Job creation and investment in imperfect financial and labor markets

When payroll variations are costly and financial constraints are tight, firms are strongly discouraged from hiring workers. Both financial and labor market imperfections restrict employment expansion. Which of these imperfections is more restrictive to employment expansion? The answer to this question will determine the job creation potential of relaxation of financial constraints compared to the removal of labor market frictions.

The theoretical framework of this article is a neoclassical dynamic model of investment on the lines of Jorgenson (1963), with adjustment costs to capital as in Lucas and Prescott (1971), extended to allow for liquidity constraints and bankruptcy, as well as labor hiring and firing decisions. This framework is suitable for evaluating the dynamics of capital, debt and two types of labor under the elimination of firing costs and the elimination of financial constraints.

The main finding of this article is that financial constraints are a bigger obstacle for job creation than labor market imperfections. While removing the labor market costly adjustment helps firms to create jobs and increase capital accumulation by releasing internal resources for investment, binding financing constraints hinder job creation. The main implication of this finding is that the elimination of financial constraints can play a very important role in designing employment policies as it can create more jobs than just removing firing costs.

Weighing the employment impact of financial imperfections with respect to labor market imperfections is a contribution of the present article. Most research on the labor market impacts of financial restrictions evidenced job losses produced by financial shocks or by productivity shocks amplified in imperfect financial markets. No research so far has shown the relative strength of each of these imperfections, which is very important to outline policy interventions that face obvious political constraints. Eliminating or reducing firing costs, a policy that is often derived from analyzes that remark the severity of labor market imperfections is a very sensitive issue in countries with strong labor unions as is the case in Western Europe, and thus may not be politically feasible. Accordingly, to enhance policy designs to consider financial imperfections together with labor market imperfections, it is crucial to have a clear understanding of the job creation potentials of these two binding restrictions.

The literature on the interaction between financial and labor market imperfections is broad and rapidly growing. Several empirical studies, as surveyed by Hubbard (1998) and Bernanke et al. (1999), established that a firm’s financial structure influences its investment decisions, thus rejecting the Modigliani and Miller (1958) proposition. Because external funds are more expensive than internal funds, the ability of a firm to adjust its capital is indeed sensitive to its internal collaterizable resources. On the other hand, investment also exhibits inaction and infrequent spikes, which are indicative of non-convex adjustment costs to capital (Caballero, 1999; Abel and Eberly, 1994; Cooper and Ejarque, 2004; Cooper and Haltiwanger, 2006). Similar research was conducted about labor adjustment costs, particularly hiring and firing costs, which were shown to reduce permanent employment, a particularly important topic in European countries (Bentolila and Bertola, 1990; Bentolila and Saint-Paul, 1992; Hopenhayn and Rogerson, 1993; Cabrales and Hopenhayn, 1997; Aguirregabiria and Alonso-Borrego, 2014).

The economic literature on the interaction of finance and labor markets grew substantially at the beginning of the 2000s and expanded to consider search and matching frictions besides firing costs. The common denominator of this research was to show both theoretically and empirically relatively large employment impacts of financial shocks and from the amplification of productivity shocks by financial constraints (Cantor, 1990; Sharpe, 1994; Nickell and Nicolitsas, 1999; Smolny and Winker, 1999; Yashiv, 2000; Pica, 2001; Acemoglu, 2001; Wasmer and Weil, 2004; Barlevy, 2003; Benito and Hernando, 2008; Caggese and Cuñat, 2008; Benmelech et al., 2011; Bakke and Whited, 2012; Petrosky-Nadeau and Wasmer, 2013; Caggese et al., 2019). Then, the Great Recession of 2007–2008 triggered a further expansion of this research, which corroborated that liquidity constraints are a powerful amplification mechanism of financial shocks with a large impact on labor demand. Following the Lehman crisis, banks were affected by huge losses that forced them to tighten lending to firms, which had to downsize their workforce. Firms with larger credit market frictions, expressed in more dependency on external finance or associated with less healthy lenders, were observed to experience larger employment drops (Chugh, 2013; Chodorow-Reich, 2014; Duygan-Bump et al., 2015; Yashiv, 2016; Giroud and Mueller, 2017). Calibrated and estimated full-blown dynamic models that allowed for interactions between financing and search, matching or nonconvex labor adjustment costs also show that a credit crunch causes larger reductions in employment (Buera et al., 2015; Michaels et al., 2019; Elsby et al., 2019).

Research conducted in Spain on labor and financial imperfections has arrived at similar conclusions as this literature. Spanish firms are shown to face significant financial constraints, so that financial variables have an important effect on firms’ investment, particularly on small firms (Alonso-Borrego and Bentolila, 1994; Estrada and Vallés, 1995). This research is also consensual that labor adjustment costs are large and exert a negative impact on employment (Bentolila and Saint-Paul, 1992; Bover et al., 2002; Aguirregabiria and Alonso-Borrego, 2014) and that tight liquidity constraint strongly limit the expansion of investment and permanent employment in Spain (Hernando and Tribó, 1999; Benito and Hernando, 2007, 2008; Barceló, 2007; Hernando and Martínez-Carrascal, 2008). In recent years, Bentolila et al. (2018) have found that during the Great Recession in Spain there were larger job losses at firms attached to fewer solvent banks than at firms related to stronger banks [¹].

This article performs a joint analysis of financial and real imperfections that affect both investment and hiring and firing decisions. This is formalized by a dynamic model in which firms decide on investment, labor and debt under financial constraints, bankruptcy conditions and adjustment costs to capital and labor. There are also two types of workers: One type is permanent, subject to firing costs and is more productive and the other type is temporary, hired under fixed-term contracts, which can be fired frictionlessly. These two types of labor are imperfect substitutes, but firms do generally prefer permanent and more productive workers. Temporary labor contracts allow firms to avoid hiring permanent workers who are costly to fire as needed, and thus serve as a relaxation of financial constraints, releasing resources to increase investment, reduce borrowing and hedge against adverse shocks. As firms accumulate capital, they also manage to build their team of permanent workers, substitute temporary workers and reduce their debt. Unlike debt, which can be reduced to zero, large firms do not totally substitute temporary workers for permanent workers because of their imperfect substitutability.

The data for this estimation comes from the Central de Balances del Banco de España (CBBE) (Balance Sheet data from the Bank of Spain) and includes financial variables, as well as information on permanent and temporary employment, from 1983 to 1996 [²]. This is a unique data set, particularly well suited to analyze the joint effects of credit constraints, firing costs and temporary employment on job creation. Several researchers also used it to analyze capital market imperfections, as well as labor market issues (Dolado and Bentolila, 1992; Bentolila and Dolado, 1993; Gómez and Dolado, 1995; Aguirregabiria and Alonso-Borrego, 2014).

The estimation consists of a simulated maximum likelihood procedure based on the policy rules of the theoretical model. The estimated model exhibits a good replication of the data. With the structural parameters, two counterfactual scenarios are evaluated. The first counterfactual scenario − the elimination of firing costs − initially reduces permanent employment substantially followed by a significant recovery, as well as capital increases and debt decreases. The second experiment − relaxing financial constraints − generates a sustained increase of permanent employment and capital together with a sharper decrease in firms’ debt. Noticeably, the level of permanent employment produced by a relaxation of financial constraints is considerably higher in the long-run (69%) than the one produced by the elimination of labor market imperfections (54%). Wage increases resulting from an expanded labor demand diminish these increases in permanent employment, yet in the long-run, the expansion of permanent employment is always larger when financial imperfections are removed than when firing costs are removed. These quantitative results may be probably seen as very large, yet they convey an important message for policy analysis: Instead of emphasizing the removal of firing costs, which may face strong social resistance, particularly by labor unions, relaxing financial constraints may be politically feasible and at the same time an effective way of addressing firing costs and expand permanent employment. Recent research corroborates the crucial connection between financial tightening and increasing unemployment. Furthermore, the recent financial crises, the Great Recession and even the recent pandemic illustrate that actual policies by central banks and governments recognize the importance of firms’ finance for job creation. Counter-cyclical policies to maintain and expand employment rely heavily on financial easing and expanding credit to firms rather than aiming at reducing firing costs.

The remainder of the article is organized as follows. Section 2 explains the model and characterizes the optimal solution. Section 3 discusses the estimation procedure and the sources of identification. Section 4 describes the data used in the estimation and documents their basic trends. Section 5 presents the results of the estimation and an assessment of how well the model fits the data. Section 6 evaluates the regime changes mentioned above. The main conclusions of this article are summarized in Section 7.

Firms maximize the expected discounted value of their stream of dividends $∑t=0∞EtDt(1+ρ)t$⁠, by choosing a sequence of investment I, permanent labor H, temporary labor L and debt B, at a discount rate ρ [³] Dividends are defined as:

that is revenues from products that depend on capital K and on two types of labor, permanent labor H and temporary labor L, net of investment, adjustment costs to capital and to permanent labor, both labor costs and net debt variation. The firm’s risky environment is contained in a random total factor productivity θ that follows a Markov process $P(θ′|θ)$⁠, parameterized as an AR(1) process: $ln⁡θ′∼N(μ+ϕln⁡θ,σ2)$⁠. Production technology is a Cobb–Douglas production function in capital and efficiency units of labor, with parameters α and β, respectively. Permanent and temporary labor is transformed into efficiency units of labor by a constant elasticity of substitution (CES) technology with parameters γ and λ, which reflect worker types’ different productivities. Flexible or temporary workers are unskilled and rigid or permanent workers are skilled; between them, there are no promotions or transitions.

Investment involves a quadratic adjustment cost $ck(K,K′)=a2I2K$⁠, where $I=K′−(1−δk)K$ and δ_k is the depreciation rate of capital. Wage rates for permanent and temporary labor are w_H and w_L, respectively. Firms do not have any way to alter these wages; they are wage-takers, which is motivated by a fully elastic labor supply or regulated wages [⁴]. The firm can hire and fire temporary workers and hire permanent workers at no cost, but their firing is costly: $ch(H−1,H)=Fmax⁡[(1−δh)H−1−H,0]$⁠, where F is the firing cost in terms of unit variations in permanent labor. Workers quit their jobs at an exogenous rate δ_h without producing any cost to firms [⁵].

Besides a labor market imperfection reflected in firing costs, there is a credit market imperfection that consists of an exogenous constraint on financing, formalized as a non-positive lower bound on dividends, $D≥−D¯$⁠. This constraint reflects that firms can use their own funds costlessly. Once these funds are exhausted, privately held or owner-operated firms do not have access to any external funds, while publicly-traded firms face an infinite cost of equity issuance [⁶].

In the current period, the firm pays debt B at interest rate r, determined both in the past period and contracts next period’s debt $B′$ at the interest rate $r′$⁠. The firm does not lend money in any way; that is, it is constrained to have a nonnegative level of debt: $B′≥0$⁠.

The timing of events is as follows: the firm enters a period with a level of capital K, debt $(1+r)B$ and permanent labor H₋₁, productivity θ is realized and the firm stays in business if its value is at least the value of an outside option and exits otherwise; the surviving firm chooses investment, new debt and the two types of labor; the exiting firm shuts down forever and defaults on the entirety of its debt [⁷].

The value of the firm is expressed in the following Bellman equation:

where $Ω(θ)$ represents the value of the outside option, which is defined as the value of starting a firm at that idiosyncratic productivity with zero capital, zero permanent workers and zero debt: $Ω(θ)=V(0+,0,0,θ)$⁠. In this environment, the value of the firm is increasing in capital and productivity, non-increasing in lagged permanent labor and decreasing in total debt payments, i.e. $VK>0,Vθ>0,VH−1≤0,V(1+r)B<0$⁠.

Let the reservation productivity, the lowest productivity that allows the firm to stay in business, be $θ¯(K,H−1,(1+r)B)≡{θ|V(K,H−1,(1+r)B,θ)=Ω(θ)}$⁠. Then the firm keeps operating if $θ≥θ¯$⁠, so that the probability of survival in the next period is $Pr⁡(θ′>θ¯′|θ)=1−Φ(σ−1(ln⁡θ¯′−γln⁡θ−μ))$⁠, where $Φ(.)$ is the normal cumulative distribution function. By the implicit function theorem applied to the definition of $θ¯$⁠, we obtain the following derivatives: $θ¯K′<0,θ¯H′≥0,θ¯(1+r′)B′′>0$⁠, which imply that the survival probability is increasing in the capital, non-increasing in permanent labor and decreasing in debt and the interest rate.

Firms and lenders establish a firm-specific debt contract by which lenders earn zero expected profits. Assuming that competitive lenders face an elastic supply of funds at the risk-free rate ρ, the interest rate $r′$ charged on debt $B′$ is determined by the condition:

The first term represents the expected return to the lender; that is, the probability of survival times the return on borrowing; the second term is the opportunity cost of the funds. This equation pins down the firm-specific interest rate as a supply function: $r′(K′,H,B′,θ)={r′|G(r′)=0}$⁠. The interest rate is decreasing in the capital, increasing in debt and ambiguous in permanent labor; more precisely, $rK′′<0,rB′′>0,rH′≥0,rθ′<0$⁠. The interest rate ranges between ρ if survival were guaranteed and infinity if the firm exists the industry with certainty, in which case B = 0, lenders deny any credit to that firm so that a Ponzi scheme is ruled out. The interested reader will find more details on the solution of the model and on the interest rate function in Appendix 3 [⁸].

As investment and permanent labor adjustments are limited by internal funds, the firm alleviates financial constraints using temporary workers and reducing debt. As the firm increases capital and builds a team of permanent workers, temporary employment and debt go down. Once the firm attains its desired level of capital and permanent workers, these two variables are no longer sensitive to the firm’s financial position.

Figure 1 illustrates the optimal solution for H as a function of H₋₁. In models of adjustment costs under unconstrained financial markets, firms with a level of permanent labor lower than H₀ immediately adjust to H₀, whereas firms with a level of permanent labor higher than H₁ immediately adjust to H₁. However, under financial constraints, small firms do not adjust their permanent labor to H₀, as they want to avoid costly firing in future periods of low productivity. Even in the absence of hiring costs, financial constraints make firms hire permanent workers slowly and rely relatively more on temporary workers. In their turn, firms with an excessive number of permanent workers have to keep them because financial constraints impede them in paying the firing cost associated with payroll downsizing. These firms can only afford to make slow reductions of permanent employment, that is, to a level above H₁.

Too few or too many permanent workers are both a liability and a sign that the firm’s net worth is low. Hence, firing costs paid currently or in the future, constrain persistently the level of permanent workers to the financial position of the firm.

The estimation method uses the policy rules of the dynamic model to construct a likelihood function at each parameter iteration (Hajivassiliou and Ruud, 1986; Keane and Wolpin, 1994). The estimated model’s parameters are the maximizers of this simulated likelihood function.

The estimation recovers the model’s parameters by minimizing the distance between observed and simulated trajectories of four observables: capital, debt, permanent employment and temporary employment. The simulation starts with the first observation after the year 1983 of these four variables and the first productivity level, which is recovered thereof. It is an estimation of a transitional process from one regime without temporary labor − in 1983 and 1984 − to one with temporary labor, established by an unanticipated change in 1985. Using the parameters of the productivity AR(1) process, for each firm, 100 simulated productivity sequences are generated. As it is unlikely that each of these simulations matches the observations exactly, it is assumed that observed values are measured with error.

Wage rates for permanent and temporary labor, w_H and w_L, are fixed, respectively, at the average and minimum wages per annum in Spain during the sample period: 2.58 for permanent labor and 0.89 for temporary labor [⁹]. Similarly, the annual risk-free interest rate ρ is set at 3.7% and the depreciation of capital δ_h at 15.6% per annum.

There are 11 estimated parameters of the model, which are identified from the four policy rules for the observable variables, characterized by nine possible regimes of permanent labor adjustment and possibly binding dividend and debt constraints [¹⁰].

The firing cost F is identified from the larger reductions in permanent labor, reductions in capital and the associated increases in debt, while the quit rate for permanent labor δ_h is identified from the lowest drops in permanent labor, understood by the model as inaction in labor adjustment. Similarly, the quadratic adjustment cost parameter a is identified from observed investment conditional on observed current capital. The financial constraint $D¯$ is identified from the amount of debt, as this constraint is binding when debt is positive, while when debt is zero, capital and permanent labor are insensitive to debt. Cobb–Douglas production function parameters, α and β, are mainly pinned down by the observed capital-labor ratios by firm size, while CES parameters for the two types of labor, γ, λ, are mainly determined by the observed mix of the two types of labor by firm size [¹¹]. Finally, the parameters of stochastic productivity, ϕ, μ, σ, are determined by the observed comovements of observed capital, the two types of labor and debt.

The estimated set of parameters is $Θ={α$⁠, β, γ, λ, δ_h, F, a, ϕ, μ, σ, $D¯$⁠, σ_K, σ_H, σ_L, $σB}$⁠; that is, the model’s parameters and the standard deviations of the measurement errors. The computation of the likelihood function exploits the discretization of the variables performed to solve the theoretical model (Appendix 5). The likelihood function is maximized using the Powell algorithm (Press et al., 1992).

The data come from balance sheet records kept at the Bank of Spain (CBBE). This data set contains 94,192 observations for more than 200 variables about the financial structure and employment of 19,473 firms from 1983 until 1996. The selection of the data leaves in the sample manufacturing private firms that do not change activity, do not merge or split and have more than five consecutive observations. The final sample consists of 1,217 firms with 10,787 observations. The data for capital and debt are given in millions of pesetas of 1987, computed using the industrial price index as a deflator. The employment information is given in terms of permanent and temporary workers, which in each category is the annual average number of workers hired by the firm. A further description of the selection of the data, the definition of the variables and the structure of the panel is provided in Appendix 2.

Table 1 presents descriptive statistics for the main variables in original amounts, ratios and variations. This table shows the values of the variables by size, measured as thirds in the distribution of capital and by period: before the labor market reform (1983‐1984), up to five years after the reform (1984‐1989) and 1990‐1996. This period is characterized by an important growth of capital, 3.3% by year, is the growth rate higher between 1985 and 1989 and a decline of debt [¹²]. Notice that in relative terms, debt by the worker is higher for medium-sized firms, whereas the debt-capital ratio is monotonically decreasing in firms’ size. As predicted by the model, firms with a high level of capital rely less on debt for their financial needs than small firms, which may be thirsty for financial resources.

In this same period, temporary labor substitutes permanent labor and, moreover, experiences a very high expansion, which is responsible for most of the expansion in total labor in the 1980s and 1990s. This is in line with Bentolila and Saint-Paul (1992) and Boeri and Garibaldi (2007), who documents an overshooting of employment or transitional “honeymoon” in the initial period after the regime change. Further evidence and discussion of these trends are provided in the next section.

Table 2 reports the maximum likelihood parameter estimates and their asymptotic standard errors. The capital coefficient is estimated at around 0.26, whereas the labor coefficient is around 0.65. These Cobb–Douglas parameters display decreasing returns to scale to capital and labor, as they add up to less than one. The value of 0.84 for γ, greater than β, indicates that substitutability between the two types of labor is predominant, however, imperfect. The estimate for λ shows that one unit of temporary labor is roughly around 20% as productive as one unit of permanent labor. In terms of profits, firms have a clear incentive to hire permanent workers, but, as seen below, their wages and especially their firing costs are relatively high.

The rate of quits of permanent labor of 0.2% is below its usual measures in job survey data. This value suggests that reductions in permanent labor result mainly from firms’ decisions. Firing costs are estimated to be 5.19, which doubles the annual wage. These firing costs do not just include observed severance payments, but all possible costs associated with the process of firing a worker, including the cost for the firm that workers appeal firing decisions and continue working. Capital adjustment costs are 2.39, which are relatively high, corresponding to slow adjustments in the capital.

The persistence parameter of the production process is 0.88, with a mean of 0.12 and a standard deviation of 0.28. The lower threshold on dividends is estimated at around 100 million pesetas of 1987, which may apparently indicate relatively loose liquidity constraints; however, it is important to bear in mind that adjustment costs both for apital and labor are relatively high, so these constraints are very likely to be binding [¹³]. Standard deviations of measurement errors are around the values of the standard deviations of the four variables explained in the descriptive section and are particularly low for capital and permanent labor. As we will see graphically and numerically, the model is able to fit the data fairly well.

Figure 2 reports the paths for actual and predicted average capital, debt and permanent and temporary labor by year. As explained above, the predicted variables in the first two years coincide with the actual because the estimation is made conditional on these observations coming from a regime without temporary labor. The model displays good replication of the data. The predicted path for debt fluctuates around the actual one, with some underprediction in later years, which looks clearer in Figure 2(c) which shows the debt-capital ratio over time. There is an increase in this ratio from 1983 until 1985 and from then onward a decrease. The predicted path for the two types of labor is also close to the actual path, with some overprediction for both types of labor in the past years. This trend is also clear in Figure 2(d) showing the actual and predicted percentage of temporary labor over the total labor force. These graphs illustrate the model is fairly successful in replicating the data.

Table 3 also shows that the model fit is fairly good, especially for capital and permanent and temporary labor. The sample standard deviations of the predicted errors of each variable are reported in the last row of each table; they are substantially lower than the estimated standard deviations of the measurement errors.

Having recovered the underlying parameters of the model and assessed their success in replicating the data, we can evaluate the impact on employment of two counterfactual scenarios:

1. removing labor imperfections; and

2. eliminating financial constraints.

As both regime changes increase the labor demand for the two types of labor, in the absence of labor supply data, we can also assess how their impact variates with a 10% wage increase and a 20% wage increase. The first scenario is attained by solving the dynamic programming problem using the estimated parameters, with the exception of firing costs, which are set to zero: F = 0 and increasing permanent and temporary workers’ wages, w_H and w_L, respectively, in the given percentages. The elimination of the dividend constraint is accomplished numerically setting $D¯$ at a very high absolute value while allowing wages of both workers to increase in the same percentages as in the previous scenario.

The predicted paths of the four variables are generated in a similar fashion as in the estimation, by a simulation of the sequence of productivities starting off with the true observed values of 1983 and 1984 and use the model’s policy rules of each counterfactual regime. The sequences of new predictions are reported in Table 4 and depicted in Figure 3 for removing labor market imperfections and Figure 4 for removing financial constraints, for 0%, 10% and 20% increases in wages.

The removal of firing costs reveals that firms have too many permanent workers, which they would like to eliminate, but cannot because of the high firing costs that this would represent. After its initial reduction, permanent labor increases again, as capital grows and debt decreases slowly to a positive level. When there are no wage increases, firms clearly substitute permanent labor for temporary labor, without much increase in capital or decrease in debt. While wage increases imply lower levels of permanent and temporary labor and lower drops of debt at all years, the response of capital is non-monotonic: Wage increases from 0% to 10% boost capital levels, suggesting a stronger substitution effect, but wages increase from 10% to 20% decrease capital levels suggesting that the scale effect dominates over the substitution effect.

The removal of the dividend constraints generates a large increase in capital with an immediate reduction of debt to zero. This is associated with increases both in permanent and temporary employment. That is, in this counterfactual scenario, firms do not downsize their permanent workforce suddenly as in the previous scenario. What clearly prevails is the complementary between capital and labor and between the two types of labor. These variations are particularly large if wages are fixed. Wage increases of 10% and 20% reduce this growth of capital and of both types of labor, especially of temporary labor, suggesting a clear dominance of the scale effect.

Table 4 reports specific values of the four variables predicted in these scenarios for three selected years. Removing financial imperfections clearly implies larger increases of permanent labor in 1996 than for removing firing costs, particularly when there are no wage increases: 54% increase of permanent labor for removing firing costs and 69% permanent labor increase for removing the dividend constraint. The long-run expansion of temporary employment is also larger for removing the dividend constraint, 91%, than for removing firing costs, 86%. Interestingly, increases in capital are in both cases 30%, while debt understandably disappears for removing dividend constraints, while it is reduced by 88% for removing firing costs.

Wage increases, particularly large increases, of 20%, erode the expansion of permanent employment substantially when the dividend constraint is removed, to 30% in 1996, which is still above the corresponding expansion of permanent employment of 28% when firing costs are removed. Removing the dividend constraint with wage increases of 20% imply much smaller long-run increases in capital and in temporary labor: only 16% for capital and only 9% for temporary labor in 1996, in contrast with 30% for capital and 14% for temporary labor when firing costs are removed. At any rate, even when accounting for increases in wages resulting from an increased labor demand, the long-run expansion of permanent employment is always larger when financial imperfections are removed than when firing costs are removed. There is no sudden large permanent employment drop when financial constraints are totally removed as it occurs when firing costs are totally removed.

This shows that removing financial constraints can create more permanent employment than removing firing costs. Access to external finance can activate imperfect labor markets without actually removing firing costs. Conversely, a financial tightening in the presence of large firing costs, as experienced in recent years, implies potentially large contractions in permanent employment.

The present article has analyzed the interaction between financial and labor market imperfections by means of a tractable estimated dynamic model of investment and labor demand. Capital accumulation and permanent job creation are restricted by a limited ability to use external finance and by the costly firing of permanent workers. Firms have to improve their internal financial position to increase their capital and adjust their number of permanent workers.

A comparative evaluation of two counterfactual scenarios reveals that relaxing financial constraints increases long-run permanent employment by 69% while removing firing costs increases long-run permanent employment by up to 54%. As these two scenarios imply both displacements of the labor demand, an upward-sloping labor supply will imply higher wages and, thus lower increases in permanent employment. Even when this effect is accounted for, increases in permanent employment are still larger under the removal of financial constraints than under the elimination of firing costs. With large wage increases of 20%, long-run permanent employment increases by 30% when financial constraints are removed and by 28% when firing costs are removed.

The present comparative assessment of the severity of financial constraints vis-à-vis the severity of labor imperfections has been carried out in a framework that is abstracted from features such as costly equity issuance, equilibrium interest rates, wage bargaining, search or matching frictions or promotions from temporary to permanent contracts. However, existing research in enhanced setups that formally incorporate the mentioned features unanimously underscores the negative effect of financial constraints on job creation, and thus the robustness of the main result of the present article on the job creation potential of removing financial constraints.

These relatively large quantitative outcomes are certainly the result of the drastic removal of liquidity constraints and firing costs. Real variations in finance tightness and firing costs may be less severe, so that observed outcomes may be actually more moderate. Yet, some policy implications emerge from the present analysis. The removal of firing costs or employment protections has been the cornerstone of labor policies in several Western European economies. Policy measures to reduce unemployment included reducing firing costs, lowering government intervention in wage determination and reducing unemployment transfers. In particular, most of the observed reforms did not attempt to reduce the costs of firing those already used, protected by strong unions, but to create a new type of contract that, once expired, allows firms to costlessly lay off newly hired workers. The result of these reforms was the emergence of dual labor markets consisting of permanent workers who are difficult to hire and especially difficult to fire and temporary workers, on probation for a fixed number of months, after which they are either promoted to be permanent or dismissed. Logically, these reforms created a strong incentive for firms to hire more temporary workers; however, the fact that firms in these economies not only operate in imperfect labor markets but also in imperfect financial markets, further limited the creation of permanent jobs to the extent of firms’ financial resources.

The present article has shown that removing firing costs, politically unfeasible in most Western European economies, would imply a sharp initial reduction of permanent employment followed by important increases thereof. Certainly, this initial reduction of employment will make such regime change even less acceptable to permanent workers. On the contrary, relaxing financial constraints would produce long-run larger capital and permanent employment expansions than the previous change, but without the initial mass layoff of permanent workers. This analysis suggests that policies designed to increase job creation will benefit from not limiting themselves to labor market policy measures but also consider financial variables and investment. Focusing on relaxing financial constraints may seem to be an indirect policy; nonetheless, it may probably be a more politically feasible and at the same time effective way to address labor market imperfections.

The present estimation and evaluation were performed using data for times of economic expansion, with only minor downturns compared to the Great Recession or the recent pandemic crisis, for which there is very visible evidence that limited finance limits job creation and reinforces job loss. Yet, we can see that even in times of sustained economic expansion, financing constraints and labor imperfections were substantial in hindering larger permanent employment increases. On the other hand, in terms of policy implications, most counter-cyclical policies carried out during these recent stress periods, both the Great Recession and even the pandemic, consisted of expanding credit issuance to the economy, with the specific purpose that firms do not fire workers. No policy in the world attempted to alleviate the firm’s financial position by reducing the cost of firing workers.

The author started this project when he was a Research Fellow at the Bank of Spain, before joining the Federal Reserve Bank of Philadelphia. He is thankful to the Bank of Spain for several fruitful discussions and for providing the dataset. He also thank participants of several seminars and conferences for useful comments and suggestions. Financial support of the Spanish Ministry of Science and Technology (Grant SEC 2001-0674) is gratefully acknowledged. Valeria Arza provided excellent research assistance at an early stage of this project. All errors and omissions are solely mine. The views expressed in this article are those of the author and not necessarily those of the Federal Reserve Bank of Philadelphia or The Federal Reserve System.

http://dx.doi.org/10.13140/RG.2.2.10422.63044