This document analyzes the effects of rushed privatizations on firm performance using data from Poland from 1995-2005. It finds:
1) Firms privatized during spikes or "rushes" in privatization in 1997 and 2001 experienced larger declines in total factor productivity (TFP) and labor productivity compared to non-rush privatized firms, supporting the hypothesis that rushing privatizations can overwhelm government capacity.
2) Event study analyses show rushed privatized firms experienced productivity declines beginning 2 years before privatization and continuing for at least 5 years after compared to state-owned or private firms.
3) Overall, the results provide evidence that rushing privatizations has negative effects on firm productivity and performance, which is
1. Are Rushed Privatizations Substandard?
Analyzing Privatization under Fiscal Pressure
Jan Hagemejer
Jan Svejnar
Joanna Tyrowicz
EEA, New York
February 26, 2017
2. Background
• Normal flow of privatizations can be handled by governments
• What is the outcome when a large number of firms needs to be
privatized fast?
• Focusing minds v. overwhelming government’s capacity (Gupta et al., 2008)?
• Important for policy (national governments, IMF programs, EU)
• Greece, Portugal, transition economies
• Government revenue v. efficiency of firms?
3. What do we do?
• Examine effects of “rushed” privatization on firm performance
• Use firm-level data from Poland during 1995-2005
• Annual panel data on almost 46,000 manufacturing, service, mining,
and utility firms
• Focus on total factor productivity (TFP) and labor productivity
• Compare privatized companies to both non-privatized (SOE) and
private firms
4. Insights from literature
• Earlier studies -- privatized firms outperform SOEs (usually not
controlling for endogeneity/selection and causal effect likely
overstated)
• Firms may react strategically to expecting privatization
• Buyers and SOEs select each other
• Unsuccessful privatizations may not be detected in the data
5. Insights from literature (2)
• Recent studies – privatization to foreign owners improves efficiency,
while privatization to domestic owners does not and may reduce it
(controlling for endogeneity/selection)
• Budgetary constraints conducive to divesting (Greece 2013, Portugal)
6. Our data
• Firm census data collected by Poland’s Statistical Office
• All manufacturing, service, mining, and utilities firms with 50 or more
employees (full time equivalent)
• Information on income statements, balance sheets, employment,
ownership, industry, location, …
• We have constructed a panel data set showing changes in ownership
• Majority/minority state, domestic private or foreign owned
• Data anonymized => cannot be directly compared to Treasury data on
publicly traded firms
7. Spikes in privatization
• Treasury and CSO firm census data indicate that there were spikes
(rushes) in privatization in 1997 and (in terms of revenue from
privatization) also in 2001
• We use these years to construct a binary variable for rushed (1.0) and
non-rushed (0.0) privatization years
• We also construct a continuous variable, capturing the number of
privatizations carried out in a given year
[Ideally, we would want to carry out the analysis also using proceeds
(revenue) from privatization, but these data are unavailable to
researchers]
9. Econometric issues – control groups
• Use 2 control groups
• Non-privatized SOEs (firms that were still SOEs at the end of our sample period)
[we do not use other SOEs, e.g., those privatized 2 years later, because of possible
anticipation effects, Ashenfelter dip]
• Private firms (those where government ownership never observed)
• Privatization creates a natural anchor in time: timing is set
• Need an analogous anchor for both control groups
randomly allocate placebo privatization year to firms in control groups (i.e., mimic
time intensity of actual privatizations)
• Anchor for all firms permits us to identify effects of privatization by
comparing ex ante to ex post performance of firms
10. Econometric issues -- matching
• Firms privatized in rush years may be systematically different from those
privatized in no-rush years
• E.g., government may privatize larger firm to raise more funds
=> employ propensity score matching (PSM) to balance out possible disparities across treated
and non-treated firms
introduce weights from PSM into our difference in difference specification
• Use kernel matching with Mahalanobis metric on industry and within 0.5 caliper
• Matching variables – lagged assets, capital, employment, and profit (government
and investors know these prior to taking decision on privatization)
• Recode continuous variables as categorical and interact => improve balancing
properties
11. Estimation strategy
• Adopt a difference-in-difference (DID) strategy – estimate the differences in
trend performance between rushed privatized firms and non-rush
privatized firms
• Double difference in that compare before-after difference in performance
for two sets of differently privatized firms
• Where VA = value added, K = capital (fixed assets + intangible assets), L =
labor (full-time equivalent employees) and TE = time effects
• Of key interest δ = own effect of rush privatizations along time before and
after privatization
ln 𝑉𝐴 𝑖,𝑡 = 𝛽0 + 𝛽1 ln 𝐾 𝑖,𝑡 + 𝛽2 ln 𝐿 𝑖,𝑡 + 𝛾𝑝𝑟𝑖𝑣𝑖 + 𝜃𝑟𝑢𝑠ℎ𝑖 + 𝛿𝑝𝑟𝑖𝑣𝑖 ∗ 𝑟𝑢𝑠ℎ𝑡 ∗ 𝑇𝐸 + 𝜖𝑖 + 𝜀𝑡 (1)
Y/L𝑖,𝑡 = 𝛽0 + ln 𝛾𝑝𝑟𝑖𝑣𝑎𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖 + 𝜃𝑟𝑢𝑠ℎ𝑖 + 𝛿𝑝𝑟𝑖𝑣𝑎𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖 ∗ 𝑟𝑢𝑠ℎ𝑡 ∗ 𝑇𝐸 + 𝜖𝑖 + 𝜀𝑡 (2)
12. Estimation strategy (2)
• Given the model specification, 𝜖𝑖 and εt are uncorrelated.
• Possible source of bias remains firm response to productivity shocks
(Olley and Pakes 1996, Levinsohn and Petrin 2003); given the before-
after approach, this problem is not likely to affect δ.
13. Table 1: Overall (combined) before-after effects
Reference group STATE
STATE
with matching PRIVATE STATE
STATE
with matching PRIVATE
Performance variable TFP TFP TFP LP LP LP
lnk 0.117*** 0.126*** 0.095***
(0.006) (0.012) (0.002)
lnempl 0.714*** 0.750*** 0.808***
(0.011) (0.034) (0.005)
After privatisation (all firms) 0.021 0.106*** 0.032* 0.054*** 0.147*** 0.054***
(0.015) (0.026) (0.017) (0.016) (0.026) (0.017)
After privatisation (rush years) -0.053** -0.078** -0.102*** -0.040* -0.070** -0.091***
(0.023) (0.032) (0.024) (0.024) (0.032) (0.024)
Observations 27,826 24,327 127,797 27,866 24,342 128,362
R-squared 0.238 0.925 0.310 0.097 0.736 0.055
Number of idn 4,620 25,126 4,623 25,245
Privatized 1340 1304 1340 1341 1304 1341
14. Empirical results (2)
• To obtain more detailed information about rush privatization effects,
we next interact rush privatization dummy with annual time dummies
(time relative to the event of rush privatization)
19. Concluding observations
• Rushed privatization appears to have a negative effect on TFP and
labor productivity
• Hypothesis of “overwhelming government capacity” supported
• Useful to get evidence from other countries
• Need to consider the negative effect in formulating policies
• Other aspects may be important (e.g., generating revenue fast) but the TFP
effect needs to be taken into account in making decisions
20. Thank you for your attention!
Welcome all questions and comments?