Regulatory Barriers and Entry into a New Competitive Industry

Regulatory Barriers and Entry into a NewCompetitive Industry

John Bennett and Saul Estrin*

AbstractWe model the effects of license fees and bureaucratic delay on firm entry into a new competitive industry,whose profitability is initially unknown. A license fee alone reduces the number of first movers and thesteady-state number of firms. The combination of license fee and delay may cause some entrepreneurs topurchase licenses speculatively, only using them to enter production later if profitability is revealed to besufficiently favourable. Alternatively, some entrepreneurs may wait, possibly buying a license only afterprofitability is revealed; but it is never found that some entrepreneurs adopt one of these strategies andsome the other.

1. Introduction

Entry processes in advanced economies have commonly been modeled in terms of thecharacteristics of firms, such as plant size, age, and ownership type (Dunne et al.,1989). In developing economies, however, institutional weaknesses also have to betaken into account (Tybout, 2000), and entry rates have been found to be lower wheregovernments impose more onerous regulations on entry (Klapper et al., 2006). In thispaper we analyze how entry by new firms in a developing economy may be affectedby two forms of regulatory barrier: the imposition of a license fee for entry andbureaucratic delay in issuing the license once the fee has been paid. As would beexpected intuitively, both of these forms of barrier are found generally to have anadverse impact on entry, but our focus is on how the timing of entry is affected. Inparticular, we show how bureaucratic delays may lead to speculative behavior amongpotential entrants, with more entry licenses being purchased than may then be used.An implication is that data on new firm registration in economies where entry barriersare high may overstate reality.

To analyze these issues we formulate a model of the entry of firms into a newlyestablished market. It is assumed that, initially, there is uncertainty about the profit-ability of the industry; but, with a lag, entry by first movers reveals this profitability toall potential entrants. In the first version of the model each entrepreneur must decidewhether to be a first mover, sinking costs before the profitability of the industry isknown, or to wait, entering late if revealed profitability justifies the sinking of costs.

* Estrin: Management Department, London School of Economics, Houghton Street, London WC2A2AE,UK. Tel: +44-(0)20-7955-6629; Email: [email protected]. Bennett: Department of Economics and CEDI,Brunel University, Uxbridge, Middlesex, UB8 3PH, UK. We thank the Department for InternationalDevelopment for supporting this research under DFID/ESCOR project number R7844. Earlier versions ofthe paper were presented at the EBRD, London, February, 2005; the Latin American and Caribbean Eco-nomic Association Annual Conference, Paris, October, 2005; and at the ASSA meeting, Boston, January,2006. We are grateful to the participants, and particular to Dan Berkowitz and Hadi Esfahani, for theircomments. We also gratefully acknowledge the insightful reports of two anonymous referees, and com-ments and suggestions from Robin Burgess, Ravi Kanbur, Daniel Kaufmann, Mark Roberts, StefanoScarpetta and Kathy Terrell. The usual disclaimer applies.

Review of Development Economics, 17(4), 685–698, 2013DOI:10.1111/rode.12059

© 2013 John Wiley & Sons Ltd

mailto:[email protected]

However, to enter, an entrepreneur must purchase a nontradeable license, and weanalyze the effects this has on the pattern of entry (and exit). We find that both first-mover entry and the steady-state number of firms are weakly decreasing in the levelof the fee, as well as in other cost parameters and in the extent to which the future isdiscounted.

We then formulate a second version of the model including a delay between feepayment and receipt of a license. The license fee and the delay influence in differentways the balance of net advantages of first-mover entry compared with waiting. Inparticular, the existence of delay may generate the speculative purchase of licenses bysome entrepreneurs, buying early, before the profitability of early entrants isrevealed.1 Depending on the level of profitability then observed, some or all of thespeculators may enter without further delay, while any remaining speculators neverenter. We show that in equilibrium speculative purchase of licenses may occur, orthere may be late (non-speculative) purchase of licenses, but not both.

Unlike in the entry models of Jovanovic (1982) and Ericson and Pakes (1995), weassume entrepreneurs are ex ante identical: instead of learning about their individualabilities, they learn about the environment they all face. This also distinguishes ourformulation from models of stochastic competitive equilibrium where firms experi-ence idiosyncratic shocks (e.g. Hopenhayn, 1992). Our analysis is more closely relatedto literature on information revelation and strategic delay. In particular, Chamley andGale (1994) model investment by individual agents where information about the stateof nature is revealed through investment by others. However, they focus on the rolesof period length and the number of potential entrants; they do not allow for competi-tion between agents, which is at the heart of our model.

The trade-off between the potentially higher profits for first movers and the ben-efits to second movers of learning from the experience of first movers is emphasizedby Jovanovic and Lach (1989). In a continuous-time framework they show how com-petition generates S-shaped diffusion of an innovation. Also, Rob (1991) develops adynamic model in which entry by each firm marginally improves the informationavailable to late entrants, and he allows for earnings differentials owing to luck. Hefinds that equilibrium entry decreases monotonically over time. In these two modelsthe main concern is with the general properties of the equilibrium, whereas we formu-late our model with more specific assumptions in order gain some leverage on theissue of how licenses and delay may affect entry.

Our framework is conceived to apply to developing economies because entry andregulatory barriers are more onerous there than in developed ones (Djankov et al.,2002; Djankov, 2009). In Latin America and the Caribbean the average cost of entry(imposed by the government), including payment of license fees, is 36.6% of percapita income and the delay in being allowed to enter averages 61.7 days, while thecorresponding figures for OECD countries are 4.7% of per capita income and 13 days(World Bank, 2008). There is also significant cross-country heterogeneity in develop-ing economies. In Brazil, for example, entry costs are low (6.9% of per capita income)but delays high (120 days), while the opposite obtains in the Gambia (215% of percapita income but only 27 days of delay). We focus on the birth of a new industry toconcentrate attention on the new firm entry (and exit) process. New industry creationtakes a different form in a developing country because, rather than product or processinnovation, it commonly occurs through adaptation of foreign technologies to localcircumstances (Castellacci, 2011). Hausmann and Rodrik (2003) provide examplessuch as information technology in India, the garment industry in Bangladesh and thecut-flower industry in Colombia.

686 Saul Estrin and John Bennett


While our assumptions regarding license fees and entry delays in developing econo-mies can be easily justified, there is less direct evidence on the existence of specula-tion. We are aware from developed economies that licenses for entry can lead tospeculative purchases, e.g. in the auctioning of spectrum rights in US telecommunica-tions (Crampton et al., 1998), and licenses are more pervasive in developing econo-mies (Djankov et al., 2002) but documentation of speculation is poor. Nonetheless,there is considerable evidence of churning among new entrants, with simultaneouslyhigh rates of entry and exit. In several Sub-Saharan African and Latin Americancountries about 20% of new entrants enter and leave in the same year (World Bank,2013). Since “entry” here is characterized as registration to acquire of a license, while“exit” is measured as being economically inactive for a certain period, this evidence isconsistent with speculative behavior.

In section 2 we specify the characteristics of the industry. In section 3 we examineequilibrium in the model with and without license fees, and in section 4 we addbureaucratic delay. Section 5 discusses some implications, considers the roles of someof our assumptions, and suggests extensions for future research.

2. The Industry

Consider a new industry, with no incumbent firms at time t = 0. The supply of poten-tial entrepreneurs is assumed large relative to the number that actually enter theindustry. Entrepreneurs (and firms) are indexed i = 1, 2, . . . . Any entrepreneur mayinnovate, setting up a firm to enter and produce at t = 1.2 This requires a sunk cost kand payment of license fee f ≥ 0. If k and f are incurred at t = 1, the firm will need toemploy a unit of skilled labour in any period t so as to produce.

The output of any active firm i at time t is

y tti = = …θ, , , ,1 2 (1)

where θ is the realization, at t = 1, of a stochastic variable Θ which is uniformly distrib-uted with support 0 2, θ[ ]. Given that at least one entrepreneur sinks cost k at t = 1,the realization θ becomes common knowledge at t = 2. Θ captures the idea that,although the industry may exist in other countries, its suitability to local conditionsand institutions can only be discovered by experimentation; it represents uncertaintyrelated to the quality and reliability of inputs and their productivity under local condi-tions, including the institutional and organizational infrastructure. Note that θ is notfirm-specific. Unlike in Jovanovic (1982) or Ericson and Pakes (1995), entrepreneursdo not learn about their own abilities; rather, they learn about their environment.Apart from θ at t = 1, the values of all variables and parameters in the model arecommon knowledge.

We assume that profitability is limited by an increasing supply price of skilledlabor.3 Output demand is assumed perfectly elastic, with price fixed at unity, so yt

i canalso be interpreted as revenue. In effect, the industry produces a traded good in asmall open economy. The wage wt per unit of skilled labor at t is

w n tt t= + > =δ α δ α, , , , ,0 1 2 … (2)

where nt is the total number of firms in the industry at t. Throughout, we approximateby treating the number of firms as continuous.

Any number of entrepreneurs can enter at any time. For a first mover (an entrantat t = 1) θ is stochastic. For a potential second mover (an entrant at t = 2) the

REGULATORY BARRIERS 687


realization θ is known. No information becomes available after θ is revealed betweent = 1 and t = 2, so there is no reason to prefer entry or exit after the start of t = 2. Thus,for t ≥ 3, nt = n2, y yt

i i= 2, and wt = w2. Writing π ti and π t

j for the respective profits at tof a first mover i and a second mover j, we have

π θ δ α π θ δ απ θ δ α π θ

1 1 2

2 2

2 3iti

jtj

n k f n t

n k f

= − − − − = − − = …

= − − − − = −

; , , , ;

; δδ α− = …n t2 3 4, , , .(3)

Writing σ ∈ (0, 1) as a discount factor, the respective present values at t = 2 of theprofit stream of first mover i, if it stays in production, and second mover j are

V ni2 2

11

=−

− −( )σ

θ δ α , (4)

V n Kj2 2

11

=−

− −( ) −σ

θ δ α , (5)

where

K k f≡ + . (6)

In each case that follows, we examine the entry by first movers, the subsequentpattern of entry and exit, and the steady-state number of firms n* in the industry. n* isthe number of firms in the industry after uncertainty has been resolved and anyadjustments associated with license delay have played out.

3. Market Equilibrium

In this section we assume that although a license fee must be paid, there is no delay inreceiving it. As there is no uncertainty after θ is revealed (between t = 1 and t = 2), theequilibrium number of firms is the same at t = 3, 4, . . . as at t = 2. We therefore focuson the values of n1 and n2, and solve the model by backward induction from t = 2. Att = 2, each entrepreneur maximizes the present value of his or her profit stream, givenθ. Taking into account the behavior of each entrepreneur at t = 2 for all possible θ, att = 1 each entrepreneur maximizes the expected present value of his or her profitstream. We assume parameter values are such that there is positive entry at t = 1.

The value of n2 depends on the range of values within which θ falls. We distinguishcases (a)–(d) as follows:

Case (a) All n1 first movers would exit because even one of them alone in the indus-try would make a loss at t = 2. From (3), this occurs if

θ δ α θ< + ≡ a. (7)

If (7) holds, all first movers exit. Second movers have a cost disadvantage relative tofirst movers (having to incur a set-up cost) so there is no second-mover entry.

Case (b) Some of the n1 first movers remain in the industry at t = 2 (thus, θ ≥ θa) butsome exit. If all n1 first movers were to remain in the industry the n1th would make aloss, i.e. from (3),



θ δ α θ< + ≡ ( )n nb1 1 . (8)

Thus, in this case θ ∈ [θa,θb). Because, at t = 2, a potential second mover is at a costdisadvantage relative to any first mover, there is no entry by second movers. Thenumber of firms n2 is such that V i

2 0= , so that, from (4),

n a b2 0= − > ∈( ]θ δα

θ θ θfor , . (9)

Case (c) All first movers remain in the industry at t = 2 (thus, θ ≥ θb) but θ is notso great as to induce second mover-entry. From (5), ∂V nj

2 2 0< . Thus, if V j2 0≤ for

n2 = n1, there will be no second movers. This condition can be written

θ δ α σ θ≤ + + −( ) ≡ ( )n K nc1 11 . (10)

Provided θ ∈ [θb, θc), n2 is independent of parameter values.

Case (d) If θ θ θ∈( ]c, 2 the first movers will all remain in the industry at t = 2 andthere will be entry by second movers until V j

2 0= . Hence, from (5),

n K n2 21

1= − − −( )[ ] ≡α

θ δ σ ˆ . (11)

At t = 1 firm n1 is the marginal firm among those that enter. Writing π tn1 for its

profit at time t, and taking into account the four cases for possible draws of θ, theexpected present value of this firm’s profit stream is

V n n d n d n dn n n n

b

c

11

1 1 10

2

2 1 2 21

2 11 1 1( ) = ( ) +

−( ) + ( )∫ ∫θ

π θ σσ

π θ π θθ

θ

θˆ

θθ

θ

c

2

∫⎡⎣

⎤⎦{ }. (12)

Here, the term in parentheses after each π tn1 denotes the number of firms in the time

period considered. The first integral covers profit at t = 1, while the second and thirdintegrals relate to profit at t = 2 in cases (c) and (d) respectively. (If case (a) or (b)applied, firm n1 would exit, and so no term is specified.) For (12) to hold we assumethat 2θ θ> c, i.e. the realization of θ may be sufficiently large for second-mover entryto occur. For simplicity, we do not consider other cases. π1

1n .( ) and π 21n .( ) are given by

(3), while n̂2 is given by (11). Thus,

V n n k f K Knn

1 1 12 11

14

1 12

( ) = − − − −( ) − −( ) + − +( )θ δ αθ

σ σ σ δ αθ

. (13)

From (13), dV n dnn1 1 1

1 0( ) < . n1 adjusts such that, in equilibrium, V nn1 1

1 0( ) = , thesolution being n n f1 1= ( )ˆ , where

ˆ .n fK K

Kz1

22 4 12 2

2( ) = − −+( ) + −( )

+( ) ≡ − −θ δα

θ θ σ σα θ σ

θ δα

(14)

We assume parameter values are such that n̂ f1 0( ) > . Combined with our assump-tion that 2θ θ> c,4 using (10) and (14), this implies that



2 1θ δ α α σ− > > −( )z K. (15)

From (11) and (14), we can find, for case (d), how many second movers will enter.Denoting this number by ˆ ˆ ˆm f n f n f2 2 1( ) = ( ) − ( ), we obtain

ˆ , ;

ˆ

m fK K

K

m f

c2

2

4 2 12 2

2

0

( ) =−( ) + − −( )[ ]

+( ) ∈( ]

( ) =

θ θ θ σ θ σα θ σ

θ θ θfor

ffor θ θ≤ c.

(16)

Let n*m( f) denote the steady-state number of firms in the solution for license fee f.Since by t = 2 all uncertainty is resolved, n*m( f ) = n2. The value of n*m( f ) depends onwhich of cases (a)–(d) obtains.

Our first proposition gives the comparative statics of the model.

Proposition 1. n̂1 is decreasing in α, δ, k and f, and increasing in σ. In case (a) n*m = 0.In case (b) n*m is decreasing in α and δ, and independent of σ, k and f . In case (c)n nm* 1= ˆ . In case (d), n*m is decreasing in α, δ, k and f, and increasing in σ.

As we might expected intuitively, the initial number of entrants n̂1 is decreasing inall cost parameters. As costs k and f are incurred only at t = 1, a higher discountfactor σ reduces their relative importance in the present value of the expected profitstream and so is associated with a greater number of initial entrants.

The steady-state number of firms n*m is non-increasing in cost parameters, withresults depending on which cases (a)–(d) obtains. In case (a) there is no production insteady state. In case (b) some, but not all, first movers remain in the industry. Thenumber remaining is decreasing in cost parameters α and δ. However, these firms donot incur costs k and f at t = 2, 3, . . . , and, since they are intra-marginal first movers,marginal variation in k or f does not affect their entry decision at t = 1. Hence, theirnumber is independent of k and f. For similar reasons, their number is not affected bymarginal variation in the discount factor σ. In case (c) the steady-state number offirms n*m equals the number of first movers n̂1, comparative statics being as alreadydiscussed. In case (d) the steady-state number of firms n*m includes both first- andsecond-movers; n*m is decreasing in all cost parameters, and, as costs k and f are onlyincurred up-front when a firm enters, it is increasing in σ, as already explained withrespect to n̂1.

Setting f = 0, we obtain the laissez-faire solution, which is depicted in Figure 1,where period-1 entry is labeled nm

1 0( ) and the steady-state solution is labeled n*m(0).For at least one firm to stay in the industry we must have θ ≥ δ + α. nm

1 0( ) is inde-pendent of θ because it is determined before θ is revealed. The vertical distancebetween n*m(0) and nm

1 0( ) shows the amount of entry at t = 2. For low values of θ thisis negative, i.e. there is exit.

Figure 2 illustrates the effect of a positive license fee on the market solution. Thezero-license fee case is shown by the broken line, and the positive-license fee case isshown by the thick line. The existence of the license fee reduces the number of firstmovers, and so the horizontal section of n*m( f) is below that of n*m(0). Also, pro-vided θ θ> ( )b n̂1 , the license fee reduces the steady-state number of firms.

4. Delay and Speculation

In the formulation above, license fee f plays the same role as sunk cost k, but wenow assume a one-period delay between paying the fee and getting the license.



(We require that f > 0, for otherwise all firms buy a licence at t = 0.) Thus, first moverspay the fee at t = 0 and begin production at t = 1. As above, θ is observed betweent = 1 and t = 2. With this amendment to the model, one possibility is for other firms towait to observe θ and, if it is favorable, pay the fee f at the beginning of t = 2. Anyentry by second movers would occur at the beginning of t = 3, with the set-up cost kthen being incurred.

However, entry may take a different form. An entrepreneur may decide to “specu-late,” applying for a license, but not producing before θ is realized, and then onlygoing into production if θ is large enough. Thus, the entrepreneur may pay the fee atthe beginning of t = 1, and then either begin production at the start of t = 2 or notbegin production at all. We can therefore distinguish three types of entry: by firstmovers, by speculators and by “late movers” (entrepreneurs who wait to see the reali-zation θ before possibly paying f ).5

At t = 2 a first mover has a cost advantage k over a speculator, and so, if a specula-tor enters, we know that all first movers remain in the industry. At t = 2 and t = 3 aspeculator has the cost advantages of f and k over a late mover, and so, if any latemovers enter, we know that all speculators remain. To solve the model, we begin bydisregarding late movers entirely, considering only first movers and speculators. We

nt

n* m(0)

qa qb qc 2q q

n 1(0)

Ο

Figure 1. Laissez faire ( f = 0)

qa qb qc 2q qΟ

nt

n* m(0)

n* m( f )

Figure 2. A Positive License Fee



shall then show that in the solution we may have speculators, or we may have latemovers, but not both at the same time.

Suppose that s1 entrepreneurs buy licenses at t = 1. For any one of these specula-tors, h, entry at t = 2, yields profit

π θ δ απ θ δ α

2 2

3 3 4

h

th

n k

n t

= − − −= − − =

;

, , , ,…(17)

and, at t = 2, the present value of its profit stream is

V n k nh2 2 3

1= − − − +

−− −( )θ δ α σ

σθ δ α . (18)

The number of speculators that then enter at t = 2 depends on what realization of θoccurs. The realizations can be divided into three cases.

Case (Sa) If θ is low enough, V h2 0< for all h ∈ (n1, n1 + s1], so that none of the

speculators enter. Since, at t = 2, a speculator has a cost disadvantage relative to afirst-mover, the highest value of θ at which this case obtains is when all n1 first moverswould nonetheless remain in the industry. Hence, case (Sa) is defined by writingn2 = n3 = n1 in (18) and finding the values of θ for which V h

2 0< , i.e.,

θ δ α σ θ< + + −( ) ≡ ( )n k nSA1 11 . (19)

Case (Sb) In this range (19) is violated, and some, but not all, s1 speculators enter. Ifall speculators were to enter, the least efficient would make a loss in present valueterms; i.e. from (18),

θ δ α σ θ< + +( ) + −( ) ≡ +( )n s k n sSB1 1 1 11 . (20)

The number of firms adjusts such that in (18), V h2 0= for h = n2, i.e.

n k ns2 2

11= − − −( )[ ] ≡

αθ δ σ ˆ . (21)

Case (Sc) Here, all s1 speculators enter; i.e. (20) is violated.Moving back to t = 1, we now consider the payoff from speculation. Provided

2θ θ> SB, the expected present value for the marginal speculator (the s1th) is

V n s f n s n ss n s n s1 1 1 2 1 1 3 1 1

1 1 1 1 1

2 1+( ) = − + +( ) +

−+( )⎡

⎣⎢⎤⎦⎥

+ +σθ

π σσ

πθSSB

d2θ

θ∫ ,

The first term in [.] is profit including set-up cost k, while the second is the streamof discounted profits after the set-up cost has been incurred. Hence,

V n s fs SB1 1 1

21

1 2+( ) = − +

−( )−( )σ

θ σθ θ

. (22)

From (20) and (22), dV n s dss1 1 1 1

1 0+( ) < . s1 adjusts such that, in equilibrium,V n ss

1 1 11 0+( ) = ; that is, assuming an interior solution,



n s kf

1 1

121

2 1 21+ = − − −( ) − −( )⎡

⎣⎢⎤⎦⎥

⎧⎨⎪

⎩⎪

⎫⎬⎪

⎭⎪α

θ δ σ θ σσ

. (23)

The ranges of θ relevant to the behavior of a first mover follow immediately fromthe cases already specified in this and the previous section.

Case (Fa) All first movers exit (and no speculators enter). They do this if θ < θa, asspecified in (7).

Case (Fb) Some, but not all, first movers exit (and no speculators enter). This occursif θ ≥ θa, but θ < θb(n1), as specified in (8). The value of n1, and thus of θb(n1), will nowdiffer from that in the absence of delay. n2 is now given by (9).

Case (Fc) All first movers stay in production, but still no speculators enter. In thiscase θ ≥ θb(n1), but θ < θSA(n1), where θSA(n1) is given by (19).

Case (Fd) All first movers stay in production and some, but not all, speculatorsenter. This occurs if θ ≥ θSA(n1), but θ < θSB(n1), where θSB(n1) is given by (20). n2 isnow given by (21).

Case (Fe) All speculators enter. This happens if θ ≥ θSB(n1), the number of firmsn1 + s1 being given by (23).

Given these ranges, the present value, measured from t = 0, of the expected profitstream for the marginal first mover is

V n f n d

n d

n n

n n

b

SA

0 1 1 10

2

2

2 1 2

1 1

1

2

11

2

( ) = − + ( )

+−

( ) +

∫

∫

σθ

π θ

σσ θ

π θ π

θ

θ

θ11 1

2 2 1 12

ˆ ,n d n s ds n

SA

SB

SB( ) + +( )⎡

⎣⎤⎦∫ ∫θ π θ

θ

θ

θ

θ

where the profit equations (3) apply, but with f deleted. Using (20) and (23), it isfound that 2θ θ> SB, and so the final integral is valid. The first integral covers profit att = 1; the others cover profit at t = 2, 3, . . . for cases (Fc–Fe), i.e. when θ is largeenough for the marginal first mover to stay in production.

We thus find, after substituting from (23) to eliminate s1, that

V n f n k k

kn

n0 1 1

2 2

2 1

1 11

41

12

( ) = − −( ) + − − −( ) − −( )

+ − +( )σ σ θ δ α

θσ σ

σ δ αθ

.

(24)

From (24), dV n dnn0 1 1

1 0( ) < . n1 adjusts so that, for an interior solution, V nn0 1

1 0( ) = .The solution is n ns

1 1= ˆ , where, from (24),

ˆ .nk f k

ks1

22 4 1 12 2

= − −+( ) + −( )[ ] + −( )

+[ ]θ δα

θ θ σ σ σ σα θ σ

(25)

We now add the possibility of late entry into the model, while still allowing forspeculation. If entrepreneur j pays at t = 2 for a license in order to begin production att = 3, the present value of his or her profit stream is



V n k fj2 3

1=

−− −( ) − −σ

σθ δ α σ , (26)

where n3 is the number of firms at t = 3. This parallels equation (5), but incorporatesadditional discounting because of the time lag in receiving the license. If θ were to besufficiently high for such late entry to occur, an interior solution for n3 would be char-acterized by V n

23 0= , so that

n kf

31

1= − − −( ) +⎡⎣⎢

⎤⎦⎥{ }α

θ δ σσ

. (27)

However, n3 is the total number of firms in this case; i.e. it includes first movers andspeculators, as well as late movers. Now compare (27) with (21), which gives the totalnumber of firms in case (Sb), i.e. when some but not all speculators enter. Because ofthe appearance of the term f/σ, n3 in (27) is less than n2 in (21), but when there is anylate entry, our assumptions imply that all first movers have stayed in and all specula-tors have entered. Thus, the total number of first movers and speculators must be atleast as large as in (21). We have a contradiction: (27) could only apply if the numberof late movers were negative. This yields the following result.

Lemma 1. When there is speculation there is no late entry.

By speculating, a firm can produce before late movers do. Thus, speculators exploit allbusiness opportunities, making entry by late movers unprofitable.

Now consider whether in the absence of speculation there may be late entry. From(23), for any value of n1 a corresponding value for s1 is found. In the solution, n ns

1 1= ˆ ,as given by (25). Suppose that the interior solution for s1 found this way is s1 = 0. Thenthere would be entry by late movers if n3, as given by (27), exceeds n1, but, once wereach t = 3, if late-mover entry is to occur at all, it would occur if the realization of θwere at the maximum value 2θ . Assume this realization occurs. Then late moverentry occurs if n3 as given by (27) exceeds n1 as given by (23) with s1 = 0. This inequal-ity reduces to

θ σσ

> −( )18

f. (28)

If we can show that there exist parameter values for which the interior solutions1 = 0 is obtained, and also (28) holds, then we have shown that late mover entry mayoccur when speculation does not. This interior solution holds if n̂ n ss

1 1 1= + whens1 = 0. From (23) and (25), this condition reduces to

θ σ σ=−( )

±( )kg2 1

, (29)

where g = [(1 − σ)h/σ]1/2 and h f= θ . For the values of θ in (29) to be positive, wemust have g > 1, i.e. 1−( ) >σ σ θf . In conjunction with (28), this gives

Lemma 2. An interior solution s1 = 0 holds and there is late entry if θ = 2Θ and

1 18

−( )> > −( )σ

σθ σ

σf f

.



Ex ante prospects can be good enough to induce entry by some first movers, but insuf-ficiently good to induce other firms to purchase a license speculatively. Nonetheless,ex post, when the value of θ is realized, the industry may be profitable enough toattract late-mover entry. Under these conditions it is partly the prospect of latemover-entry that inhibits the initial speculative purchase of licenses.

Putting Lemmas 1 and 2 together, the following proposition obtains.

Proposition 2. With a positive license fee, if it is necessary to wait to be granted alicense, speculative purchase of licenses may occur, or there may be late (non-speculative) purchase of licenses, but not both.

Up-front payment of the license fee by speculators can give a strategic advantage overlate entrants that is so strong that all late entry is squeezed out. If it is not this strong,then speculation is eschewed by all entrepreneurs, with late entry then taking place ifthe realization of θ is sufficiently large. In the latter case, even though prospects, exante, are on average so poor as to prevent speculation, a realization of θ may thenoccur that favorable enough to induce late entry.

It might be supposed that bureaucratic delay would decrease the number of firms insteady state, for it increases entry costs for all entrants. However, delay imposes someexpected costs on late movers and speculators that are not imposed on first movers.Whereas first movers must wait only one period before entering, late movers suffer atwo-period delay before they may enter, meanwhile forgoing any potential profits;and while first movers pay for a license fee that they then use for entry and potentiallyto earn profits, speculators pay a fee for a license that they may not use. These costshave a negative effect on entry by late movers/speculators, and this boosts theexpected profits, and numbers, of first movers. The interplay of these factors preventsus from coming to clear-cut conclusions regarding the effect of delay on the steady-state number of firms.

If there is late entry (with no speculation) the pattern of entry is broadly similarto that shown in Figure 2. If there is speculation (but no late entry) the pattern ofentry is shown in Figure 3, where n fs

1 ( ) is period-1 entry in the delay-case. Thisdiffers significantly from the previous figures in that, for the highest range of θ(θ θ θ∈ ( )[ ]( )SB n1 2, ) the line depicting the steady-state number of firms n* is horizon-tal. This is because, for such high θ all the entrepreneurs that have speculatively

qa qb qSA qSB 2q qΟ

nt

n* s( f )n 1s( f )

Figure 3. Speculation



bought a license will enter, and, as we have seen, the solution to the model is such thatthere is no entry by late movers.

5. Concluding Comments

We have analyzed how license fees and bureaucratic delays affect entry into a newindustry in a developing economy. License fees alone have a negative effect on entryand may lead to later entry by some firms. This may exacerbate the deficiencies in theprocess of creative destruction that are emphasized by the World Bank in its latestWorld Development Report.6 We also show that bureaucratic delays may result insome entrepreneurs purchasing licenses speculatively, only using them to enter pro-duction late if the profitability of the industry is revealed to be sufficiently favorable.In equilibrium, some entrepreneurs may speculate or some may wait, not decidingwhether to purchase a license until after profitability is revealed; but it is never foundthat some entrepreneurs adopt one of these strategies and some the other. This sug-gests that official statistics about new firm entry, which derive from the new companyregistration data, may over-estimate actual entry.7 With currently available data, it isnot possible to explore this proposition empirically because there is no publicly avail-able information tracing economic activity post-registration. To build such a data setwould require a major collection effort, but if we are correct in predicting that entryrates will be over-estimated in weak institutional environments, this may be a valu-able research activity.

Two of our simplifying assumptions bear further comment. The first, the uniformdistribution of Θ, is a common simplification in related literature (e.g. Hausmann andRodrik, 2003; Hausmann et al., 2007). In our model we would not expect the use ofother distributions to affect Proposition 1 significantly, provided the analysis is trac-table. However, we conjecture that Proposition 2 might not hold. Suppose, e.g. that adistribution has a small probability of very high values of θ, above 2θ in our model,the mean still being θ . Then average prospects in the industry are unaffected, butnow a very high realization θ is possible. This favors potential late entry, rather thanspeculation. We conjecture, therefore, that late entry may occur in this case even inthe presence of speculation.

Second, consider the assumption of linearity of the wage equation. Suppose, e.g.that we amend the wage equation (2) so that dwt/dnt is increasing in nt. In the dia-grams, the upward-sloping lines then diminish in slope as θ rises, for the entry of eachfirm drives up costs for the next firm. Although the parametric specification isaffected, the general nature of Proposition 1 still obtains, though the implications forour speculation/delay result are unclear.

A weakness of our analysis is that it does not address the issue of why the gov-ernment requires licenses for entry and delays issuing them. In principle, the gov-ernment might be welfare-maximizing, erecting barriers to entry to limit theduplication of fixed costs, and our could be extended to explore this issue (Bennettand Estrin, 2006). However, as stressed by public choice theory, the rationale forimposing entry barriers may instead be to generate private benefits to politiciansand the bureaucracy (Djankov et al., 2002). The motivation for introducing licensingmight be in order to refuse to issue a license, or to delay its issue further, unless abribe is given (Shleifer and Vishny, 1993). Thus, the license fee f in our model mightbe interpreted as including a bribe component. With this interpretation, the modelmight be developed to explore the interplay between politicians’ desire to generate



bribes, exhibited through endogenously determined levels of the license fee and theassociated delay, and entrepreneurial entry behavior. The possibility that, after theyhave entered, first movers would bribe politicians to raise the license fee that otherentrants would then have to pay, could also be examined.

References

Bennett, John and Saul Estrin, “Regulatory Barriers and Entry in Developing Economies,”CEDI working paper 06-02, Brunel University (2006).

Bhaumik, Sumon Kumar, Shubhashis Gangopadhyay, and Shagun Krishnan, “Reforms andEntry: Some Evidence from the Indian Manufacturing Sector,” Review of Development Eco-nomics 13 (2009):658–72.

Campos, Nauro and Saul Estrin, “Entry and Barriers to Entry in Emerging Markets,” Eco-nomic Systems, 31 (2008):343–45.

Castellacci, Fulvio, “Closing the Technology Gap?” Review of Development Economics 15(2011):180–97.

Chamley, Christophe and Douglas Gale, “Information Revelation and Strategic Delay in aModel of Investment,” Econometrica 62 (1994):1065–85.

Crampton, Peter, Evan Kweral and John Williams, “FCC Spectrum Auctions in the US,”Journal of Law and Economics 41 (1998):647–75.

Djankov, Simeon, “The Regulation of Entry: A Survey,” World Bank Research Observer 24(2009):183–203.

Djankov, S., Rafael La Porta, Florenzio Lopez-de-Silanes, and Andrei Shleifer, “The Regula-tion of Entry,” Quarterly Journal of Economics 117 (2002):1–35.

Dunne, Timothy, Mark Roberts, and Larry Samuelson, “The Growth and Failure of US Manu-facturing Plants,” Quarterly Journal of Economics 104 (1989):671–98.

Ericson, Richard and Ariel Pakes, “Markov-Perfect Industry Dynamics: a Framework forEmpirical Work,” Review of Economic Studies 62 (1995):53–82.

Estrin, Saul and Tomasz Mickiewicz, “Shadow Economy and Entrepreneurial Entry,” Reviewof Development Economics 16 (2012):559–78.

Hausmann, Ricardo and Dani Rodrik, “Economic Development as Self Discovery,” Journal ofDevelopment Economics 72 (2003):603–33.

Hausmann, Ricardo, Jason Hwang, and Dani Rodrik, “What You Export Matters,” Journal ofEconomic Growth 12 (2007):1–25.

Hopenhayn, Hugo A., “Entry, Exit, and Firm Dynamics in Long Run Equilibrium,”Econometrica 60 (1992):1127–50.

Jovanovic, Boyan, “Selection and the Evolution of Industry,” Econometrica 50 (1982):649–70.

Jovanovic, Boyan and Saul Lach, “Entry, Exit, and Diffusion with Learning by Doing,” Ameri-can Economic Review 79 (1989):690–99.

Klapper, Leora F., Luc Laeven, and Raghuram G. Rajan, “Entry Regulation as a Barrier toEntrepreneurship,” Journal of Financial Economics 82 (2006):591–629.

Rob, Rafael, “Learning and Capacity Expansion under Demand Uncertainty,” Review of Eco-nomic Studies 58 (1991):655–75.

Roberts, Mark J., and James R. Tybout (eds), Industrial Evolution in Developing Countries:Micro Patterns of Turnover, Productivity, and Market Structure, Oxford: Oxford UniversityPress (1996).

Shleifer, Andrei and Robert W. Vishny, “Corruption,” Quarterly Journal of Economics 108(1993):599–617.

Tybout, James R., “Manufacturing Firms in Developing Countries: How Well Do They Do,and Why?” Journal of Economic Literature 38 (2000):11–44.

World Bank, Doing Business 2008, Washington, DC: World Bank (2008).———, World Development Report: Jobs, Washington, DC: World Bank (2013).



Notes

1. Since the licenses are non-tradeable, the down-side of a speculative purchase is that thelicense may be unused. If licenses were tradeable, they might be purchased speculatively forpotential later resale.2. At any time t there are more potential entrants than the market can sustain. We assume asmall exogenous asymmetry between firms limits entry to the point at which the present valueof the expected profit stream of the marginal entrant is zero. We need not assume that entry upto this point is sequential; but we speak in terms of sequential entry for simplicity.3. The effect of skilled labor supply on entry in India, post-1991, is emphasized by Bhaumiket al. (2009).4. The two assumptions are required for the number of first movers to be positive (otherwisethere is no industry) and for the possibility that profitability can turn out sufficiently high forsecond movers to enter.5. Note that there is no such scope for speculation in the first version of the model, where thereis no delay between purchase of a license and entry. It would not be rational for a firm to specu-late by purchasing a license at t = 1 because, instead, it can just wait to observe the realizationof θ between t = 1 and t = 2 and then, if θ is sufficiently favorable, buy the license and produceimmediately.6. “The wide dispersion of productivity among businesses, the large number of unsustainablemicro-enterprises and the stagnation of larger firms all suggest that the process of market selec-tion and creative destruction . . . is weak in most developing countries” (World Bank, 2013,p. 112).7. Although our analysis suggests that entry rates might be lower in developing than developedeconomies because regulatory barriers are higher, the available evidence suggests quite highrates of entry in developing economies (Roberts and Tybout, 1996; Campos and Estrin, 2008).The contradiction may be explained by this potential problem with the data. High entry rates indeveloping economies may also be a consequence of a larger shadow economy (Estrin andMickiewicz, 2012).



Documents

Regulatory Barriers and Entry into a New Competitive Industry