrated for selected [[w.bar].sup.c]-values by the diagram in Figure A2.

We conclude from this that the advantage of a two-period contract shrinks when the variance of the market wage shrinks. Intuitively, this is what one would expect. It pays more to offer a long-term contract when the "uncertainty" of the market wage increases.

APPENDIX B: TRANSLATED INSTRUCTIONS

The instructions presented below are those for Treatment I. It is easy to see how they should be used for the remaining treatments. The German instructions use a shorthand for variables which we related to German vocabulary used. We kept this notation for reasons of authenticity rather than substituting them by those in the theoretical analysis, related to the English vocabulary.

In this experiment, you can earn real money, which will be paid out in cash at the end of the experiment. The experiment lasts for 10 periods. How much you earn depends on your decisions and the decisions of the other participants. Every participant makes her decisions isolated from the others, sitting at separate computer terminals. Communication among participants is not allowed.

A participant will be randomly assigned the role of an employer (AG) or a worker (AN). She will be informed about her role at the beginning of the experiment and will keep this role until the end.

Every participant receives an initial endowment of 150 CU (currency units).

General Procedure

At the beginning of each period, the period's market wage M will be announced to all members of a group. For this wage rate each worker will be able to find employment in case of not being contracted by an employer. Only the market wage of the present period will be announced, market wages of future periods are not known. Every employer is randomly matched with one of the workers and offers her a contract. This consists of the fixed wage, the contract duration, and the revenue share of the produced quantity. Each worker can accept or refuse this contract offer. If she accepts, she chooses the planned product quantity. Employers and workers are paid according to the contract. When a contract expires, employers and workers are asked if they want to interact again with the same person in the next period. If both agree, they are matched again in the next period.

First-Round Procedure

1. The random market wage for the present round is announced, which can vary between 13 and 30.

2. Employer-worker pairs are randomly selected.

3. The employer offers a contract characterized by the following items:

* A fixed wage F (in CU), where F [greater than or equal to] 0.

* A share a, where 0 [less than or equal to] a [less than or equal to] 1 of the total production.

* The contract duration L, where 1 [less than or equal to] L [less than or equal to] number of remaining periods.

4. Workers see the contract offered by "their" employers and decide to accept it or not. If the contract is not accepted, the worker receives the prevailing market wage M, and the employer has zero return.

5. If the contract is accepted, the worker chooses the production quantity Q, which is sold for 10 CU. The division of the production quantity is determined by a.

The worker's return from an accepted contract in this round is given by:

F + 10 x aQ - 1/2 x [Q.sup.2].

The employer's return is given by:

10 x (1 - a)Q - F.

If the worker refuses the contract, she receives the prevailing market wage. The employer has zero return.

6. Per-period earning and the sum of all previous periods' earnings are presented (in CU) on the computer screen.

Procedure in the Following Rounds

For participants not yet restricted by a long-term contract, the procedure in the following rounds does not differ from the first round, where the offered contract duration is restricted to the number of the remaining rounds. For employer-employee pairs bound by a long-term contract, the employer may increase the fixed wage after the market prevailing in this round has been announced. Then the workers decide how much to produce.

If a contract has expired, both partners are asked if they want to be matched again with the same partner in the next period. If both agree, they can enter a new contract in the next round. Otherwise, both are randomly matched with another participant.

Please consider that in a long-term contract the worker's production share and the contract duration do not change, while the fixed wage may be raised by the employer in each round. In a contract modified in this way, the increased fixed wage is automatically accepted. The worker may change the production quantity in every contract period.

History

During the experiment you can call up your "history" at any time by pressing the button at the lower border of your computer screen or by pressing the FI key. The following information about previous rounds is given: prevailing market wage, fixed wage, production share, remaining contract periods, acceptance of offers, quantity produced, employer's return, employee's return.

Payoff

You will be paid immediately after the experiment is finished. The return of all rounds is added up and converted into Euro at the exchange rate of 0.05 Euro per CU. Payment is made anonymously.

Questionnaire

Before the experiment starts, you will be asked, via the computer screen, some questions on the rules of the experiment. If you do not understand a question, please ask the experimenter.

TABLE 1
Full Distribution of Ambiguity Urn Decisions
from Phase 1

 Frequency

Chose ambiguous urn for 5 Coded as "0" in
 both red and black at 2.5% binary variable
 40/60 "More Ambiguity
 Averse"
Chose ambiguous urn for 6
 both red and black at 3.0%
 50/50
Chose ambiguous urn for 3
 either red or black at 1.5%
 50/50
Switched to ambiguous 122
 urn at 40/60 61.9%
Switched to ambiguous 46 Coded as "1" in
 urn at 30/70 23.4% binary variable
 "More Ambiguity
 Averse"
Switched to ambiguous 3
 urn at 20/80 1.5%
Switched to ambiguous 0
 urn at 10/90 0.0%
Never chose ambiguous 5
 urn 2.5%
Unable to classify (i.e., 7
 switched back and 3.6%
 forth)
Total 197

Summary of Notation

a Worker's production share, 0 [less than or equal to] a [less
 than or equal to] 1
F Fixed wage, 0 [less than or equal to] F [less than or equal to]
 60
L Contract duration, 1 [less than or equal to] L [less than or
 equal to] number of remaining periods
M Market wage M [euro] 13, 14, ... , 30
Q Production quantity, 0 [less than or equal to] Q [less than or
 equal to] ...
GE Currency units
AN Worker
AG Employer

APPENDIX C: REGRESSION RESULT

TABLE A1
Logit Regression of Contract Acceptance

 1 2

Constant (a) -0.909 (0.459) -0.779 (0.428)
 0.047 0.069
Offered duration -0.461 (0.062) -0.467 (0.062)
 <.001 <.001
Fixed wage 0.163 (0.015) 0.162 (0.014)
 <.001 <.001
Revenue share 8.410 (0.705) 8.424 (0.705)
 <.001 <.001
New contract -0.752 (0.162) -0.770 (0.161)
 <.001 <.001
Market wage -0.143 (0.016) -0.143 (0.016)
 <.001 <.001
I 0.129 (0.202) 0.125 (0.202)
 0.523 0.537
[I.sub.+] 0.376 (0.232) 0.370 (0.231)
 0.104 0.110
F -0.059 (0.200) -0.064 (0.200)
 0.767 0.748
[F.sub.+] 0.030 (0.247) 0.028 (0.247)
 0.904 0.910
Period 0.019 (0.024) --
 0.429
Likelihood ratio 337.159 336.532
p value <.001 <.001

 3 4

Constant (a) -0.842 (0.439) -0.723 (0.408)
 0.055 0.077
Offered duration -0.468 (0.062) -0.473 (0.062)
 <.001 <.001
Fixed wage 0.162 (0.015) 0.161 (0.014)
 <.001 <.001
Revenue share 8.421 (0.707) 8.433 (0.707)
 <.001 <.001
New contract -0.762 (0.162) -0.779 (0.160)
 <.001 <.001
Market wage -0.141 (0.016) -0.141 (0.016)
 <.001 <.001
I -- --
[I.sub.+] -- --
F -- --
[F.sub.+] -- --
Period 0.018 (0.024) --
 0.456
Likelihood ratio 332.342 331.786
p value <.001 <.001

(a) The first entries in Table Al represent the [beta]-values of the
regression, the numbers in parentheses denote the standard deviation,
and the term below indicates the p value.

REFERENCES

Agell, J. "On the Benefits from Rigid Labor Markets: Norms, Market Failures, and Social Insurance." Economic Journal, 109, 1999, 143-64.

Alewell, D., C. Friedrich, W. Guth, and W. Kuklys. "Fair Wages and Multiple Fairness Standards--A Lab Study of Co-Employment of Hired and Rented Hands." Schmalenbach Business Review, 59, 2007, 2-28.

Anderhub, V., S. Gachter, and M. Konigstein. "Efficient Contracting and Fair Sharing in a Principal-Agent Experiment." Experimental Economics, 5, 2001, 5-25.

Berninghaus, S. K., S. Bleich, and W. Guth. "Wage Flexibility in Ongoing Employment Relations--An Experiment with a Stochastic Labor Market," in Arbeitsvertrage, Wirtschaftswissenschaftliches Seminar Ottobeuren, Vol. 37, edited by W. Franz, H. J. Ramser, and M. Stadler. Tubingen, Germany: Mohr Siebeck, 2008.

Bewley, T. F. "Why Not Cut Pay?" European Economic Review, 42, 1995, 459-90.

Bleich, S. "What Is Good Work? Different Aspects of Labor Contracts in Theory and Experiment." Ph.D. Thesis, University of Karlsruhe, Department of Economics, 2009.

Bolton, P., and M. Dewatripont. Contract Theory. Cambridge, MA: The MIT Press, 2005.

Briggs, C., and J. Buchanan. "Australian Labor Market Regulation: A Critical Assessment." Parliament of Australia, Parliament Library, 2000.

Brown, M., A. Falk, and E. Fehr. "Relational Contracts and the Nature of Market Interactions." Econometrica, 72, 2004, 747-80.

Esping-Andersen, G., and M. Regini. Why Deregulate Labour Markets? Oxford, UK: Oxford University Press, 2000.

Falk, A., S. Gachter, and J. Kovacs. "Intrinsic Motivation and Extrinsic Incentives in a Repeated Game with Incomplete Contracts." Journal of Economics Psychology, 20, 1999, 251-84.

Falk, A., and S. Gachter. "Reputation and Reciprocity--Consequences for the Labor Relation." Scandinavian Journal of Economics, 104, 2002, 1-26.

Fehr, E., G. Kirchsteiger, and A. Riedl. "Does Fairness prevent Market Clearing? An Experimental Investigation." Quarterly Journal of Economics, 108, 1993, 437-60.

Fehr, E., L. Goette, and C. Zehnder. "A Behavioral Account of the Labor Market: The Role of Fairness Concerns.'" Annual Review of Economics, 1, 2009, 355-84.

Gait, D. "Telecommunications Regulatory Structures in New Zealand." Presentation to International Telecommunications Society's Interconnection Workshop, Wellington, 10 April 1995.

Irlenbusch, B. "Discussion of Wage Flexibility in Ongoing Employment Relations--An Experiment with a Stochastic Labor Market," in Arbeitsvertrage, Wirtschaftswissenschaftliches Seminar Ottobeuren, Vol. 37, edited by W. Franz, H. J. Ramser, and M. Stadler. Tubingen, Germany: Mohr Siebeck, 2008.

Konigstein, M., and M. Villeval. "Efficiency and Behavioral Considerations in Labor Negotiations." Journal of Economic Psychology, 31, 2010, 599-611.

McCabe, P. G. "New Zealand: The Unique Experiment in Deregulation," in Telecommunications in the Pacific Basin: An Evolutionary Approach, edited by E. Noam, S. Komatsuzaki, and D. A. Conn. New York: Oxford University Press, 1994.

Sheskin, D. J. Parametric and Nonparametric Procedures. 3rd ed. Boca Raton, FL: Chapman and Hall, 2004.

Spiller, P. T., and C. G. Cardilli. "The Frontier of Telecommunications Deregulation: Small Countries Leading the Pack." The Journal of Economic Perspectives, 11, 1997, 127-38.

(1.) Similar problems can, of course, show up in love relationships (see Berninghaus, Bleich, and Guth 2008). Here, we will not discuss such analogies any further.

(2.) In our view, the "global economy" recently inspired an erosion of formerly evolved norms with firms threatening to dislocate their enterprises to countries with lower wages.

(3.) See, e.g., Galt (1995), McCabe (1994), and Spiller and Cardilli (1997) for the telecommunication industries, and Briggs and Buchanan (2000) and Esping-Andersen and Regini (2000) for labor markets.

(4.) In the experiment, an employer is free to offer a pure gift exchange, for example a contract with a 0-revenue share for the employee, which actually hardly ever occurs in the experiment, or provide working incentives.

(5.) Since in an experiment, the number of periods is limited, long-term contracts cannot be open-ended.

(6.) Wage rates are drawn in each period from a uniform probability distribution such that the [{[w.sup.c.sub.t]}.sub.t] are i.i.d, random variables.

(7.) Except for very rare instances like waiters in U.S. restaurants who sometimes must buy their tables, wages cannot be negative, although negative wages might alleviate the moral hazard problem of employment contracts.

(8.) Think about soccer professionals who would not be hired unless their contracts have expired.

(9.) In the case of only one pair in a session splitting up, this pair is of course rematched involuntarily--without, however, being aware of this (rematching).

(10.) As often in experimental economics we do not actually predict rational behavior but consider it as an important benchmark to compare and locate actual behavior.

(11.) For a detailed derivation of the results, see Bleich (2009).

(12.) For further analysis, see Bleich (2009) for the particular parameter values of our experimental design, only a two-period contract would be offered with [s.sup.F] + = 2/3 and [w.sup.F]+ = 0, with the fixed wage remaining at zero and the piece rate decreasing to one half for the rest of the contract's periods.

(13.) Although one faces only five potential candidates for the other role, a participant hardly ever met the same participant twice (without, of course, being informed about this) due to the frequency of constant pairs, either by contract or by mutual consent. Furthermore, when meeting again, the pair faces a different market wage than before.

(14.) The exchange rate is 0.05 Euro per 1 ECU (experimental currency units).

(15.) Table 1 is by no means complete (for instance, it allows for downward flexibility of [s.sup.j.sub.i] only when [w.sup.j] can be decreased as well). Nevertheless, we think that it contains the most relevant cases (salaries are usually more strictly regulated than additional incentives).

(16.) Thus the values in Table 2 represent the average values of the variables proposed in contract offers.

(17.) A rather simple idea is that direction of contract changes (improved or worse contract terms) parallels the development of the (increased or decreased) market wage. We could only confirm this, however, for treatments [I.sub.+] and F, where there is a positive correlation between changes in [w.sup.c] and changes in the workers' utility.

(18.) For a detailed statistical analysis see Section "Acceptance Decisions."

(19.) For a detailed statistical analysis see Section "Interdependence Between Degree of Regulation and Contract Duration."

(20.) [chi square]-test on the independence of moving directions of [increment of [w.sub.t]] and [increment of [w.sup.c.sub.t]] at 5% significance level.

(21.) For [F.sub.+], one has to add pure wage rate increases (10%), "both plus" (4%), although "both mixed" could also contain some wage rate increases.

(22.) For Dunn's method see Sheskin (2004).

(23.) These contracts give a worker just the period's market wage.

(24.) Here, we excluded effort choices leading to detrimental losses and reflecting highly irrational behavior.

(25.) We included a last contract period dummy to isolate endgame effects. To avoid collinearity problems with the last period dummy, we did not include contract duration.

(26.) We obtain the same result when deriving the slope of iso-profit curves of an employer.

(27.) Remember that the optimal effort [e.sup.*] = ps/c depends on the piece rate but not on the fixed wage.

(28.) See Result 2 and Result 3.

(29.) [chi square]-test at 5% significance level.

(30.) One could argue that only reverting F < I+ yields an alternative flexibility ordering.

(31.) There is a sharp decrease in efficiency from treatment I to treatment I+ which is difficult to explain. Employees might feel "disappointed" employers do not exercise the upward flexibility of piece rates and might lower their effort.

(32.) This result is based on the following calculations of average employment duration over the treatments:

Treatment N 1 [I.sub.+] F [F.sub.+]/Employment duration 1,621 2,253 2,188 2,392 2,299

(33.) By shifting the probability mass of the original uniform wage distribution equally onto larger w-values, the resulting distributions with larger [[w.bar].sup.c]-values dominate the preceding distributions with respect to first-order Stochastic Dominance.

TABLE 1
Overview of All Treatments

 Flexibility

Treatment [w.sup.j.sub.i] [s.sup.j.sub.i]

N No No
I [arrow up] No
[I.sub.+] [arrow up] [arrow up]
F [arrow up] [arrow down] No
[F.sub.+] [arrow up] [arrow down] [arrow up] [arrow down]

 Average Payoff (Euro)

Treatment Number of Sessions Employer Employee Both

N 6 18.14 18.11 18.12
I 11 14.98 20.13 17.03
[I.sub.+] 6 13.88 19.17 16.52
F 11 13.79 18.94 16.38
[F.sub.+] 6 18.55 18.71 18.63

TABLE 2
Average Experimental Results for All Treatments

 p Value N 1 [1.sub.+]

Market wage * <.001 21.712 22.119 21.890
Fixed wage * 0.002 13.039 11.449 9.667
Revenue share * 0.007 0.493 0.532 0.534
Accepted (%) 0.560 72.0 73.1 76.8
Offered duration * <.001 1.311 1.539 1.430
Effort * <.001 8.000 6.466 6.506

 F [F.sub.+]

Market wage * 20.917 20.096
Fixed wage * 12.075 11.718
Revenue share * 0.523 0.541
Accepted (%) 70.7 71.8
Offered duration * 1.737 1.973
Effort * 6.568 7.695

* Significant differences between treatments according to
a Kruskal- -Wallis test (at a 5% level of significance).

TABLE 3
Offered Contract Durations in Treatments

 N I [I.sub.+] F [F.sub.+]

Average 1.311 1.539 1.430 1.737 1.973
Average optimal (a) 5.498 5.468 5.536 5.637 5.878
Sign-Test on <0.001 <0.001 <0.001 <0.001 <0.001
 max. duration

(a) Because of effort-smoothing, each employer should in
each period offer a contract with maximum duration.

TABLE 4
Optimality of Contract Acceptance Rates

 N I [I.sub.+] F [F.sub.+]

Actual 72.0 73.1 76.8 70.7 71.8
 acceptance (%)
Optimal 87.5 88.8 89.9 69.2 62.8
 acceptance (%)
Sign-Test <0.001 <0.001 <0.001 0.245 0.011

TABLE 5
Effort Choice

 N I [I.sub.+] F [F.sub.+]

Actual effort 5.226 5.928 5.523 6.189 5.780
Optimal effort 5.119 5.373 5.353 5.468 5.499
Pearson Con. 0.833 0.589 0.797 0.590 0.707
P (Pearson) <0.001 <0.001 <0.001 <0.001 <0.001
P (Sign-Test) 0.360 <0.001 0.0013 <0.001 <0.001

TABLE 6
Multiple Linear Regression of Effort

 1 2 3

Constant 4.883 (1.836) 4.973 (1.529) 3.951 (1.388)
 0.008 0.001 0.004
Fixed wage 0.059 (0.028) 0.066 (0.027) 0.063 (0.026)
 0.037 0.013 0.016
Revenue share 8.184 (1.958) 8.434 (1.905) 8.193 (1.900)
 <.001 <.001 <.001
Period -0.398 (0.104) -0.407 (0.103) -0.412 (0.103)
 <.001 <.001 <.001
I -1.839 (0.925) -1.801 (0.921) --
 0.047 0.051
[I.sub.+] -1.560 (1.038) -1.520 (1.036) --
 0.133 0.143
F -1.565 (0.932) -1.515 (0.923) --
 0.093 0.101
[F.sub.+] -0.521 (1.056) -0.463 (1.036) --
 0.622 0.655
Market wage 0.027 (0.056) -- --
 0.629
Last contract -0.389 (0.655) -- --
 period dummy 0.553
[R.sup.2] 0.0252 0.0249 0.0215
Adjusted [R.sup.2] 0.0197 0.0206 0.0197

TABLE 7
Short- and Long-Term Contracts

Percentages N I [I.sub.+] F [F.sub.+]

Short-term offer (T = 1) 76.7 74.3 75.1 62.9 53.7
and conditional 77.2 77.5 80.3 78.9 78.2
 acceptance (a)
Long-term offer (T > 1) 23.3 25.7 24.9 37.1 46.3
and conditional 55.0 60.4 66.1 56.8 64.4
 acceptance
Renewed offer 22.2 30.1 34.2 30.8 18.6
Only one opts for renewal 52.1 41.5 41.8 39.1 50.0
The one is the employer 45.9 56.2 65.2 69.6 87.3
Employers opting for 58.9 74.9 78.5 79.5 74.5
 rematching

(a) Conditional acceptance is the percentage of accepted
short-term contracts.