Jan 30, 2010 - Family-owned and domestic firms pick managers according to âfidelity modelâ ... First stage: firms may become innovative, by investing in R&D.
Temporary workers and seasoned managers as causes of low productivity Francesco Daveri Università di Parma, IGIER and CESifo
Maria Laura Parisi Università di Brescia
Ifo / CESifo & OECD Conference “Regulation: Political Economy, Measurement and Effects on Performance”, Munich, January 29-30, 2010
Outline • • • • • • •
Italy’s productivity slowdown: the issue Two views about the productivity slowdown Conceptual framework Empirical specification The data Results Conclusions
The Gdp growth gap between Italy and the other EU-G4 …
Source: Francesco Daveri, “Italy, before and after Lehman Brothers”, VoxEu.org, June 5, 2009
… is mainly a productivity story … Table 1: Decomposing Italy’s per-capita GDP growth, 1970-2004 Growth
Per-capita
rates of:
GDP
GDP per Hours per Working hour working age pop’n worked age person over total pop’n
1970-80
3.1
3.9
-0.8
0.0
1980-95
1.8
2.1
-0.7
0.4
1995-04
1.3
0.5
1.0
-0.2
… and the productivity slowdown has been very pronounced in manufacturing and in 2000-03 Growth of labor productivity in Italy 1970-2003, main industry groups 1970-80 1980-95 1995-03
1995-00 2000-03
Economy
+2.4
+1.8
+0.6
+1.1
-0.2
Manufacturing
+2.8
+3.0
+0.2
+1.0
-1.0
-- non-durables
+2.7
+3.1
+0.3
+0.7
-0.2
-- durables
+2.9
+2.7
+0.0
+1.7
-2.7
Two views on Italy’s productivity slowdown The Labor Supply view • Half-hearted LM reform has eased labor market entry of unskilled temporary workers since 1998 – This lowered the equilibrium K/L ratio (and possibly TFP)
• Reference: Robert Gordon and Ian Dew-Becker (2008)
L a te 1 9 9 0 s
Temporary contract legislation eased substantially in Europe between the end of the 1980s and the late 1990s This occurred in 1997-98 in Italy, gradual but sizable effects over time
6
5
Greece
4
Italy France 3
Spain
Portugal Belgium Germany
2
Finland Austria Sweden Netherlands
1
Switzerland
Denmark
Ireland United Kingdom 0 0
1
2
3
Late 1980s
4
5
6
Two views on Italy’s productivity slowdown The Labor Demand view • management selection and career criteria influencing negatively productivity growth – Family-owned and domestic firms pick managers according to “fidelity model” – MNEs and non-family adopt performance-based management model
• Reference: Oriana Bandiera, Luigi Guiso, Andrea Prat, Raffaella Sadun (2008)
Our goals Streamline the two views and test them against company data If LS view is correct, expect to see more pronounced productivity declines in companies with disproportionately higher shares of part-time and temporary workers If LD view is correct, expect to detect stronger productivity declines in companies with older and/or senior managers • both age and seniority may in fact matter: Daveri and Maliranta (2007), it is seniority the damaging feature for productivity in Finnish high-tech companies, not age as such • Test for this asymmetry between innovative and noninnovative firms with Italian data too
The conceptual framework Two-stage conceptual framework First stage: firms may become innovative, by investing in R&D or thanks to their cash flows. Or it may stay non-innovative Second stage: production occurs If firm non-innovative, experience enhances efficiency • YN = AN KNαLNβ, with AN = BNEN(1-α-β) If firm innovative, experience entails a benefit but also a cost in terms of foregone managerial training. – “Too much” experience is bad for productivity in an innovative firm
• YI = AIKIα LIβ , with AI =EI(1-α-β) + S (1-α-β) – S is managerial capital accumulated at the business school under the time constraint that S = T – EI
Empirical implications Firms select themselves into innovative and non-innovative ones depending on some exogenous circumstances such as the occurrence of R&D spending and the presence of enough cash flow Experience is good for labor productivity in the non-innovative firms and either good or bad for productivity in the I-firms. – –
managerial seniority = measure of managerial experience within the firm. managerial age = measure of overall managerial experience over and above the experience gained within a given firm
Labor market liberalization and other measures that tend to favor the entry of unskilled workers is equally bad for labor productivity in both types of firms
Baseline empirical framework ⎛Y ⎞ ⎛K⎞ ⎛ Temporary⎞ ∆2 ln⎜ ⎟ = α∆2 ln⎜ ⎟ + β + γAgei + λSeniorityit − 2 + µ ⎜ ⎟ + ε it L ⎝ L ⎠it ⎝ L ⎠it ⎝ ⎠it − 2
Dependent variable: long log-difference of LP, 2001-03 Age is the average age of board members when appointed Seniority is the average (with respect to all the firm board members) number of years in the board as of 2001 (the initial year of our sample). Temporary/L is the share of workers in the firm hired on a fixedterm basis (full time + part time) as of 2001 Additional controls for 21 industries (Ateco2007 code), geographical macro areas, size dummies and firm membership in a group
Empirical strategy 1.
2. •
• 4. 5.
OLS estimation first, with and without restrictions on AGE, SENIORITY and TEMP parameters, testing whether parameter “instability” b/w innov/non-innov is there or not Endogenous formation of “innovation groups” (innovative, non-innovative) firms may introduce innovations or not depending on whether they are more productive, young or intensive at investing into R&D activity and human capital, or maybe because they have more cash flows Hence “switching regression model” with endogenous switching (Maddala 1983) 2SLS estimation to instrument for potential endogeneity of capital stocks as well Extensions
Switching regression ⎧ ⎛Y ⎞ ⎛K⎞ ⎛ Temporary ⎞ ln ln α β γ λ µ Age Senior ∆ = ∆ + + + + ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ + ε1it if innovative i i 2 1 1 1 1 ⎪ 2 L L ⎝ ⎠it ⎝ L ⎠it ⎝ ⎠it−2 ⎪ ⎪⎪ ⎞ ⎛ Temporary ⎛K⎞ ⎛Y ⎞ ln ln α β γ λ µ Age Senior + + + + = ∆ ∆ ⎟ + ε 2it if noninnovativ ⎜ ⎟ ⎜ ⎟ ⎜ ⎨ 2 i i 2 2 2 2 2 L ⎠it−2 ⎝ ⎝ L ⎠it ⎝ L ⎠it ⎪ ⎪Ε(ε1it | innovative )≠0 ⎪ ⎪⎩Ε(ε 2it | noninnovative )≠0
With following criterion to determine if firm belongs to regime 1 ⎧ D 1 = 1 if δ ' Z it + ω it > 0 ⎨ D 1 = 0 otherwise ⎩
To estimate the δs, ML probit for: E(D1)=P(D1=1)=P(δ’Zit+ωit>0) is the probability of being an innovative firm. Z = instruments for decision to innovate. E.g. R&D spending, number of R&D workers, firm age, cash flow in 2001
Switching regression After correcting for the error conditional mean and substituting the estimate of δ’Z in the correction terms (from the first stage), we estimate a system of equations through OLS (second stage) φ(δ ' Zit ) ⎧ ⎛ Temporary ⎞ ⎛K⎞ ⎛Y ⎞ Age Senior α β γ λ µ σ ln ln + + + + − + u1it if D1 =1 = ∆ ∆ ⎟ ⎜ ⎜ ⎟ ⎜ ⎟ i i 2 1 1 1 1 ε1ω ⎪ 2 L L Φ(δ ' Zit ) ⎝ L ⎠it ⎝ ⎠it ⎝ ⎠it−2 ⎪ ⎨ φ(δ ' Zit ) ⎞ ⎪∆ ln⎛⎜ Y ⎞⎟ =α∆ ln⎛⎜ K ⎞⎟ + β + γ Age + λ Senior+ µ ⎛⎜ Temporary + u2it otherwis ⎟ + σε1ω i i 2 2 2 2 2 ⎪⎩ 2 ⎝ L ⎠it L L δ Z 1 ( ' ) − Φ ⎝ ⎠it ⎝ ⎠it−2 it
After rearranging, we obtain an estimable specification to perform Wald tests of parameters instability (ξit is a standard normal): ⎛ Temporary⎞ ⎛Y ⎞ ⎛K⎞ ∆ 2 ln⎜ ⎟ = α∆ 2 ln⎜ ⎟ + β 2 + γ 2 Agei + λ2 Seniorityit −2 + µ 2 ⎜ ⎟ + (β1 − β 2 )Φ(δˆ' Z it ) + L ⎝ L ⎠ it ⎝ L ⎠ it ⎝ ⎠ it −2 ⎛ Temporary⎞ + (γ 1 − γ 2 ) Agei Φ(δˆ' Z it ) + (λ1 − λ2 )Seniorityit −2 Φ(δˆ' Z it ) + (µ1 − µ 2 )⎜ ⎟ Φ(δˆ' Z it ) + L ⎝ ⎠ it −2 + φ (δˆ' Z )(σ − σ ) + ξ it
ε 2ω
ε1ω
it
If parameters do not change across groups, back to baseline specification
Data Match firm-individual level data from AIDA with Capitalia/Unicredit Manufacturing Firm survey From AIDA we obtained: • firm level 2001-2003 balance sheets for firms with publicly available data in 2007 • individual level information on board members birth dates and nomination dates
From Capitalia-Unicredit survey we use: • firm level 2001-2003 balance sheet for firms filling the questionnaires in 2004 ( 4000) • innovation activity, introduction of innovations, R&D spending, investments • labor force, temporary workers, R&D workers
Benchmark OLS results
Diagnostics of OLS results
Remarks on OLS results Innovative firms enjoy positive productivity drift – particularly so for firms introducing process innovation Age: negative and statistically significant coeff for innov firms; positive or zero coeff for non-innov firms (positive when innovation is process innovation; zero otherwise) Seniority: shows zero coefficients almost everywhere Temporary workers: always negative and sizable for all categories of firms Chow test indicates presence of overall parameter instability (two regimes are there) Wald tests show parameter instability for age and seniority only
Switching regression – 1st stage
Remarks on 1st stage switching regressions Table 6.a: probit of 1st stage decision to innovate Column 1-2 innovation, 3-4 product innovation; 5-6 process innovation Useful to learn about relevance of instruments • Engaging in R&D and hiring R&D workers affects more product than process innovation • Cash flow is particularly important for process innovation In the second stage we use column 1 results only
Switching regressions – 2nd stage, OLS & 2SLS
Remarks on 2nd stage switching regressions Table 6.b: OLS and 2SLS of 2nd stage regressions Instruments for growth rate of K/L: initial level of K/L, firm age, investment intensity in 2001. Plus size, area and industry dummies OLS vs. 2SLS: does not seem to matter a lot Innovative firms (Correction for switching regression is negative and significant) • Age coefficient is negative and statistically significant • Seniority bears zero coefficients • Share of temporary workers still negative, sizable and strongly significant Non-innovative firms (Correction for switching regression is positive or zero) • Age, either null (OLS) or positive (IV) coefficient • Seniority: negative and quite significant with both OLS and IV (very different from OLS & at odds with expectations; puzzling result) • Share of temporary workers: negative and statistically significant
Two robustness checks Sub-sample of “influential” managers (CEO, CIO, CFO, Sales manager, Chairperson) • Much smaller sample, but no change in results (results not reported in this draft, available upon requests) ML endogenous switching model (Lokshin and Sajaja, 2004) • Estimate system (3) as it is (“ML”) and after substituting predicted K/L (“ML predicted”) • Results for innovative firms replicate those of Table 6.b • For non-innovative firms, impact of average board age is zero (2SLS was positive in Table 6.b), while seniority is still negative and significant. The share of temporary workers is still negatively related to productivity growth • In both cases, correlation between 1st and 2nd stage regression is zero. This is shown by the Wald test of independence across the equations of the system (3)
Conclusions Italy’s prolonged productivity slowdown is not cyclical Here we contrast two views We find that the Labor Supply view is consistent with our productivity data Our results indicate that managerial experience is not unambiguously correlated with productivity performance in the entire sample Definite patterns of correlation are present though, once the whole sample is split into innovative and non-innovative clusters. Age, in particular – a measure of overall experience – in the labor market appears to be negatively correlated with productivity for innovative firms and only weakly positively correlated with productivity in non-innovative firms The pattern of correlation for seniority is instead less robust and requires further investigation.
Some answers to potential remarks of the discussant The cross-sectional statistical analysis of long-differences based on firm-averaged data is not problem-free. A big issue is potential reverse causation • The statistical relations we intend to analyze posit that the temporary share of workers, age or seniority variables are the independent variables and productivity the dependent variable. • But cross-section data as such (be they observed at a given point in time or averaged over time) may only indicate correlation, not causation. • Therefore, if the estimated coefficient linking seniority and productivity is negative, this may not indicate that the firms where aged or senior managers are employed are less productive. • Rather, the negative correlation may simply signal that older or senior managers tend to stay longer in less productive and older firms, featuring outdated machines and methods of production, probably because they managed to put in place successful “relations”, while new, innovative and high-productivity plants may be more often matched to young and brilliant managers. • Hence we would be wrongly interpreting what causes what, attributing to seniority or age a causal influence on plant productivity, which may go the other way around. This is why we implement our 2SLS specification. Our expectation is that by choosing predetermined instruments we may lessen the simultaneity problems
Some answers to potential remarks of the discussant Surely, a lot of unobserved heterogeneity in plant productivity is still there in the data even once we have augmented the list of productivity determinants with dummies and other control variables. Yet the problem of interpreting the statistical results from crosssectional estimates arises if and only if the unobserved (therefore unmeasured) firm variables are correlated with the included explanatory variables. To tackle this problem, we control for a few dummy variables that capture some, though presumably not all, of the unobserved determinants of firm productivity
