The Microeconomic Implications of Labor Regulations: Cross-Country Evidence from Within the Firm
Francine LafontaineJagadeesh Sivadasan
Ross School of Business, University of Michigan
AEA Meetings, Jan 2007 (Updated Jan 2009)
2
Introduction
Goal: Assess the effects of regulations that create rigidities in the labor market
Important and controversial question: a number of papers have considered the effects of such regulations at the macro level (e.g. Botero et al (2004), Lazear (1990))
We focus on the very micro level, using data from a single fast-food chain with operations in more than 40 countries around the world: most importantly, we quantify the effect of the regulations on firm
labor adjustment decisions also examine effect on extent of operations
3
Introduction
Important advantages of our empirical setting: Fundamentally same production technology Same output (as close as can get) Labor intensive - labor issues really matter We have very detailed data: weekly revenues and
costs information for each outlet outside the US (more than 2500 of them, over 4 years, in 43 countries)
4
Introduction
Important advantages of our empirical setting: The high frequency of our data has two main
advantages: Gets around problems with annual data, which can hide
a lot of within year turnover (Hamermesh and Pfann (1996))
Allows us to adopt estimation strategies with lots of controls (i.e. fixed effects at the outlet, outlet-year, or outlet-year-season)
5
Introduction
We focus on questions that are particularly suited to our data Companion paper – addresses the effect of regulation
on productivity and labor demand In some sense, more direct place to look for effects But in reality, theory ambiguous on these effects,
whereas clear predictions on hysteresis and effect on responsiveness to output changes
And our data – labor costs, not labor levels => potential biases due to “poor” measure of wages
6
Introduction – Preview of Results
We find strong evidence that in countries with more rigid labor laws: Outlets’ choices of labor levels are less responsive,
from period to period, to changes in revenues, and there is more hysteresis in labor levels, that is labor in
one period is more related to previous period labor We also find some evidence that in such
countries: the Company enters later and operates fewer outlets
-- and uses “local” franchising more as well
7
Organization of this Talk
Basic Model and Predictions A Note on Dynamics Data and Definition of Variables Results for Labor Adjustment – Botero & GCS Key Identification Issues
Contrasting with Materials Adjustment Results IV Results
Results on Company’s Extent of Operations Conclusions
8
Basic Model
Draws on Heckman and Pagés (2003), who drew on Holt, Modigliani, Muth and Simon (1960)
Given our weekly data, we take capital as given (or contributing to the Hicks neutral productivity term), and write a 2 input Cobb-Douglas production process with labor and materials
where Yt is the quantity of output produced by the firm in period t, Lt is labor, and Mt represents materials used
tttt MLY
9
Basic Model
Assume iso-elastic demand curve:
where Pt is the price per unit output in period t, represents demand shifters and is the price elasticity of demand
The firm’s profit to be maximized is
where Wt is the wage rate faced by the firm in period t, and St is the per unit cost of materials
ttttttt MSLWQP
/1tt QP
10
Basic Model
Each week, manager chooses labor and materials so FOCs for these are binding
Assume horizontal labor and materials supply in each local market (each outlet buys little, and even as a group they are a tiny part of the market)=> obtain optimal labor and materials demand function
in terms of the primitives (prices, demand & production function parameters)
=> can write total labor and material cost equations conditional on output
11
Basic Model
These input demand equations are
where we use bt to denote log(WtLt) and ft to mean log(StMt), and
These equations then describe equilibrium input costs in the absence of adjustment costs
' log t*t rb
t*
t rf ' log
.1
1 'and1
1 '
12
Basic Model
Now suppose that there are costs to adjusting labor. First, let the cost of being off the static optimum be quadratic in log labor
where o > 0. Second, suppose that the cost of changing labor levels from one period to the next are given by
where we expect a to be positive and increasing in the rigidity of labor regulations
2*o )( tt
ot bbc
21a )( tt
at bbc
13
Basic Model
Each outlet minimizes the sum of these costs. This yields optimal labor choice
where outlet i is in country j, and
The optimal labor cost equation above can be rewritten as
1,*)-(1 ti
jit
jit bbb
')log-1()-1( 1, jti
jit
jit brb
.o
ja
jaj
14
Basic Model
Taking, as a first approximation,
we get the following econometric specification for the labor costs of outlet i in country j at time t:
where τ j is the index of labor regulation, and is stands for store, store-year, or store-season-year fixed effects.
itistij
bitj
rtiit
jti
jit
jit
brbr
aabaaraab
1,1,
101,1010 'log)1()()1(
jj aa 10
15
Basic Model
In this regression, we expect r to be negative and b to be positive.
In other words, our simple model yields two principal implications that we bring to data: Labor costs should be less responsive to changes in revenues
in countries where regulations are more stringent Labor costs at time t should be more dependent on labor costs
at time t-1 in countries with more stringent laws (hysteresis) These predictions are intuitive, and the latter has been
tested in a number of studies of the effect of regulation on labor demand (see survey in Heckman and Pagés, 2004).
16
Note on Dynamics
Our two testable implications are derived from a very simple model Heckman and Pagés (2004) express concern that the labor hysteresis
prediction may not arise in a more general dynamic model We solved a more general dynamic stochastic programming model:
Two state variables are current productivity and last period labor. We solve numerically for four scenarios: with both symmetric (quadratic) and asymmetric (i.e. severance pay
only) adjustment costs. and for iid as well as persistent demand /productivity shocks processes
17
Note on Dynamics To approximate our actual data, using optimal policy functions, we ran regressions
on simulated behavior of 75 outlets for 104 periods across 45 regimes Zero adjustment costs Symmetric quadratic
adjustment costs Asymmetric linear
adjustment costs Fixed (lump-sum)
adjustment costs IID
shocks Persistent
shocks IID
shocks Persistent
shocks IID
shocks Persistent
shocks IID
shocks Persistent
shocks Log (Lagged labor cost) 0.0000 0.0006*** 0.657*** 0.574*** 0.874*** 0.251*** 0.117*** 0.146*** [0.0001] [0.0001] [0.039] [0.015] [0.043] [0.017] [0.008] [0.008] Log (Revenue) 0.984*** 0.984*** 0.198*** 0.264*** 0.190*** 0.288*** 0.398*** 0.677*** [0.0001] [0.0002] [0.022] [0.018] [0.030] [0.020] [0.026] [0.023] Adj. cost X Log (Lagged labor cost) 0.0001 -0.0004 1.130*** 0.690*** 1.856*** 0.398*** 0.253*** 0.421*** [0.0002] [0.0003] [0.151] [0.090] [0.104] [0.032] [0.020] [0.031] Adj. cost X Log (Revenue) -0.0001 0.0001 -0.540*** -0.602*** -0.885*** -1.083*** -1.344*** -0.971*** [0.0004] [0.0006] [0.111] [0.089] [0.118] [0.073] [0.067] [0.089] Constant -1.529*** -1.528*** -0.555*** -0.688*** -0.181** -1.267*** -1.470*** -1.337*** [0.0001] [0.0002] [0.072] [0.022] [0.072] [0.023] [0.021] [0.005] Fixed Effects Outlet-year-
season Outlet-year-
season Outlet-year-
season Outlet-year-
season Outlet-year-
season Outlet-year-
season Outlet-year-
season Outlet-year-
season Observations 351000 351000 351000 351000 351000 351000 351000 351000 Adjusted R-squared 0.999 0.999 0.801 0.932 0.782 0.929 0.852 0.90 Number of clusters 45 45 45 45 45 45 45 45
18
The Data
Mostly from internal firm records Cover over 2500 outlets in more than 40 countries
worldwide, weekly from 2000 to 2003 Data on
Revenues per week Total labor costs per week Total materials costs per week Number of items (standardized notion of output)
19
The Data
We measure the rigidity of the labor regulations in each country using the Botero et al (2004) index (see appendix in paper for details) Main advantage – computed similarly across
countries Main disadvantage – laws may not be enforced as
strongly everywhere
We verify our results using an index of hiring and firing flexibility from the Global Competitiveness Survey (2002) of business executives
20
Index of labor regulation (Botero, et al)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9Z
MB
HK
G
USA
UG
Y
SGP
IRL
AU
S
BO
L
TH
A
KO
R
LK
A
HR
V
BE
L
VN
M
DO
M
PO
L
UK
R
NO
R
NL
D
FRA
KA
Z
RU
S
21
ArgentinaAustralia
Austria BelgiumBulgaria
Bolivia
Brazil
Canada
SwitzerlandChile
China
Colombia
Czech Republic
Germany
DenmarkDominican Republic
Ecuador
SpainFinland France
Great Britain
Greece
Hong Kong
Croatia
Hungary
Indonesia
India
Ireland
Israel
Italy
Jamaica
Jordan
Japan
South KoreaSri Lanka
Lithuania
Latvia
Morocco
Mexico
Malaysia
NetherlandsNorway
New Zealand
Panama
Peru Philippines
Poland
PortugalRussian Federation
Singapore
Slovakia
Slovenia Sweden
Thailand
Tunisia
Turkey
Taiwan
Ukraine
United States
Venezuela
Vietnam
South Africa
Zimbabwe
0.2
.4.6
.8In
dex
of l
abo
r re
gul
atio
n (
Bot
ero
, et a
l)
.2 .4 .6 .8 1Hiring/firing inflexibility (2002, Global Competitiveness Report)
Fitted (OLS) Fitted (GLS, Root-N weight)
22
Panel Characteristics (baseline sample)
Year Number of observations
Number of outlets
Number of countries
2000 80,429 1,721 39 2001 85,113 1,828 37 2002 74,201 2,147 38 2003 82,305 1,938 37 Total 322,048 2,526 43
23
Descriptive StatisticsPanel B: Summary statistics (variables in logs) Variable N Mean SD P25 Median P75 Min Max
Log (Labor cost) 322,048 7.19 0.85 6.71 7.27 7.78 -5.05 10.25 Log (Revenue) 322,048 8.84 0.69 8.46 8.90 9.32 2.85 11.50 Log (Material cost) 318,749 7.72 0.66 7.37 7.78 8.16 -4.87 10.94 Panel C: Summary statistics (variables in levels) Variable N Mean SD P25 Median P75 Min Max
Labor cost 322,048 1,798.57 1,391.21 819.84 1,434.39 2,390.97 0.01 28,219 Revenue 322,048 8,485.44 5,329.17 4,730.58 7,332.80 11,156.53 17.30 98,668 Material cost 318,749 2,706.78 1,626.86 1,590.36 2,394.45 3,481.43 0.01 56,580 Index of labor regulation 322,048 0.00 0.16 -0.15 0.03 0.16 -0.25 0.42
24
0.2
.4.6
.81
.2 .4 .6 .8Index of labor regulation (Botero, et al)
Correlation between changes in labor cost and revenue
Fitted (GLS, Root-N weight)
Correlation between changes in material cost and revenue
Fitted (GLS, Root-N weight)
0.2
.4.6
.81
.2 .4 .6 .8 1Hiring/firing inflexibility (2002, Global Competitiveness Report)
Correlation between changes in labor cost and revenue
Fitted (GLS, Root-N weight)
Correlation between changes in material cost and revenue
Fitted (GLS, Root-N weight)
Correlation between changes in input costs and revenue
25
Table 4: Baseline Results
(4) (5) (6) Log (Lagged labor cost) 0.501*** 0.348*** 0.203*** [0.049] [0.036] [0.033] Log (Revenue) 0.341*** 0.360*** 0.391*** [0.040] [0.035] [0.036] Regulation X Log (Lagged labor cost) 1.013*** 0.909*** 0.687*** [0.291] [0.223] [0.203] Regulation X Log (Revenue) -0.570*** -0.488*** -0.414*** [0.144] [0.101] [0.106] Constant 0.623*** 1.550*** 2.305*** [0.227] [0.310] [0.358]
Fixed Effects Outlet Outlet-year Outlet-year-
season
Observations 322,048 322,048 322,048 Adjusted R-squared 0.945 0.952 0.959 Number of clusters 43 43 43
26
Table 4: Baseline Results
Effect of a one standard deviation (0.85) increase in Log (Lagged Labor) in percentage terms
At Regulation = mean (0.00) 42.59 29.58 17.26
At Regulation = mean + sd (= 0.16) 56.36 41.94 26.60
Impact of increase in Regulation 13.78 12.36 9.34
Effect of a one standard deviation (0.69) increase in Log (Revenue) in percentage terms
At Regulation = mean (0.00) 23.53 24.84 26.98
At Regulation = mean + sd (= 0.16) 17.24 19.45 22.41
Impact of increase in Regulation -6.29 -5.39 -4.57
27
Table 5: GCS Index Results
(1) (2) (3) Log (Lagged labor cost) 0.530*** 0.371*** 0.220*** [0.053] [0.041] [0.034] Log (Revenue) 0.355*** 0.383*** 0.418*** [0.041] [0.036] [0.038] Inflexibility X Log (Lagged labor cost) 0.988*** 0.914*** 0.710*** [0.332] [0.291] [0.235] Inflexibility X Log (Revenue) -0.734*** -0.714*** -0.629** [0.221] [0.217] [0.262] Constant 0.244 1.136*** 1.909*** [0.240] [0.316] [0.363]
Fixed Effects Outlet Outlet-year Outlet-
year-season
Observations 338,660 338,660 338,660 Adjusted R-squared 0.948 0.955 0.961 Number of clusters 48 48 48
28
Table 5: GCS Index Results
Effect of a one standard deviation (0.85) increase in Log (Lagged Labor) in percentage terms
At Inflexibility = mean (0.00) 45.05 31.53 18.70
At Inflexibility = mean + sd (= 0.13) 55.97 41.63 26.55
Impact of increase in Inflexibility 10.92 10.10 7.85
Effect of a one standard deviation (0.69) increase in Log (Revenue) in percentage terms
At Inflexibility = mean (0.00) 24.50 26.43 28.84
At Inflexibility = mean + sd (= 0.13) 17.91 20.02 23.20
Impact of increase in Inflexibility -6.58 -6.40 -5.64
29
Key Identification Issues
As per our model above, the error term is:
Thus the error term includes omitted supply-side parameter it (output elasticity with respect to labor) and demand side parameter it (elasticity of demand)
Our store or store-year-season fixed effects control implicitly for the differences in τ j across countries, and for these supply and demand parameters insofar as they are fixed within a store or store-year or store-year-season
itit
jit aae
1
1log)1( 10
30
Key Identification Issues
Unanticipated demand and productivity shocks can also add to the error term if labor is set early, then a high (low) demand shock shows up as
too little (much) labor conditional on output Since we are interested in the coefficient on lagged labor
or revenue interacted with regulation, a bias arises only when: the omitted demand and supply parameters vary within store-
year-seasons, and are correlated with lagged labor/revenue in a different way across different regulation regimes
prediction errors (with regard to demand and productivity shocks) are correlated with levels of regulation
31
Key Identification Issues
We address these potential identification issues in two different ways: using the material demand equation as a control with instrumental variables
The material demand equation is a good control because: omitted demand and supply parameters are the same for the material
as for the labor demand equation prediction error also would bias the material demand specification in a
similar way IV approach
draws from traditional approaches in the literature (Blundell and Bond 1998)
uses lagged endogenous variables that are uncorrelated with unpredicted component of current demand and productivity shocks
we have more instruments than endogenous variables, so we can perform an overidentification test – good results!
32
Table 6: Material Costs Results (Botero et al. Index)
(4) (5) (6) Log (Lagged materials cost) 0.159*** 0.112*** 0.033* [0.038] [0.036] [0.019] Log (Revenue) 0.852*** 0.901*** 0.942*** [0.027] [0.020] [0.008] Regulation X Log (Lagged materials cost) -0.211 -0.168 -0.089 [0.201] [0.197] [0.125] Regulation X Log (Revenue) -0.020 -0.004 -0.075* [0.138] [0.093] [0.043] Constant -1.032*** -1.102*** -0.856*** [0.084] [0.127] [0.132]
Fixed Effects Outlet Outlet-year Outlet-
year-season
Observations 362,711 362,711 362,711 Adjusted R-squared 0.947 0.953 0.960 Number of clusters 43 43 43
33
Table 6: Material Costs Results (Botero et al. Index)
Effect of a one standard deviation (0.66) increase in Log (Lagged Materials cost)
At Regulation = mean (0.00) 10.49 7.39 2.18
At Regulation = mean + sd (=0.16) 8.27 5.62 1.24
Impact of increase in Regulation -2.23 -1.77 -0.94
Effect of a one standard deviation (0.69) increase in Log (Revenue)
At Regulation = mean (0.00) 58.79 62.17 65.00
At Regulation = mean + sd (=0.16) 58.57 62.12 64.17
Impact of increase in Regulation -0.22 -0.04 -0.83
34
Table 7: Robustness check: OECD Sample and Interaction Terms (1)
OECD (2) (3) (4) (5)
Log (Lagged labor cost) 0.174*** 1.709*** -0.052 0.204 0.165 [0.030] [0.325] [0.063] [0.333] [0.176] Log (Revenue) 0.461*** -1.085*** 0.639*** 0.726*** 0.645*** [0.036] [0.276] [0.065] [0.205] [0.171] Regulation X Log (Lagged labor cost) 0.428*** 0.331** 0.380* 0.687** 0.700*** [0.141] [0.161] [0.200] [0.299] [0.238] Regulation X Log (Revenue) -0.354* -0.069 -0.225 -0.584*** -0.543*** [0.183] [0.108] [0.152] [0.152] [0.141] GDP X Log (Lagged labor cost) -0.158***
[0.033]
GDP X Log (Revenue) 0.156***
[0.030]
Entry barriers X Log (Lagged labor cost) 0.081***
[0.021]
Entry barriers X Log (Revenue) -0.078***
[0.025]
Wage flexibility X Log (Lagged labor cost) -0.0002
[0.064]
Wage flexibility X Log (Revenue) -0.063
[0.038]
Labor relations X Log (Lagged labor cost) 0.008 [0.040] Labor relations X Log (Revenue) -0.053 [0.032]
Fixed effects Outlet-year-season
Outlet-year-
season
Outlet-year-
season
Outlet-year-
season
Outlet-year-
season
Constant 1.996*** 2.197*** 2.273*** 2.238*** 2.291*** [0.314] [0.321] [0.296] [0.344] [0.321] Observations 236,291 322,048 265,842 321,569 321,569
35
Table 8: Robustness check: DID comparison of top and bottom decile of change in Index of Inflexibility between 2002 and 2004
LABOR MATERIALS (1) (2) (3) (4)
Log (Lagged input cost) 0.570*** 0.358*** 0.157*** 0.055*** [0.032] [0.057] [0.022] [0.016] Log (Revenue) 0.152*** 0.198*** 0.897*** 0.918*** [0.038] [0.053] [0.023] [0.021] Year 2003 -0.474* -0.138 [0.280] [0.122] Year 2003 X Log (Lagged input cost) -0.328*** -0.233*** -0.034 0.008 [0.040] [0.059] [0.023] [0.018] Year 2003 X Log (Revenue) 0.321*** 0.328*** 0.048* 0.032 [0.037] [0.067] [0.025] [0.022] DInf_p90 X Year 2003 1.435*** 1.107*** [0.494] [0.164] DInf_p90 X Log (Lagged input cost) -0.143 -0.179 0.104*** 0.103*** [0.112] [0.191] [0.037] [0.033] DInf_p90 X Log (Revenue) 0.196 0.083 -0.091** 0.009 [0.171] [0.202] [0.045] [0.044] DInf_p90 X Year 2003 X Log (Lagged input cost) 0.519*** 0.432** 0.001 -0.077* [0.127] [0.198] [0.053] [0.044] DInf_p90 X Year 2003 X Log (Revenue) -0.634*** -0.722*** -0.125** -0.078 [0.105] [0.211] [0.057] [0.060] DGDPGR X Year 2003 15.56** 9.037*** [6.029] [1.803] DGDPGR X Log (Lagged input cost) -1.970*** -0.480 1.555*** 1.061*** [0.660] [1.073] [0.303] [0.256] DGDPGR X Log (Revenue) 1.403 0.341 -1.110*** -0.135 [1.378] [1.654] [0.350] [0.320] DGDPGR X Year 2003 X Log (Lagged input cost) 1.881** 0.647 -2.257*** -2.088*** [0.857] [1.101] [0.360] [0.304] DGDPGR X Year 2003 X Log (Revenue) -3.569*** -4.088** 0.806** 0.181 [0.823] [1.780] [0.362] [0.375] Constant 1.233*** 1.968*** -1.250*** -0.941*** [0.288] [0.252] [0.132] [0.086]
Fixed effects Outlet Outlet- Outlet Outlet-
36
Table 8: Robustness check: Case study of labor reform in South Korea (1996-98)
BEFORE-AFTERLABOR MATERIALS
(1) (2) (3) (4)
Log (Lagged input cost) 0.765***0.270*
** 0.332*** 0.122***[0.049] [0.058] [0.039] [0.024]
Log (Revenue) 0.206***0.160*
** 0.717*** 0.822***[0.034] [0.041] [0.034] [0.026]
POST_REFORM X Log (Lagged input cost)
-0.404**
*
-0.129*
* -0.024 -0.018
[0.053][0.064
] [0.044] [0.029]
POST_REFORM X Log (Revenue)0.335**
*0.574*
** 0.033 0.040
[0.043][0.049
] [0.037] [0.035]D_KOREA X Log (Lagged input cost)
D_KOREA X Log (Revenue)
D_KOREA X POST_REFORM X Log (Lagged input cost)
D_KOREA X POST_REFORM X Log (Revenue)
Constant -0.1270.515*
**-
1.242***-
0.710***[0.096] [0.145] [0.100] [0.137]
Fixed Effects OutletOutlet-
year-season
OutletOutlet-
year-season
Observations 15,071 15,071 15,099 15,099 Adjusted R-squared 0.854 0.894 0.944 0.963Number of clusters 152 152 152 152
37
Table 8: Robustness check: Case study of labor reform in South Korea (1996-98)
DIFFERENCE-IN-DIFFERENCES LABOR MATERIALS
(5) (6) (7) (8) Log (Lagged input cost) 0.272*** 0.084*** 0.074*** 0.023*** [0.021] [0.016] [0.009] [0.007] Log (Revenue) 0.460*** 0.549*** 0.958*** 0.965*** [0.013] [0.013] [0.009] [0.010]
POST_REFORM X Log (Lagged input cost) 0.190*** 0.090*** 0.054*** 0.018
[0.015] [0.021] [0.014] [0.012]
POST_REFORM X Log (Revenue) -0.167*** -0.152*** -0.043*** 0.005
[0.013] [0.014] [0.012] [0.013] D_KOREA X Log (Lagged input cost) 0.493*** 0.186*** 0.257*** 0.099*** [0.054] [0.060] [0.040] [0.024] D_KOREA X Log (Revenue) -0.254*** -0.389*** -0.241*** -0.143*** [0.036] [0.043] [0.035] [0.028]
D_KOREA X POST_REFORM X Log (Lagged input cost) -0.594*** -0.219*** -0.078* -0.036
[0.055] [0.067] [0.046] [0.031]
D_KOREA X POST_REFORM X Log (Revenue) 0.502*** 0.726*** 0.075* 0.035
[0.045] [0.050] [0.038] [0.038] Constant 1.156*** 2.158*** -1.453*** -1.153*** [0.105] [0.085] [0.053] [0.053]
Fixed Effects Outlet Outlet-
year-season
Outlet Outlet-
year-season
Observations 71,273 71,273 71,200 71,200 Adjusted R-squared 0.977 0.984 0.971 0.980 Number of clusters 592 592 592 592
38
Impulse response functions based on VAR analysis
39
Implied Rigidity Estimates
The underlying structural parameters and are not simultaneously identified.
However is identified Model implies following relationship between optimal
adjustment and actual adjustment of labor
More regulation => higher => greater dampening of adjustment
o a
jj aa 10
j
Estimate of a0 Estimate of a1 Dampening factor estimate
Regulation Change
P25 P75 (percent)
Panel 1: Using results from Column 6 of Table 4 Coefficient on Log (Lagged labor cost): 0.203 Coefficient on Regulation X Lagged labor cost: 0.687 0.900 0.687 23.7 1 - Coefficient on Log (Revenue): 0.609 - (Coefficient on Regulation X Revenue): 0.414 0.453 0.325 28.3
Average of above: 0.406 Average of above: 0.550 0.677 0.506 25.2
Panel 2: Using results from Column 1 of Table 7 – OECD only Coefficient on Log (Lagged labor cost): 0.174 Coefficient on Regulation X Lagged labor cost: 0.428 0.890 0.749 15.9 1 - Coefficient on Log (Revenue): 0.539 - (Coefficient on Regulation X Revenue): 0.354 0.514 0.397 22.7
Average of above: 0.356 Average of above: 0.391 0.702 0.573 18.4
40
Table 11: Estimates of Dampening Factor
(change in labor costs in the absence of adjustment costs / actual change in labor costs)
41
A Look at the Firm’s Expansion
If more rigid labor regulations imply that individual outlets cannot adjust labor as much as they otherwise would, then all else the same, outlets in these markets will be less profitable.
This suggests the firm should Enter later Expand less rapidly Franchise more ?
in highly regulated markets
42
Labor Regulation and International Expansion: Time to Entry
Whole Sample Country = Observation(end of 2001)
Log(GDP/capita in USD) -0.31** -0.22** -0.32** -0.22**[0.07] [0.08] [0.07] [0.08]
Log(Population) -0.19** -0.11 -0.18* -0.10[0.07] [0.06] [0.07] [0.07]
Log(Distance to USA in kms)
0.28[0.19]
0.26[0.18]
0.24[0.21]
0.23[0.19]
Regulation (Botero et al) 1.10* 1.14*[0.44] [0.45]
Regulation (GCS) -0.45 -0.44[0.65] [0.70]
Constant 5.09* 3.80* 5.36* 3.95*[2.01] [1.61] [2.14] [1.72]
Observations 6906 8111 34 40R-squared 0.42 0.23 0.45 0.24
Number of Clusters 37 43
43
Labor Regulation and International Expansion: Number of Outlets
Whole Sample Country = Observation(end of 2001)
Log(GDP in USD) 0.59** 0.57** 0.51** 0.50**[0.13] [0.12] [0.18] [0.16]
Log(Population) 0.63** 0.54** 0.58** 0.53**[0.17] [0.15] [0.20] [0.18]
Log(Distance to USA in kms)
-0.45+ [0.26]
-0.40[0.27]
-0.38[0.32]
-0.30[0.32]
Regulation (Botero et al) -2.02* -2.60*[0.83] [1.07]
Regulation (GCS) -0.37 -1.99[1.47] [2.01]
Constant -8.25* -7.54** -7.10 -6.83+[3.56] [2.81] [4.64] [3.51]
Observations 7423 8628 38 44R-squared 0.41 0.38 0.34 0.30
Number of clusters 41 47
44
Labor Regulation and International Expansion: Use of Franchising
Whole Sample Country = Observation(end of 2002)
Log(GDP in USD) 0.15 0.12 0.18 0.14[0.06]* [0.05]* [0.06]* [0.05]**
Log(Population) -0.001 -0.01 0.03 0.01[0.06] [0.04] [0.06] [0.04]
Log(Distance to USA in kms)
-0.19[0.06]**
-0.28[0.07]**
-0.17[0.08]*
-0.27[0.08]**
Regulation (Botero) 0.12 0.37[0.37] [0.38]
Regulation (GCS) 0.82 0.91[0.40]+ [0.40]*
Constant 0.59 1.43 -0.35 0.76[1.53] [1.12] [1.68] [1.30]
Observations 2852 3372 29 33R-squared 0.35 0.29 0.38 0.34
Number of clusters 30 36 29 33
45
Conclusion
Using weekly data from outlets of a multinational fast-food chain, we have shown evidence of a statistically and economically important effect of labor regulations on labor decisions at the micro level
To our knowledge this is the first time that effects of such policies are documented in a cross-country context at such a micro level
46
Conclusion
Specifically, using our most conservative estimates, we find that an increase of one standard deviation in the labor regulation rigidity index reduces the response of labor cost to a one standard
deviation increase in output (revenue) by about 4.4 percentage points (from 26.4 per cent to 22.0 percent)
increases the response of labor cost to a one standard deviation increase in lagged labor cost by about 9.6 percentage points (from 17.0 per cent to 26.6 per cent)
47
Conclusion
We have also shown that results are similar whether we use the Botero et al.
index, or an alternative measure of labor regulation from the Global Competitiveness Survey
The effects do not hold for material costs, confirming that they are not spurious – the increased rigidity in labor is not driven by omitted variables that are also likely to affect other variable costs
The effects are even stronger when we estimate using an IV approach
48
Conclusion
Consistent with the impact on adjustment behavior, we also find that the Company delayed entry, and operates fewer outlets and its partners rely on franchising more
in countries with more rigid labor laws
49
Conclusion So we have shown that labor levels are more
persistent in countries that enact more rigid labor laws, an effect these policies are meant to achieve
However, increases (responses to positive shocks) as well as decreases in labor levels are affected
Consistent with our earlier findings that outlet level labor demand was lower in more heavily regulated markets, our results on timing of entry and level of operations of the Company across markets suggest that easing these laws would increase employment and output in this industry
50
Other paper (Lafontaine and Sivadasan, “Within-firm Labor Productivity across Countries: A case study): Quantifying the Labor Demand effect
We find coefficients for regulation of about -0.4 => an increase in labor regulation (Botero Index)
from its 25th percentile to its 75th percentile value reduces labor per outlet by 0. 4 * 0.31 = 12%
Alternatively, a one standard deviation increase in the labor regulation rigidity index leads to a reduction in conditional labor demand of 6.4 per cent
51
Interpreting the Expansion Effects
From column 1 (part a): an increase in labor regulation by one standard deviation increases the time to entry by 1.7 years on a mean of 8.9 years (so not quite 20% increase)
From column 1 (part b): a one standard deviation increase in the index of labor regulation reduces the number of stores by about 3.2 relative to a mean of 12.06 outlets (a >25% reduction)
52
Interpreting the Expansion Effects
We also find some evidence that the franchisor and its partners rely on franchising more the more highly regulated the labor market is
Our analyses here are based on more limited information, however, and we have yet to verify robustness