Entry and Exit of firms explained by trigger points: Dutch glasshouse horticulture. Natalia V. Goncharova and Arie J. Oskam

Transcription

1 Entry and Ext of frms explaned by trgger s: Dutch glasshouse hortculture Natala V. Goncharova and Are J. Oskam Paper prepared for presentaton at the 12 th EAAE Congress People, Food and Envronments: Global Trends and European Strateges, Gent (Belgum), August 2008 Copyrght 2008 by [Natala V. Goncharova and Ares J. Oskam]. All rghts reserved. Readers may make verbatm copes of ths document for noncommercal purposes by any means, provded that ths copyrght notce appears on all such copes.

2 Entry and Ext of frms explaned by trgger s: Dutch glasshouse hortculture Natala V. Goncharova 1,2 Are J. Oskam 2,* 1 Forts Bank, Utrecht / Brussels 2 Agrcultural Economcs and Rural Polcy, Wagenngen Unversty, The Netherlands * Correspondng and presentng author: Hollandseweg 1, 6706 KN Wagenngen, The Netherlands. Tel: ; E-mal: are.oskam@wur.nl Abstract The entry and ext decsons, consdered as nvestment decsons, are nvestgated n the paper. Takng nto account the heterogenety of entry and ext, the analyss s based on two types of entry-ext: real (related to the establshment or closng of a frm), or entry-ext n a new sector (ndcatng the dversfcaton or changng specalsaton). The theoretcal model s based on Marshallan trgger s wth Real Opton trgger s as an alternatve. The estmaton exploted the negatve bnomal model to nvestgate the role of trgger s (thresholds) on the observed number of entry or ext frms n Dutch glasshouse hortculture over 25 years. Frms should overcome dfferent thresholds dependng on types of entry and ext. Marshallan trgger s functon as good as the ones based on Real Opton theory. The estmaton of the model, whch takes nto account expected output prces, uncertanty and the nterest rate, however, provdes the best explanaton of entry and ext. That model can be consdered of a flexble varant of Real Opton theory. The model provdes plausble elastctes of entry and ext, ether real or n changng specalsaton. Keywords: entry and ext, trgger s, glasshouse hortculture 1. Introducton Entry and ext decsons of frms belong to the most nterestng, but also hghly ntrcate steps of ndvduals or frms. Moreover there are at least three 1

3 dfferent types of entry or ext decsons (Goncharova, 2007: ); each of them wth ther own dynamcs. Decsons about entry or ext are accompaned by nvestments that are lkely to be rreversble. These two dfferent decsons, whch are crucal for the frm, have profound mplcatons for economc growth. Entry and ext of frms can be benefcal for productvty growth, technologcal upgradng and employment generaton. Accordng to the OECD (2003), the entry and ext of frms accounts for 20-40% of total productvty growth n eght selected OECD countres. By consderng entry as an nvestment decson and ext as dsnvestment (=negatve nvestment), nvestment theory contrbutes to explan ndustry dynamcs. The economc lterature suggests dfferent theoretcal and emprcal approaches to explan choces of entry, ext and sze of frms ( for an overvew see, for example, Segfred and Evans, 1994). Ths paper s based on Marshall s model of long-run and short-run equlbra, whch assume that frms are nduced to enter f current revenue exceeds sunk costs ( Marshallan trgger ) and to ext f revenue falls below sunk costs. However, t s observed that frms sometmes prefer to delay an entry or ext decson, n the expectaton that prces and revenue (or costs) can change n the future. The real opton theory postulates that uncertanty wll affect the entry-ext nvestment decsons n such a way that t wll change trgger s In the model of Dxt (Dxt, 1989; Dxt, 1992), a wedge between the Marshallan trgger and observed trgger produces a zone of hysteress n whch frms do not respond to prce sgnals. The objectve of ths paper s to nvestgate whether nvestment trgger s contrbute to the explanaton of the number of enterng and extng frms for Dutch glasshouse hortculture. We try also to answer the queston whether trgger s based on real opton theory explan entry and ext behavour better than Marshallan trgger s. Dutch glasshouse hortculture can be charactersed as a dynamcally changng, hghly compettve, and captal ntensve sector. The evoluton and adaptaton of the sector to new technologes and market requrements are reflected n the process of frms entry and ext. 2

4 Secton 2 presents frst the theoretcal model and then the emprcal models of entry and ext. The negatve bnomal econometrc model s used for estmaton. Secton 3 dscusses the data, and provdes an analyss of changes n trgger s over tme as well as the comparson of dfferent types of trgger s. Secton 4 provdes estmaton results ndcatng the effect of trgger s on entry and ext. Fnally, Secton 5 closes a short dscusson, and some concludng and qualfyng remarks. 2. Modellng of entry and ext nvestment decsons 2.1. Theoretcal model The long-run compettve equlbrum s determned not only by the prce and output levels of the frms but also by the number of operatng frms. Followng MasCollel et al. ( 1995, p. 335 ) the central assumpton s: A frm wll enter the market f t can earn postve profts at the gong market prce and wll ext f t can make only negatve profts at any postve producton level gven ths prce. The long-run equlbrum prce (p*) equates demand wth long-run supply, where the long-run supply takes nto account frms entry and ext nvestment decsons. Consder an ndustry ntally n a long-run equlbrum poston, whch assumes number N 0 of operatng frms and long-run cost c (Fgure 1, a). Suppose that demand shfts upward, then the ndustry wll mmedately move to a new short-run equlbrum poston. The shock n demand causes an ncrease n prces to p S and the output per frm ncreases to q S ; ths can nfluence the nvestment decson of frms. Because frms profts ncrease, operatng frms earn more n the short-run (due to p S >c) and can even be nduced to make nvestments to expand; nactve frms can be nduced to nvest n entry. 3

5 p S p Intal Long-Run Equlbrum Short-Run Equlbrum New Long-Run Equlbrum p * = p c New Long-Run Equlbrum Intal Long-Run Equlbrum p * = c p S Short-Run Equlbrum a ) Entry N 0 q N 0 q S q N 1 Q b ) Ext q N 1 q S N 0 q N 0 Q Fgure 1: Impact of trgger s on Entry and Ext In the long run, frms enter n response to the ncrease n proft, wth the number of frms ncreasng to N 1 >N 0 ; the ndustry wll then move to the rght along a new demand curve untl t reaches the new long-run equlbrum. The graph (b) demonstrates the change n the number of frms as a result of the ext of frms as an adjustment to the new long-run equlbrum. In the long run, frms ext n response to the decrease n proft, wth the number of frms fallng to N 1 <N 0. Now, consder that a frm s proft-maxmsng nvestment decson s to enter or to reman nactve. A frm has to nvest a lump sum k and wll have a varable cost w for the producton of output. In the case of an ext decson, t must also pay a lump sum l, whch t loses due to the ext of the frm, and a varable cost w, whch wll be saved. The goal of the frm s to maxmse expected net present value (NPV). The standard Marshallan theory (Marshall, 1920) postulates that a frm wll nvest (and enter) f expected NPV s greater than zero, and n the case of an operatng frm a decson to ext wll be undertaken when NPV s negatve. Then for the entry nvestment of a frm, the trgger W H s Marshall s long run cost (when NPV>0), whch s the sum of the varable cost and the nterest on the sunk costs: W H w + ρk (1) where ρ s nterest rate. The Marshallan trgger for ext dsnvestment of a frm (NPV<0) becomes: 4

6 W L w ρl (2) Dxt (1989) and Dxt and Pndyck (1994) ntroduced a dscusson concernng a dfference between Marshallan trgger s and Real Opton trgger s. The dfference can be explaned by the presence of uncertanty that causes a frm to consder the opton of watng. Dxt (1989) provdes the followng relatonshps for the Real Opton entry P H and ext P L trgger s: P P > w + ρ k (3) H W H < w ρ l (4) L W L Dxt (1989) derves analytcally a closed form soluton for trgger s that take nto account uncertanty and the effect of changes n expectaton of 2 output prces ( µ ), uncertanty ( σ ) and nterest rate ( ρ ) on trgger s Emprcal model From equatons (1-2) we can numercally calculate Marshallan entry and ext thresholds. In the case of Entry, frms consder parameters of a potental sector to enter, consequently ρ s an average value ndcatng the current proftablty of the sector as perceved by a potental entrant. w, k are operatng costs n the frst year and the costs of captal; they represent the sunk costs of entrant frms. These ndvdual characterstcs of a frm are also mportant, because when the frm decdes on entry t takes nto account the level on whch t s gong to operate. In the case of Ext, ρ s the same as for entry frms, but w and l are operatng costs of the prevous year and rreversble costs of captal; ths represents sunk costs of the ext of frm j. To calculate losses l due to ext, we also nclude loss of proft because the frm no longer operates. The changes n the number of enterng or extng frms ndcate nvestment (or dsnvestment) decsons of frms. Accordng to the emprcal model represented n Equatons 5-6, we estmate the mpact of nvestment trgger s on entry (5) or ext (6) decsons: Entry γ + η t t t = 1, TR H, (5) Ext t γ 1 + η (6) =, jtr L, j t j 5

7 where TR, s the calculated threshold of a frm, that entered n tme t, and t H TR, s the calculated threshold of a frm j, that exted n tme t. t L j Marshallan trgger s ( W and H W L ) are calculated as shown n Equatons 1-2; Real Opton trgger s ( P H and P L ) are calculated as shown n Dxt (1989) and Goncharova (2007: 126). Addtonal varables, followng Real Opton theory, have an mpact on trgger s and percepton about the proftablty of the sector. They are the trend rates of growth of the market prce of output µ and ts varanceσ 2. t Entry s the number of frms enterng n the year t; t Ext s the number t of frms that were prevously observed to be n operaton n the year t. η - s an error term, a subscrpt ndcates an enterng frm, j ndcates an extng frm, and γ - s the parameter to be estmated. As a possble modfcaton of the model based on Marshallan trgger s, we nclude 2 µ, σ, ρ as addtonal varables n the Equatons 5-6, thereby assumng that these parameters have an mpact on the frm s decson concernng entry/ext, but ther mpact s more flexble than assumed by Real Opton theory Econometrc model Snce the dependent varable n the entry (ext) equaton s the number of frms enterng (extng), ths can take only nonnegatve nteger values. A count s understood as the number of tmes an event occurs. The ordnary least squares (OLS) method for even count data results n based, neffcent, and nconsstent estmates (Long, 1997). Thus, varous nonlnear models that are based on the Posson dstrbuton were developed for ths type of count data. The Posson regresson s y Posson ( µ ) (7) ~ µ = exp( x ) (8) for observed count y wth covarates for the -th observaton. The Posson model assumes that ts mean s equal to ts varance, whch s unlkely n realty. Ths leads to a problem of overdsperson,.e. that the 6

8 observed varance s greater than the mean ( var( y ) > E( y ) ). One reason for ths s the omsson of relevant explanatory varables. Estmates of a Posson model for overdspersed data are unbased, but neffcent wth standard errors based downward (Cameron and Trverd, 1998; Long, 1997). The most common alternatve s the Negatve Bnomal model, whch ntroduces an ndvdual, unobserved effect nto the condtonal mean. * y ~ Posson ( µ ) (9) * µ = exp( x β + u ) (10) u e ~ Gamma (1/ λ, λ) λ s the overdsperson parameter. The larger α s, the greater the overdsperson. If λ =0 then the model converges to the Posson model. A more detaled descrpton of the Posson model and the negatve bnomal model can be found n Cameron and Trverd (1998: 59) or Greene (2003: 744). 3. Data Ths secton frst gves a descrpton of the data used n estmaton and then presents an analyss of calculated trgger s, whch are used as ndependent varables n the model. We combne two data sets: FADN (Farm Accountancy Data Network) and Metellng data 1, provded by the LEI. Table 1 gves the varables used for estmatng thresholds, and for the econometrc specfcaton of the model. Metellng data provde us wth nformaton about all frms n the glasshouse hortculture sector, but also other sectors, durng the perod If a frm exted and entered durng these tme perods then we have the complete record of the frm s lfe : from brth to death. Although the coverage of glasshouse hortculture frms s good, the data content s farly small. Bascally, only the land and the numbers of employees are avalable. 1 Metellng s the Regster of Enterprses and Establshments of agrculture frms n the Netherlands. The regster covers all frms wth a sze equal to or bgger than 2 nge (Dutch Sze Unts). 7

9 Table 1: Descrptve Statstcs for Glasshouse Frms, Thresholds and Number of Entry and Ext Varable Descrpton of Varable Mean Standard Devaton Ha_tot Land per frm, ha Ha_glass Land under glass per frm, ha Proft_ha Proft per ha, 1000 Euros* Cost_mat_ha Materal cost per ha, 1000 Euros* Lab_tot Number of workers per frm, annual workers Cost_lab Labour cost per annual worker, 1000 Euros* Inv_ha Investments per ha, 1000 Euros* µ Trend rate of growth of output prces σ Standard devaton of output prces ρ Interest rate, % Entry K Ext K W H,K W L,K Number of enterng frms K=1 as real entry K=2 as entry n hortculture Number of extng frms K=1 as real ext K=2 as ext from hortculture Marshallan entry threshold, calculated for enterng frm, 1000 euros* K=1 as real entry K=2 as entry n hortculture Marshallan ext** threshold, calculated for extng frm, 1000 euros* K=1 as real ext K=2 as ext from hortculture * Monetary values are normalsed by 1985 prces ** Ext thresholds were used for estmaton as absolute values for the smplcty of the nterpretaton of results of the econometrc model The FADN s an unbalanced panel data set, amongst others, on glasshouse hortculture frms durng the perod Due to the rotaton of frms, frms stay n the sample for an average of 3-5 years. These data provde a wde range of ndvdual characterstcs of frms such as revenue, captal, nvestments, varable costs, whch we used for the estmaton of the annual level of these varables. For the calculaton of the trgger s, we used 8

10 varables from both data sets; however, due to the tme perod of FADN data, the further estmaton s lmted by the perod We dstngush and use for the analyss two dfferent types of entry and ext: 1) the genune (or real 2 ) entry and ext, 2) the entry and ext by changng specalsaton (e.g., when an exstng frm shfts to or from glasshouse hortculture producton). The varables represented n Table 1 are used for the calculaton of trgger thresholds 3. These varables characterse the average glasshouse frm, whch earns 59,000 euros proft through the use of 2.3 ha of land (0.6 ha under glass) and employs 3.4 workers per year. The average frm nvests Euros per ha n captal (such as land, glasshouses and nstallatons). The salent characterstc of Dutch glasshouse frms s that they reman small-scale famly frms (68.8% of famly labour) wth respect to labour and land, but they are hghly captal-ntensve, wth an average captal per frm of 383,000 euros (at 1985 prce levels). The next step, as an extenson of the conventonal approach, wll be to calculate Real Opton trgger s and compare them wth Marshallan ones. As can be seen, the nvestment thresholds (Table 2) vary over the years wth the common tendency of growth. The gap between Marshallan and Real opton trgger s vares and becomes bgger: f at the begnnng of the analysed perod the dfference for entry was about 5,000 euros and for ext about 2,000 euros, then at the end t had rsen to 30,000 and 14,000 euros respectvely. Followng the dscusson n Dxt (1989), the dfference between thresholds s caused by uncertanty. So the years wth the bggest gap, namely 1981, 1987, 1993, and 1996 possbly exhbt the effect of hysteress, when frms prefer to wat and would need to overcome a hgher threshold to make nvestments (n the case of entry) or dsnvestments (n the case of ext). It can be also noted that the dfference between entry trgger s s bgger than for ext trgger s; although n both cases the dfference between Marshallan and Real Opton thresholds s affected n the same years. 2 We use terms genune and real nterchangeably for the defnton of one of the types of entry or ext 3 A descrpton of the calculaton of trgger s by combnng of two data sets s provded by Goncharova (2007: ) 9

11 Table 2: Marshallan and Real Opton trgger s Year Real Entry Ponts, 1000 euros Marshallan Real Opton Real Ext Ponts, 1000 euros Hortculture Marshallan Ponts, 1000 euros MarshallanReal Opton Entry Ext Na Na 17.1 na na na Total s represent the annual average level - na not possble to calculate due to the absence of relable data on hortculture entry/ext An exstng frm that enters (exts) glasshouse hortculture has to overcome lower mpedments compared to the real entry (ext). Ths s demonstrated by the dfference n the nvestment trgger s: an exstng frm that enters the hortculture sector should nvest (on average, over the years) thousand euros, but for a real entry a frm should nvest almost twce as much, on average thousand euros. For the real ext, a frm should overcome (on average) losses of thousand euros, whch s three tmes the threshold for the ext from the hortculture sector (loss of 66.1 thousand Euros). 10

12 4. Results of estmaton econometrc models The change n the level of trgger s can encourage or dscourage ext and entry nto glasshouse hortculture, as s shown n Tables 3-4. These tables gve the negatve bnomal estmaton results for entry and ext. The results lend support to the negatve bnomal model, snce the λ parameter s sgnfcantly dfferent from zero. Ths s confrmed by the Lkelhood-rato test. The sgnfcance of overdsperson parameter λ confrms the presence of an ndvdual, unobserved effect that means non constant mean and varance n the data. By ths fact, the outperformng level of Log- Lkelhood for Negatve bnomal regresson over the Posson model can be explaned. The ext barrers were ncluded n the model as the postve values for the purpose of easer nterpretaton. The dfference among models s n the explanatory varables: Model 1 ncludes Marshallan trgger s, Model 2 ncludes Real Opton trgger s, whch are corrected for the effect of expectaton of prces, uncertanty, and nterest rate; and Model 3 explctly ncorporates the expectaton of prces, uncertanty and nterest rate n Model 1, that devates from the specfcaton of Dxt (1992). Based on Pseudo R2, t can be concluded that the Model 3 provdes the best explanaton of the varaton of entry and ext out of three specfatons. As can be seen from the estmaton results, a hgher level of entry thresholds has a negatve mpact on the number of frms that decde to enter. Increasng ext thresholds deters frms from extng the sector. In agreement wth the theory, postve expectatons about the trend of output prces nduce more frms to enter and fewer frms to cease operaton. 11

13 Table 3: Effect of Ponts on Real Entry and Ext Varable Dependent varable: Independent varables: TR Model 1 W H, *** (0.0004) Real Entry Model 2 P H,1 Entry1 Model 3 W H, ***-0.001*** (0.0004) (0.0006) µ * (6.776) σ (5.087) ρ 0.095** (0.046) Constant 5.372*** (0.203) λ (0.028) Lkelhoodrato test of λ = 0: Ch2(01) Log lkelhood: 5.371*** (0.203) (0.027) 3.402*** (1.142) (0.018) Model 1 W L, * (0.0006) 5.253*** (0.154) 0.034*** (0.011) Real Ext Model 2 Model 3 P L,1 Ext * (0.0006) 5.245*** (0.152) 0.034*** (0.011) W L, *** (0.001) *** (3.785) *** (2.554) (0.023) 7.405*** (0.629) (0.357) *** *** *** *** *** 70.52*** - Posson model Negatve bnomal regresson Pseudo R N ) estmated standard devatons n parentheses 2) *** denotes coeffcent sgnfcant at 1% level, ** at 5% and * at 10% level Hgher nterest rate, whch s an ndcator of the proftablty of a sector, has a postve connecton on entry, and a negatve one for ext (except a real 12

14 ext, whch s not sgnfcant). Uncertanty (σ ) has a postve (and not sgnfcant) result for real entry, but a negatve one for entry nto hortculture. Table 4: Effect of Ponts on Entry nto and Ext from Hortculture Varable Dependent varable: Independent varables: - TR - µ - σ - ρ Constant λ Lkelhoodraton test of λ = 0: Ch2(01) Log lkelhood: - Posson model - Negatve bnomal regresson Entry nto Hortculture Ext from Hortculture Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 W H, *** (0.0007) 7.639*** (0.148) 0.039*** (0.012) P H,2 Entry *** (0.0006) 7.626*** (0.144) 0.039*** (0.012) W H, (0.001) (4.682) (3.958) 0.049* (0.028) 6.433*** (0.813) (0.008) W L, *** (0.003) 6.600*** (0.210) 0.093*** (0.030) P L,2 Ext * (0.003) 6.597*** (0.205) 0.092*** (0.030) W L, *** (0.002) *** (7.048) *** (4.265) * (0.039) 9.070*** (1.186) (0.014) *** *** *** *** *** *** Pseudo R N

15 Ths can be explaned by the statement of Wennberg et al. (2007) that the negatve effect of uncertanty on the lkelhood of entry wll turn postve at a hgh level of uncertanty for real entry but not for the entry of exstng frms. Therefore the results can be understood as an ndcaton of hgher uncertanty for the real entry, compared to the entry nto hortculture. The hgher varaton of nput prces deters frms from exts; ths effect s larger for extng due to a change n specalsaton. Ths means that frms prefer to delay the decson to ext, because of expectatons of postve changes n prces. The presence of nvestment thresholds predetermnes a certan number of frms that are able to overcome these thresholds and that decde to nvest and enter (or to dsnvest and ext). Changes n nvestment thresholds affect frms and change ther behavour n such a way that an addtonal number of frms wll enter or ext. Ths effect of changes n trgger s can be demonstrated by analysng elastctes (Table 5). Table 5: Elastctes for trgger s after Negatve Bnomal Estmaton (Model 3 4 ) Real Entry Real Ext Entry n Hortculture Ext from Hortculture Dependent varable: En1 Ex1 En2 Ex2 Independent varable: TR H,1 TR L,1 TR H,2 TR L,2 - trgger W (0.11) (0.18) (0.82) (0.64) The establshment of a new frm can be expected f the real entry threshold decreases by 3,700 Euros. The real ext nvestment threshold should decrease by 1,900 Euros to nduce an addtonal frm to cease tradng. The dfference n elastctes demonstrates the fact that exstng frms respond more to changes n trgger s, because t s easer for these frms to overcome nvestment barrers. The changes n entry barrers should be bgger than for ext barrers to have an mpact on a frm s decson as can be seen from smaller values of elastctes for entry compared to ext thresholds. Another observaton from the table s that the exstng frms that enter or ext the hortculture sector are more senstve to the changes n nvestment 4 Model 3 s represented n Table 6, because, as s shown n Tables 4,5 and 7, Model 3 outperforms other specfcatons 14

16 thresholds. It can be expected that wth a 2,700 Euro decrease n the hortculture nvestment threshold (TR H,2 ), two more frms wll enter the hortculture sector, whle to encourage the establshment of the two addtonal frms the threshold (TR H,1 ) should decrease by 7,400 Euros. The same holds true for the ext: we can expect the ext from the hortculture sector of the two addtonal frms f the nvestment threshold (TR L,2 ) s bgger n absolute value by an amount of 1,000 euros; but for real ext TR L,1 should change by 3,800 euros. Table 6: Predcted and Actual mean of Number of Entry and Ext frms Number of Entry or Ext: Real Entry - actual (62.1) - predcted by: Model (46.1) Model (46.4) Model (46.9) Real Ext (73.6) (37.5) (36.7) (57.5) Entry nto Hortculture (143.5) (133.5) (129.8) (86.7) Ext from Hortculture (89.8) (56.7) (55.8) (51.3) By analysng the Table 6, we can compare how close the predcton can be compared to the actual average of events. It can be seen that real entry and ext events have closer predcted values than hortculture entry and ext. Ths can be related to the slower reacton to changes n nvestment thresholds, as dscussed above. As a comment to the dscusson about the real opton approach, we can see that the use of RO trgger s only slghtly mproves the predcton of entry and ext, whle assumng that characterstcs of the sector nfluence the frm s decson nstead of changng trgger s (Model 3) gves the most accurate predcton. The preference for Model 3 can be also supported by the dfferences n values of Log-lkelhood and Pseudo R2 provded n Tables

17 5. Dscusson, conclusons and further research We have examned emprcally the entry-ext process n Dutch glasshouse hortculture as an nvestment decson of a frm that should overcome an nvestment threshold. Clearly nvestment trggers barrers mpact on a frm s decson to nvest and enter, or to dsnvest and ext. An ncrease n the barrers dscourages frms from takng any acton; they prefer to delay the decson, whch s assocated wth rreversble nvestments. The models that nclude Marshallan and Real Opton trgger s were compared. The explctly calculated nvestment thresholds provde nsghts nto the barrers that a frm should overcome and show the ncrease of competton n the sector, partally due to the use of captal-ntensve technology n glasshouse hortculture. We dstngushed two types: real (or genune) entry-ext; glasshouse hortculture sector entry-ext. The heterogenety of entry and ext nvestments has two consequences. Frst, frms wll overcome dfferent thresholds that can nduce or deter frms from entry or ext. Second, the change n thresholds results n a dfferent number of enterng or extng frms, e.g. exstng frms whose specalzaton changes, resultng n them enterng hortculture are more senstve to the change n nvestment thresholds compared to frms, whch potentally can enter the sector and whch are consderng establshng a new busness. The dfference n degree of rreversblty of the dfferent types of entry and ext can be one of the reasons for ths. The mpact of thresholds can be a confrmaton of the effect of rreversblty on an nvestment decson: f a threshold (as a sum of operatonal and fxed costs) s possble to be reversed, a frm wll not take t nto account. The emprcal results do not provde reasonably strong support to real opton theory, whle the model that suggests the drect mpact of the sectorcharacterzng varables, such as expectaton of output prces, uncertanty and nterest rate, explans entry-ext decson better. The effect of these varables s larger for the real entry and ext compared to the change n specalzaton entry-ext. Moreover, uncertanty has a negatve mpact on ext and entry nto hortculture, but turns out to be postve for the real entry. One of the possble 16

18 suggestons, whch can be further explored n future research, s that for a hgher level of uncertanty, the negatve effect of uncertanty on the lkelhood of entry can turn postve. The elastctes of changes n the level of trgger s on the number of entres or exts shows hgher elastctes for exts then for entres and also hgher elastctes for shfts from and to hortculture than for real entres and exts: results whch are ntutvely very plausble. Further research can be conducted on deepenng the knowledge of the ndvdual frm s decson for entry and ext whch dfferentates the heterogenety of entry and ext. Thus t can have an mportant mpact on the length of survval of frms, and on ther post-entry performance. Investgatng ndvdual frms provde more opportuntes to reflect on results. References Cameron, A.C. and Trverd, P.K. (1998). Regresson Analyss of Count Data. Cambrdge: Cambrdge Unversty Press. Dxt, A. (1989). Entry and Ext Decsons under Uncertanty. The Journal of Poltcal Economy. No 97(3): Dxt, A. (1992). Investment and Hysteress. Journal of Economc Perspectves. No 6(1): Dxt, A.K. and Pndyck, R.S. (1994). Investment under Uncertanty. Prnceton: Prnceton Unversty Press. Gocharova, N.V. (2007) Investment patterns n Dutch glasshouse hortculture. PhD thess, Wagenngen Unversty. Greene, W.H. (2003). Econometrc Analyss. Upper Saddle Rver, New Jersey: Prentce Hall. Long, J.S. (1997). Regresson Models for Categorcal and Lmted Dependent Varables. Advanced Quanttatve Technques n the Socal Scences. Sage Publcatons. Marshall, A. (1920). Prncples of Economcs. Macmllan, New York. Mas-Colell, A., Whnston, M.D. and Green, J.R. (1995). Mcroeconomc Theory. Oxford Unversty Press, Oxford. 17

19 OECD (2003). The Sources of Economc Growth n OECD Countres. Pars, OECD. Segfred, J.J. and Evans, L.B. (1994). Emprcal Studes of Entry and Ext: A Survay of the Evdence. Revew of Industral Organzaton. No 9: Wennberg, K., Folta, T.B. and Delamr, F. (2007). Real Opton Model of Stepwse Entry nto Self-Employment. Proceedngs of the conference: Entrepreneurshp Research Conference, Unversty of Maryland. 18