Out of furniture business? Anne Strøm Prestvik The main research question : What are the factors that affect growth and demise of small scale producers? – The case of Central Java wood furniture industry. 1
Theory on growth and demise/exit Jovanovic : Total cost is related to the ability of the entrepreneur ∂ ψ = ψ η + ε > ( ). ( ) , 0 c c y ∂ η t t t η : entrepreneurial ability • In the discussion here, it is very unclear how ability is affecting cost and expectations of costs. • It appears to me that t ε is playing a role but I cannot see what it is. 2
Predictions of Jovanovic’s model Most of these make intuitive sense. But it is again very difficult to see how all these predictions are coming out of single (unified) underlying model. More explanation is required. 3
What about the role of fixed cost/capital? It is generally true that small developing country enterprises will have very little investment in fixed capital. If the outlook for current and future profitability is poor they will quit. The same will not be true larger enterprises. It would appear that the idea of fixed cost is implicit in Jovanovic’s model – is it? 4
Empirical model: Two models are estimated. 1. What determines the probability of survival/exit of furniture enterprises between 2005 and 2008? 2. What are the determinants of growth (- growth can be negative here -) in individual enterprises? 5
1. Exit probability: A Probit model - Firm-level factors (“idiosyncratic shocks”) - Sector-level factors (“correlated shocks”) - Economy-wide factors • It appears that only firm-level and some sector-level characteristics from 2005 are included in the model. • It is obvious that decision to exit is dynamic process. Should the events taking place during 2005 and 2008 be somehow incorporated in the model? 6
2. Predicting growth of enterprises: Growth between 2005 and 2008 • There can be a problem of selection here. I understand that the new entrants during this period are not in the dataset (?) • Are we looking at the growth of those enterprises that survived? • This will lead to the problem of sample selection bias. • One could run a two-stage (Heckman type) model where first stage would be the selection equation. 7
Recommend
More recommend
