WPS4792
Policy ReseaRch WoRking PaPeR 4792
Explaining Enterprise Performance
in Developing Countries with Business
Climate Survey Data
Jean-Jacques Dethier
Maximilian Hirn
St�phane Straub
The World Bank
Development Research Department
Research Support Unit
December 2008
Policy ReseaRch WoRking PaPeR 4791
Abstract
This paper surveys the recent literature which examines impact firm performance. Section 1 of this paper outlines
the impact of business climate variables on productivity the theoretical framework that underpins the investment
and growth in developing countries using enterprise climate literature. Section 2 describes the available
surveys. Comparable enterprise surveys today cover datasets and surveys the key findings of the empirical
some 70,000 firms in over 100 countries around the literature, first macroeconomic and then microeconomic
world. The literature that has analyzed this data provides studies. Particular attention is paid to the robustness
evidence that a good business climate drives growth of the reported results. Section 3 highlights important
by encouraging investment and higher productivity. econometric issues common to this literature and
Various infrastructure, finance, security, competition suggests a research agenda and possible improvements in
and regulation variables have been shown to significantly survey design.
This paper--a product of the Research Support Unit, Development Research Department--is part of a larger effort by the
World Bank to use enterprise surveys to identify constraints on productivity and growth in developing countries. Policy
Research Working Papers are also posted on the Web at http://econ.worldbank.org. For information, contact jdethier@
worldbank.org.
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Produced by the Research Support Team
Explaining Enterprise Performance in Developing Countries
with Business Climate Survey Data
Jean-Jacques Dethier,
Maximilian Hirn
and
St�phane Straub 1
JEL Codes: L5, O4, O12
Keywords: Investment Climate. Growth and Productivity. Economic Development.
1J.J. Dethier (jdethier@worldbank.org) and M. Hirn (mhirn@worldbank.org) are with DEC, The World
Bank, Washington, D.C. and S. Straub (stephane.straub@univ-tlse1.fr) with Arqade, Toulouse School of
Economics, Toulouse.
Introduction
This is a survey paper of the literature discussing the impact of the business climate on
productivity and growth in developing countries. In recent years, an unprecedented data
collection effort has yielded a set of comparable enterprise surveys covering close to
70,000 firms from over 100 countries in all continents. As a result, a number of studies
have started to analyze the impact of the business climate variables contained in these
surveys on different dimensions of firm performance. The general aim of this literature is
to generate policy prescriptions based on the identification of the main constraints facing
firms. Although many of these studies identify relevant constraints, contradictory or
fragile results are also found, pointing to some weaknesses in the methodology applied in
some papers as well as in the original survey questionnaire design itself.
The objectives of this paper are to review the literature, take stock of the lessons learned,
highlight strengths and shortcomings, and propose potential improvements. To do so, we
start by providing a theoretical framework to think about the impact of the business
climate on productivity and growth in developing countries (Section 1). We then survey
the existing empirical literature that is based on investment climate survey data. We
discuss the empirical macro literature in order to put the firm level investment climate
studies into context. The main findings of the micro literature are then outlined, and the
robustness of the results considered (Section 2). Finally, we highlight the main
econometric issues raised by the current literature, put forward a number of ideas to
advance research on the investment climate and suggest possible improvements in survey
design (Section 3).
Section 1. Economic Growth and the Business Climate
This first section describes the general theory of the business climate and outlines a
theoretical model linking critical variables with economic performance and growth. A
number of structural, institutional, and behavioral variables shape and drive economic
growth. The critical variables that collectively define the so-called business or
investment climate are, broadly speaking, (1) infrastructure, (2) access to finance, (3)
security (absence of corruption and crime) and (4) the regulatory framework, including
competition policies and the protection of property rights. The main hypothesis of the
investment climate literature is that the business climate affects activity throughout the
economy--particularly incentives to invest. An improvement in the business climate
increases returns to current lines of activity and so increases investment in these. It also
creates new opportunities � for example, through trade or access to new technology. It
influences the psychology of entrepreneurs � the Keynesian `animal spirits � affecting
their assessment of whether innovation will pay off. It puts competitive pressure on firms
that have enjoyed privileged positions as a result of import or other protection, or special
access to government officials. As a result of greater competition, it may cause some
firms, perhaps those closer to technological frontiers, to shine--even as others fail.
2
Given the complexity of effects that changes in the business climate elicit, different
firms, industries, and regions will be affected in different ways. Moreover, business
climate�fueled growth is not simply a shift toward some technological frontier.
Developing countries must overcome or reduce all kinds of obstacles to efficiency,
dynamic and otherwise, without any illusions that the economy will soon arrive at a
frontier. Indeed, changes in the investment climate may have their most crucial impact far
from the technological frontier.
A weak business climate, on the other hand, may not only discourage investment, it can
also lead businesses to take costly or counterproductive steps to defend themselves from
the consequences of its weaknesses. If social order and control are weak, firms typically
have to invest heavily in defensive measures such as private security (as in parts of Latin
America or the former Soviet Union). If the power supply is unreliable, firms will invest
in their own generating capacity (as in many parts of South Asia). If it is difficult to get
goods through or to ports, trade is discouraged (as in many countries in Sub-Saharan
Africa) and larger, more costly inventories are held. Many such constraints on
development are not quickly or easily reversed.
To formalize the idea of an economy in which the business climate significantly impacts
output and productivity � the key hypothesis in the recent literature � it makes sense to
present a simple aggregate model that explicitly includes the business climate, using an
approach that is standard in endogenous growth theory.2 It should be noted that aggregate
growth models inevitably involve production possibility frontiers, and as discussed, in
developing economies crucial economic gains may take place far from some hypothetical
technological or efficiency frontier. Moreover, more specific questions may have to
appeal to different and more disaggregated models � each with its own insights � rather
than shoehorning all interesting phenomena into one particular model. The following
aggregate model, however, does provide a broad sense of the main economic
relationships investigated by the literature surveyed below. Even microeconomic studies
that focus exclusively on firm level variations in productivity and growth implicitly
hypothesize a macro relationship of this type in the aggregate.
Suppose that output, Y, is a function of the capital stock, K, labor, L, and the business
climate, as measured by a single variable M. We write this function, where t is time, as:
(1) Y = e t (M) F (K, M)
Note that we implicitly distinguish between the non-infrastructure aggregate capital
stock, K, and the infrastructure capital stock KI, which is a determinant of M.3 As the
output function is written here, the business climate affects both output and productivity
levels, through the function F( ), and the rate of change of output, through the function (
). To keep things simple, we can portray the rate of change of the business climate in the
model as being governed by:
�
(2) M = g (, M(, x), Y),
2See Aghion and Howitt, 1998. Also Stern, Dethier and Rogers, 2005: 207.
3Straub, S. 2008: 6-8
3
where is a vector of policy actions government can take, x is a vector of all other
factors determining the investment climate, and g is the function that converts these
inputs into the change in M. While policy can cause immediate, discrete changes in M,
say if there is a one-off change in the law, it also acts through the rate of change of M.
The rate of change of M can depend on both M itself and the level of income in the
society (see below). We also assume that:
�
(3) K = h(sY, M)
where s is the aggregate savings rate, meaning that the capital stock increases with
savings in the economy. Moreover, as Durlauf et al. (2008) have pointed out, the
investment climate can also affect physical capital accumulation rates by influencing
decisions to invest.
Finally, we model the rate of change in the labor force as a function of population growth
n, and the investment climate. Changes in the investment climate as defined here can
have profound implications for labor force growth, for instance, if regulations with
respect to foreign workers are amended (causing an influx or exodus of foreign labor), or
if infrastructure development connects regions with low labor force participation rates
with more dynamic ones.
�
(4) L = i(n, M)
These four equations describe a dynamic growth model with three state variables, K, L
and M.
The growth rate can be increased by policy, captured in the vector , which improves the
business climate directly, and by shifting the rate of change of M. An increase in M can
increase both the growth rate, through ( ), and the level of productivity, directly through
F( ). Improvements in the business climate could generate further improvements through
political economy mechanisms if they increase the number of people and firms with a
stake in a better climate. For example, if trade reforms create an export-oriented sector of
the economy, that sector may increase pressure for further reforms to trade policy or
trade-related infrastructure. And higher incomes might lead to pressure for an improved
business climate in other ways, as people seek rules governing the protection of wealth or
capital (hence the presence of Y in function g( ) in equation 2). The model could in
principle capture phenomena such as an endogenous business climate and virtuous (or
vicious) circles of growth. We could also generalize the notation to cover a vector of
capital goods, vintage models, many dimensions of the business climate, and so on.
To the extent that (changes in) the business climate affect different firms differently, the
aggregate model, with its reliance on a representative firm, is not adequate. As stressed in
Banerjee and Duflo (2005), such a model can hardly account for the behavior of firms in
a world where either markets or governments fail, or people face psychological
difficulties to take advantage of opportunities. In such a case, the impact of constraints
such as infrastructure limitations, lack of access to finance, or political economy issues
on individual firm's decisions can be analyzed in non-aggregative models. Relatively
simple, distinct, disaggregated models, addressing the relevant constraints of interest, can
4
provide a variety of insights. As Section 2 will demonstrate, the empirical studies that
exploit data from investment climate surveys, while usually implicitly set in a macro
context as the one described by our model, are examples of the added value provided by a
disaggregated, microeconomic approach.
Section 2. Survey of Recent Enterprise-Level Business Climate Studies
Firm-level enterprise and business climate data has proved a rich resource for research on
the characteristics and constraints of firms in the developing and transitioning world. This
section surveys the recent literature with a focus on empirical work that exploits the data
to explain firm performance as a function of different aspects of the business climate.
Sub-section 1 places the recent micro-level business climate literature into the context of
the macro-studies that largely preceded it, thus highlighting the place of the firm-level
literature as a whole and the added value it provides. Sub-section 2 discusses the
available datasets. Sub-section 3 outlines the most serious econometric challenges
encountered throughout the firm-level literature. Sub-section 4 then presents the key
results of the literature for the four main sets of business climate variables that have been
investigated: (1) Infrastructure, (2) Financial Constraints, (3) Corruption and Crime and
(4) Competition and Regulation.
1. Firm Level Analyses in the Context of the Macro-Institutions Literature
The microeconomic business climate literature surveyed here is framed by more macro-
oriented analyses. It is useful to briefly contrast the two literatures to put the firm-level
studies into context and better understand their specific potential and advantages. The
macro-literature has provided some interesting insight, but is characterized by a number
of inherent limitations that microeconomic studies can help overcome.
The macroeconomic literature has generally attempted to use cross-country samples to
explain GDP-based outcome variables4 with broad, country-level indicators of
institutional quality, the policy environment and infrastructure. The majority of such
analyses have found significant effects of these variables on economic performance, even
though recent studies have been more cautious in their interpretation of the evidence.
In a review paper for the European Investment Bank, Romp and de Haan (2005) find the
consensus view in the macro-literature to be that `public capital furthers economic
growth'.5 With respect to institutions and the policy environment, Pande and Udry (2005)
speak of `compelling evidence' and a `persuasive case' that
`long run growth is faster in countries that have higher quality contracting
institutions, better law enforcement, increased protection of private property
rights, improved central government bureaucracy, smoother operating formal
4Typically GDP per capita (e.g. in Acemoglu, Johnson & Robinson, 2001), GDP per worker (e.g. in Hall & Jones,
1999), or the growth rates of these two variables (e.g. in Knack & Keefer, 1995, or Mauro, 1995).
5Romp and de Haan, 2005: 52
5
sector financial markets, increased levels of democracy, and higher levels of
trust'6
Likewise, Dollar et al. (2005) state that a `range of empirical studies...find a relationship
between long-run growth...and measures of institutional quality'7. The World
Development Report 2005 underlines that the macro-analyses `generated useful insights �
the most important is that secure property rights and good governance are central to
economic growth.'8
A recent paper by Durlauf, Kourtellos and Tan (2008) is more cautious. Investigating `the
strength of empirical evidence for various growth theories when there is model
uncertainty with respect to the correct growth model'9, the authors judge that `previous
findings on the direct importance of institutions to growth are fragile'10. They do,
however, conclude that there is at least `some evidence that institutions...play a role as
determinants of growth rates' even if `their effect is likely to flow through their influence
on physical capital accumulation rates and not via TFP growth directly.'11 Straub,
Vellutini and Warlters (2008) find some evidence for a positive effect of infrastructure on
growth, especially in poorer countries, but conclude that in general, the `results from
studies using aggregate data lack robustness'12. Romp and de Haan (2005) highlight that
recent estimations of infrastructure elasticities are much lower than earlier calculations
that did not account for endogeneity effects appropriately.13 Some econometric problems,
such as the failure to account for model uncertainty in cross-section studies14, persist in
the literature. Moreover, Romp and de Haan underline the considerable heterogeneity of
the results across economies, arguing that the precise `channels through which
infrastructure affects economic growth'15 are still not understood very well. The
consensus view that a broadly defined `business climate � institutions, infrastructure and
the social environment � significantly affect economic performance is thus qualified by
lingering concerns about the robustness and generality of specific results and the precise
channels through which the estimated effects occur.
The macroeconomic approach is characterized by a number of inherent limitations that
suggest that microeconomic, firm-level analyses are required to achieve more robust
results and more precise policy recommendations. These inherent limitations of the
macro-literature include:
6Pande, R. and C. Udry. 2005: 31.
7Dollar, D., M. Hallward-Driemeier and T. Mengistae. 2005: 22.
8World Bank. 2004: p.25;
9Durlauf, S.N., A. Kourtellos and C.M. Tan. 2008: 329.
10Ibid.:338.
11Ibid.: 342.
12Straub, S., C. Vellutini and M. Warlters. 2008: 23.
13This is primarily because conceptual and econometric problems such as reverse causality or inefficient proxy
variables have been at least partly addressed in more modern studies.
14Romp, W. and J. de Haan. 2005: 57.
15Ibid.: 58.
6
o The explanatory variables at the country level obscure important dimensions of
heterogeneity such as variations across different regions within a country16 and/or
across different types of firms (by firm size, firm age and so on).
o The limited number of countries restricts the sample size of country-level analyses,
especially cross-sectional ones, and thus the robustness of the results.17
o Aggregate business climate indicators are often imprecise, rely on de jure
information, or subjective judgments about the weighting of variable components,
and lack direct input about actual conditions as experienced by affected parties such
as firms.
o Many country level indicators `contain little or no variation over time and thus are
completely or almost indistinguishable from country-, sector- or region-specific
effects that may reflect other features than the business environment.'18
o The instruments most often used consist of geographical or historical pre-conditions
(latitude, colonial history, settler mortality, etc.), which limits the ability of the
empirical models to identify the consequences of institutional change for growth.19
Thus Durlauf et al. (2008) state that `it is most likely the case that the limits to what
information can be extracted from aggregate regressions requires more attention to
microeconomic and historical studies'20. Similarly, Straub (2008) argues that `the main
limitation [of the macroeconomic literature] is ...the fact that the interesting questions
cannot be addressed with data at that level of aggregation'21. Pande and Udry (2005)
highlight that `this [macro] literature has served its purpose and is essentially complete'
and argue for the necessity of `empirical research based on micro-data in development
economics' to `make progress'22. The World Development Report 2005 underlines that
these `limits [of macroeconomic analyses] inspired the search for more disaggregated
evidence on the quality of a location's business climate and ...the impact of that climate
on the investment decisions and performance of firms.'23
The crucial pre-requisite for finding `more disaggregated evidence' is the availability of
raw disaggregated data. The following section will introduce the main firm-level datasets
on which almost the entire micro-literature on business climate is based.
16Dollar, Hallward-Driemeier and Mengistae. 2005: 2.
17Ibid.
18Commander and Svejnar. 2007: 3.
19Pande and Udry. 2005: 8.
20Durlauf, Kourtellos and Tan. 2008: 344.
21Straub, 2008: p.35.
22Pande and Udry. 2005: 3 and 31.
23World Bank. 2004: 21.
7
2. Datasets
Overview of Existing Datasets
Before the 1990s, standardized firm-level business surveys spanning multiple countries
were practically non-existent. This began to change with an initial series of largely self-
contained projects which carried out business surveys for certain sets of countries and
with various thematic scopes.
Four key projects of that period were sponsored by the World Bank: First, a first set of
Africa-focused surveys carried out from 1992 to 1995 by the Africa Regional Program on
Enterprise Development (RPED); second, the first round of the Business Environment
and Enterprise Performance Survey (BEEPS) for 22 transition countries in 1999; third,
the World Business Environment Surveys (WBES), implemented for 80 countries and the
West Bank/Gaza territories from late 1998 to early 2000; fourth, a number of Firm
Analysis and Competitiveness Surveys (FACS) in the Development Economics Research
Group (DECRG). While these projects yielded unprecedented and highly useful data for
the countries and issues they were designed for, they suffered from limited comparability
amongst each other due to differing questionnaire designs and priorities.
The key development of the early 2000s was a push for greater standardization in order to
build up a single, centralized database of comparable business climate surveys from
around the world. For this purpose, a set of core questions was `pooled and
consolidated'24 from the earlier surveys. This set of core questions became the crucial
component of the new, standardized business climate questionnaires known as
Productivity and Business climate Surveys (PICS). In a specific country survey, around
50-60% will consist of the core modules (some 80 questions), the rest of nationally
specific ones that can be added flexibly to the core instrument depending on each
country's data needs. The core instrument was also partly incorporated into the latest
rounds of surveys that had started earlier, for instance BEEPS, the second and third round
of which contain most of the core PICS questions.
Launched in 2001, 25 the new PICS surveys have been used to acquire detailed firm-level
data in 15 to 20 countries a year. The results have been collected in a central database
(www.enterprisesurveys.org) along with those of earlier, comparable projects such as
BEEPS II and III. All surveys in this database are now commonly referred to as
Enterprise Surveys (ES),26 although the old terminology (PICS, BEEPS etc.) persists to
some extent. The database currently holds information from almost 70,000 firms from
24http://iresearch.worldbank.org/InvestmentClimate [`About ICS', 07/22/2008]
25Hallward-Driemeier, M. and R. Aterido. 2007. [http://www.businessenvironment.org/dyn/be/docs/158/Hallward-
Driemeier.pdf, 07/21/2008]: 5.
26Not to be confused with the World Business Environment Surveys (WBES) mentioned above, which were a one-off
project in 1999-2000.
8
over 100 countries in six different regions. Aterido et al. (2007) have outlined key
features of the database in a recent paper: 27
`The median sample size is 350 firms, with several large countries having
substantially larger samples...The sample of firms in each country is stratified by
size, sector and location...The unit of analysis is the "Establishment" in the
manufacture and service sectors. Most firms are registered with local authorities,
although they may be only in partial compliance with labor and tax authorities.'
The core questions are generally answered by the manager or owner of the establishment
in face-to-face interviews. Accounting data may be provided by the establishment's
accountant and/or human resource manager. Some countries have attached nationally
specific modules answered by workers (for instance the Thailand 2007 PICS survey).
Among the earlier surveys, there is still some variation of the core questions, so that
comparative analyses of multiple business climate variables may require a focus on a
subset of the total database. Aterido et al. suggest a highly comparable subset of around
50,000 firms in 80 countries.28
Structure and Content of the Core Business Climate Survey Instrument
The standardized core survey instrument is organized into two distinct parts.29 The first
part provides general information about the firm and the business climate it faces. The
second part collects accounting information such as production costs, investment flows,
balance sheet information and workforce statistics. The questions about the firm and the
business climate in the first part include:
- General information about the firm: age, ownership, activities, location.
- Sales and supplies: imports and exports, supply and demand conditions, competition.
- Business climate constraints: evaluation of general obstacles
- Infrastructure and services: power, water, transport, computers, business services
- Finance: sources of finance, terms of finance, financial services, auditing, land
ownership
- Labor relations: worker skills, status and training; skill availability; over-employment;
unionization and strikes
- Business-government relations: quality of public services, consistency of policy and
administration, customs processing, regulatory compliance costs (management time,
delays, bribes), informality, capture.
- Conflict resolution/legal environment: confidence in legal system, resolution of credit
disputes
- Crime: security costs, cost of crimes, use and performance of police services
- Capacity, innovation, learning: utilization, new products, planning horizon, sources of
technology, worker and management education and experience.
27Hallward-Driemeier, M.; and R. Aterido. 2007: 20. See also: Aterido, R., M. Hallward-Driemeier and C. Pag�s.
2007: 10-11.
28: Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 10.
29PICS Implementation Manual (Nov. 2003) http://iresearch.worldbank.org/InvestmentClimate/Help/pics_manual.pdf
[07/21/2008]
9
Both subjective perceptions of managers and objective data on various business climate
indicators are recorded.
Box 1: Basic infrastructure variables � subjective and objective
With respect to basic infrastructure, there is one key subjective perceptions variable in the core
survey. It includes three indicators: electricity, transport and telecommunications.
� Rate whether the following issues are a problem for the operation and growth of your
business on a five point scale from `No Obstacle' up to `Very Severe Obstacle': (a)
Telecommunications, (b) Electricity, (c)Transportation (d) ... [14 other non-infrastructure
issues incl. customs/trade regulation, labor regulation etc.]
There are also a number of objective indicators:
� During how many days last year did your establishment experience the following service
interruptions, how long did they last, and what percent of your total sales value was lost last
year due to: (a) power outages or surges from the public grid? (b) insufficient water supply?;
(c) unavailable mainline telephone service?
� Does your establishment own or share a generator? If yes, what percentage of your
electricity comes from your own or a shared generator?
� What share of your firm's water supply do you get from public sources?
� What percentage of the value of your average cargo consignment is lost while in transit due
to breakage, theft or spoilage?
� Does your enterprise regularly use e-mail or a website in its interactions with clients and
suppliers
� Based on the experience of your establishment over the last two years, what is the actual
delay experienced (from the day you applied to the day you received the service or approval)
and was a gift or informal payment asked for or expected to obtain each of the following?
(a) A mainline telephone connection, (b) An electrical connection, (c) A water connection;
(d)... [three other non-infrastructure issues]
Specific national surveys may add infrastructure questions to augment the core-instrument. Moreover,
some changes to the core instrument have been made over time, thus some surveys include additional
indicators, in the case of infrastructure for instance:
� What is your average cost of a kilowatt-hour (KwH) of electricity from the public grid?
� If yes [on generator ownership], what was the generator's original cost to your establishment?
There has been considerable debate about possible weaknesses of subjective, perception-
based indicators compared to objective, quantitative data. Concerns have been raised
whether subjective data may be vulnerable to `waves of pessimism and euphoria', to
inconsistencies across regions and countries because firms compare themselves to
different benchmarks (so called "anchoring effects"30), or to managers' inability to form
accurate subjective estimates.31 For instance, managers may fail to separate internal
weaknesses of the firm (e.g. inability to provide proper documentation) from external
business climate constraints (e.g. inefficient bureaucracy. These problems are a specific
concern when conducting econometric estimations based on cross-sectional data, and
30See for example Bertrand and Mullainathan, 2001.
31Gelb, Ramachandran, Kedia-Shah and Turner. 2007: 2.
10
addressing them may require the use of panel data to control for individuals' or firms'
fixed effects.
Exploring such concerns, Gelb et al. (2007) examine subjective data yielded by the core
Enterprise Survey perceptions-question cited in Box 1. They conclude that while
`perceptions of critical business climate constraints may not always correspond fully to
"objective" reality', firms `do not complain indiscriminately' and response `patterns
correlate reasonably well with several other country-level indicators related to the
business climate'. Likewise, Aterido et al. (2007) underline that:
`subjective rankings are highly correlated with objective measures in 16 of the 17
variables [and] also significantly correlated with external sources, including
Doing Business indicators. Pierre and Scarpetta (2004) use 38 countries and
confirm that countries with more restrictive labor regulations are associated with
higher shares of firms reporting labor regulations as constraining'32
However, even if objective and subjective measures are significantly correlated, it is
important to remember that the latter remain prone to bias. For example, a study by
Olken (2006) compares corruption perceptions among villagers in Indonesia with
objective measures of corruption in road construction projects. It shows that although
subjective and objective measures are positively correlated, there are also systematic
individual-level biases in the latter. Similar issues are very likely to arise in firm-level
surveys.
In spite of these problems, subjective indicators can still play a useful role in identifying
important constraints through descriptive statistics. For instance, Carlin, Schaffer and
Seabright (2006) have highlighted the ease with which a subjective ranking of constraints
by firms allows a comparison of the importance of different constraints. This is not
readily possible with objective indicators that measure various elements of the business
climate in variable-specific units.33 For instance, it is much easier to directly ask firms to
rank the perceived severity of the constraint posed by the power supply relative to
corruption, rather than trying to rank it based on two objective measures such as the
number of power outages relative to the amount of bribes paid. Carlin et al. also argue
that while over-optimism or pessimism may affect estimates of the absolute level of
measured constraint severity, there is no reason to think that average differences between
constraint rankings are likely to be biased. Thus subjective data may be helpful to shed
light on the relative importance of different constraints within economies. However, even
if they can play an important complementary role, subjective indicators are probably less
useful than objective ones in standard econometric analyses.
32Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 11-12.
33Carlin, W., M.E. Schaffer and P. Seabright. 2006: 13.
11
3. The Enterprise-Level Literature on Business Climate: Recent Results
This subsection summarizes the most important results of the recent business climate
literature which relates firm performance to investment climate indicators. Given that
many studies have very specific and limited samples, one must be careful before drawing
general conclusions. However, a large variety of samples can be shown to yield
essentially similar or complementary results. The subsection is structured by types of
constraints, looking in turn at `Basic Infrastructure' (Electricity, Telecommunications,
Transport, Water), `Financial Constraints', `Corruption and Crime' and `Competition and
Regulation'. In each case, a summary of relevant descriptive statistics precedes a review
of the regression results.
Basic Infrastructure
Carlin, Schaffer and Seabright (2006) have analyzed descriptive statistics based on
subjective indicators from 55 Enterprise Surveys.34 As reported in Box 1, the main
perceptions question of the ES core instrument asks about three basic infrastructure
indicators � electricity, telecommunications and transportation. Among the three,
electricity emerges as the most important perceived infrastructure problem.35 It is viewed
as a particularly severe problem in the poorest countries of the sample, including 9 out of
10 African nations36, 4 out of 5 South Asian countries37, Kosovo and Albania. Gelb et al.
(2007) confirm that electricity constraints decrease in perceived severity as GDP per
capita rises.
Perceived Severity of Electricity Constraint (0 to 3 scale) by Country Income Groups 38
34 Carlin, Schaffer and Seabright (2006) calculate a relative and an absolute measure of the importance of particular
constraints. The relative measure calculates the importance of each constraint relative to the average perceived
constraint severity of the country in question. Each constraint is then ranked by the total number of countries in which
it is perceived as more severe than average. The absolute measure ranks constraints by the number of countries for
which the constraint is ranked higher than the average perceived severity of all constraints in all countries (2.2).
35 Ibid.: p.15.
36 The exception is South Africa
37 The exception is Oman
38 Carlin, Schaffer and Seabright. 2006: 6. Country income classification is from the World Bank, July 2005, based on
GNI. Note that Carlin et al. state that the graph is based on a 0 to 3 scale, however, the actual question in the core
survey uses a 0 to 4 scale. It is not clear whether this is an error or whether Carlin et al. amended the original scale.
12
With the exception of Ireland, transport is rated as an above average constraint only in `a
handful of poor or war torn economies'. Telecommunications does not appear at all in
Carlin et al.'s main ranking, possibly indicating the extent to which the rapid spread of
mobile phones has reduced the importance of this constraint. It also underlines the need
to update the objective survey questions referring to `mainline' telephone services only.
An analysis of infrastructure statistics based on objective indicators is provided by Lee
and Anas (1992) and Lee, Anas and Oh (1996; 1999) for Nigeria, Indonesia and
Thailand. Their analyses are not based on the standard Enterprise Survey (ES) data, but
on three dedicated surveys that were carried out in the late 1980s and early 1990s. The
infrastructure information they collected, however, is very similar to the one available in
the ES database.39
Lee et al. focus on the incidence of public infrastructure deficiencies, the extent of private
provision responses to these deficiencies and the costs thereby imposed. They find large
variations in the availability and quality of public infrastructure across the three
countries, across regions within the countries and across firm sizes. In general, Nigeria
tended to have a worse public infrastructure performance and a correspondingly higher
incidence of private provision than Thailand and Indonesia. The authors speculate that
the comparatively worse problems in Nigeria are related to the country's (then) tighter
restrictions on private provision arrangements. Aimed at protecting inefficient public
suppliers, these restrictions prevented the emergence of private infrastructure provision
regimes more efficient than the simple `one firm, one generator' model. 40 The authors
argue in favor of `[opening] up the markets for power, water and other various
infrastructure services' in order to improve service reliability and reduce system
congestion.41 However, Lee et al. do not explore this suggestion in detail and do not
discuss possible implications and problems (such as equity-efficiency trade-offs).
A key finding of the study is the disproportionate way in which smaller firms are affected
by infrastructure deficiencies. In all three countries, small firms relied far more on the
public supply than larger ones and were thus subject to the bulk of the power failure
incidents.42 Lee et al. argue that this was not because the `burden of poor electricity or
water supplies is less per unit of output' for smaller firms, but rather due to economies of
scale in private provision of electricity and water, which means it is relatively cheaper for
larger firms to avoid the public system and provide their own power and water. This
result finds support in the much broader analyses of 80+ Enterprise Surveys by Aterido et
al. (2007) who examine the deviation of perceived constraints from the average ranking.
They find that small firms report electricity as a greater relative constraint than larger
firms.43 The authors make the intuitive argument that smaller firms are `more likely to be
39The key descriptive statistics can be calculated in both cases: number of firms that own a generator; number of firms
that own a private well; production time and sales value lost due to public infrastructure interruptions et cetera.
Stratification and sample sizes are also similar in both cases (a couple of hundred enterprises per country and year,
stratified by industries, regions/cities and firm size).
40Lee, K.S., A. Anas and G. Oh. 1999: 2141.
41Ibid.: 2149.
42Ibid.: 2138. See also World Bank (2004): Box 6.10.
43Note that Figure 6.4 in the World Development Report 2005 shows that a greater percentage of large than small
firms rank infrastructure constraints as `major' or `severe'. However, this is not inconsistent with smaller firms
13
in areas without access to electricity or to be dependent on an unreliable public grid'44,
given that they lack the scale economies to operate a generator efficiently. Lee et al.
(1999) suggest that since a very large share of new jobs in developing countries are
created by small firms, the negative impact of infrastructure deficiencies on employment
creation is potentially huge.45 Regrettably, the potential links between the disproportional
infrastructure problems of small firms and job creation are not followed up by the authors
and no tentative cost estimates in terms of jobs are provided.
Lee et al.'s analyses demonstrate the concrete insights that can be gained from descriptive
firm-level statistics of infrastructure variables. Their specific results are somewhat dated
by now, but the concerns they address remain relevant today. The Enterprise Survey
database, which contains similar statistics for more than 100 countries, is an extremely
valuable new resource in that respect.
While descriptive statistics are useful to establish basic facts, regression analyses have
provided a more detailed view of the relationship between infrastructure and firm
performance indicators. Escribano and Guasch (2005) use ES data from Guatemala,
Honduras and Nicaragua to calculate ten different measures of firm productivity. The
productivity measures are then regressed on a number of controls as well as a broad array
of objective business climate variables of which four are infrastructure indicators - the
log of average duration of power outages, the log of the number of days to clear customs
for imports, the log of shipment losses as fraction of sales as well as a dummy for internet
access. The regressions are carried out both for a pooled sample and for each of the three
countries separately. The authors find that in all regressions the infrastructure variables
always have the expected signs, and the vast majority is significant. The authors are
confident enough in their results to interpret the elasticities straightforwardly, noting that
for the pooled sample:
`a one percent increase in the average duration (hours per day) of power outages
decreases productivity between 0.02 and 0.1 percent, depending on the
productivity measure used. It mainly affects old plants...a one percent increase in
the fraction of shipment losses will decrease productivity between 1.23 and 2.53
percent. This is most important in old and small firms...firms with access to
internet are between 11% and 15% more productive [than] those firms without'46
Escribano and Guasch go to great length to avoid bias and inconsistency in their analysis.
They take care to avoid simultaneity problems, control for country, industry and year
effects with dummies and also including at least two firm characteristics as controls (age
and share of imported inputs). Regressions are run on all variables at a time to avoid
omitted variable bias. Moreover, they use region-industry averages of the IC variables as
instruments to alleviate reverse causality. Escribano and Guasch show that infrastructure
perceiving infrastructure as a greater relative constraint. There may be structural reasons � such as larger firms' greater
demands on the various elements of the business climate � that on average lead larger firms to report higher absolute
constraint rankings in the various categories.
44Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 15.
45Lee, K.S., A. Anas and G. Oh. 1999: 2140.
46Escribano, A.; and J. Guasch. 2005: 55.
14
has a significant impact on productivity, explaining some 9 percent of it in total � the
second highest percentage after `red tape, corruption and crime'47.
However, a number of points should be noted when interpreting this result. Firstly, the
huge impact of internet access on productivity suggests that this dummy functions as a
proxy for better equipped, higher-technology firms rather than just representing internet
access per se.48 This advises caution when deriving policy interpretations. For instance, if
the large productivity improvements captured by the internet dummy are actually related
to much broader technological differences between firms, then prioritizing putting
internet-connections into every firm could be a misguided step. Secondly, one should be
careful not to generalize too much from these results and recall that they are based on
only three (lower) middle income countries. The relationships may be considerably
different in very poor countries where subjective measures indicate that electricity is a
more severe problem.49
Hallward-Driemeier, Wallsten and Xu's (2006) study underlines that specific aspects of
infrastructure are greater constraints in some countries than in others. They use data from
an Enterprise Survey of China to regress different firm performance indicators (TFP,
investment rate, sales growth, employment growth) on a number of controls50 and on
objective business climate indicators, including two infrastructure ones: loss of sales due
to transportation or power problems, and share of employees that use computers51. The
authors find `no evidence that physical infrastructure affects firm performance' but `the
impact of technological infrastructure [on productivity] appears to matter significantly'.
Hallward-Driemeier et al. conclude that this `roughly' conforms to their knowledge of
China, which has relatively few bottlenecks in roads and power after the recent build-up
of physical infrastructure, but still faces a `binding constraint' in terms of `technological
infrastructure'.
In a sample of five Eastern European countries (Kyrgyz Republic, Moldova, Poland,
Tajikistan, Uzbekistan), Bastos and Nasir (2004) obtain a similar result as Escribano and
Guasch. Regressing TFP on three controls (firm age, export status and ownership) and
three aggregate business climate indicators (`competition', `infrastructure' and `rent
predation'52), they find that all three IC measures have the expected sign and are
significant at the 1% level. `Infrastructure' accounts for the second largest share of the
variation in firm-level productivity, behind `competition' but before `rent predation'.
What undermines the results of Bastos and Nasir, however, is the fact that their `two-
47Ibid.: 73.
48Likewise, the other indicators may also capture some additional variation from unobserved variables. In essence, this
means that there may still be some omitted variable bias that distorts the estimated parameters, or alternatively, there
may be no (or almost no) bias, but the included variable may only be an instrument/proxy for the actual cause of the
productivity effect.
49For example, the marginal impact of power outage duration on productivity could be much higher after certain
thresholds, which may not be reached in middle-income countries.
50Ownership (share of ownership that is domestic private; share of ownership that is foreign), logs of firm age, city
population and city GDP per capita
51in each case city-industry averages are employed to lessen reverse causality
52The authors use principal component analysis to obtain their aggregate indicators. See Annex 2 for the specific
indicators on which the aggregate measures are based.
15
step' estimation is vulnerable to simultaneity bias as pointed out by Escribano and
Guasch. Moreover, they do not control for country effects. Thus, while their indicators
may capture some genuine cross-country differences in the three business climate
categories, they are also vulnerable to bias if other cross-country effects (such as trade
policy, political instability etc.) influence productivity and are also correlated with their
indicators.
Dollar et al. (2005) use a sample of Enterprise Surveys from Bangladesh, China, India
and Pakistan to regress total factor productivity on a number of controls and five
objective business climate variables, including the two infrastructure indicators `log of
the cost of power losses as a % of sales' and log `time required to obtain a phone line'.
The authors find that even after controlling for firm characteristics, geography variables
and country-level effects, power losses have a significantly negative effect on
productivity. This seems to confirm the importance of electricity in poor countries and
more generally the significance of infrastructure for explaining variation in productivity.
However, the telecommunications variable has a perversely positive and significant
coefficient, but this counter-intuitive result `is not robust across all specifications'.
Results based on firm performance variables other than productivity by and large confirm
the significant role variations in infrastructure play in explaining differences in firm
success. Reinikka and Svensson (2002) use the 1998 Ugandan Industrial Enterprise
Survey for a sophisticated short study of the effects of poor infrastructure and deficient
public services on the level of private investment. The authors first construct a model to
garner hypotheses for the empirical analysis that follows. The model characterizes two
decisions � whether a firm buys private, complementary infrastructure capital, and how
much it invests in non-complementary, productive capital in the next period. The first
empirical estimation thus runs a probit regression of `ownership of a generator' on the
number of days of power interruptions from the public grid, the firm's employment size,
the percentage of foreign ownership, a dummy indicating whether the firm exports part of
its output, firm profit and age. The model hypothesis is confirmed as public power
outages show a significantly positive relationship with the probability of owning a
generator.53 Moreover, a firm is significantly more likely to own a generator if it is a
larger firm, an exporter or has a higher percentage of foreign ownership. With respect to
the investment decision, the empirical analysis also confirms the model hypotheses. For
firms without a generator, investment is found to be negatively related to the number of
days of power interruptions. However, `an increase in the number of days lost has no
statistically significant effect on investment for firms with their own generators'54. This
comes at a cost, however, for if the public power supply is good (i.e. conditional on few
lost days), firms that have installed expensive private electricity infrastructure invest less
than firms without a generator. On the whole, Reinikka and Svensson (2002) conclude
that their analysis of Ugandan firm level data shows that poor public capital `significantly
reduces productive investment by firms'.55 They deduce that a poorly functioning public
53In the empirical analysis the actual public power outcomes are taken as a proxy for ex-ante beliefs of firms that
invest in complementary private infrastructure capital in the first period.
54Reinikka, R. and J. Svensson. 2002: 65.
55Ibid.: 67.
16
infrastructure sector is likely to hinder a private supply response to more general
macroeconomic reforms.
Aterido et al. (2007) use employment growth as their dependent variable and regress on a
large number of controls and objective business climate variables including three
infrastructure ones: log of days with power outages, log of % of sales lost due to power
outages and log of days without water. Their analysis is based on a sample of at least 80
Enterprise Surveys, considerably more than other papers. With respect to infrastructure,
the authors find a significantly negative effect of power outages on employment growth
for medium sized firms, and at least the expected sign for small and large firms.
Hallward-Driemeier and Aterido (2007) carry out a similar study with particular focus on
Africa which produces interesting results. They regress employment growth on a variety
of controls and business climate variables, including losses from power outages (% of
sales), frequency of outages and whether a firm owns a generator as infrastructure
indicators. They can confirm that a higher incidence of power losses is associated with a
negative impact on employment growth. Interestingly, the authors find that African firms
seem to have adapted to this problem to some extent so that given the frequency of
outages, African employment growth is stronger than expected relative to the rest of the
world. This has partly to do with the comparatively high incidence of generator
ownership in Africa, which reduces the impact of power shortages from the public grid.
However, another reason seems to be that a higher frequency of outages seems to have
contributed to a disproportional concentration of African employment growth in very
small firms, which are less capital intensive and thus less vulnerable to power outages in
terms of employment effects.
Export status � whether or not a firm exports goods abroad � has also been studied as a
dependent variable. As trade integration has often been associated with economic growth
� both at the firm and country level � authors have been interested to test whether a
relationship between business climate and export status can be found. Dollar et al. (2006)
draw on a sample of firm level surveys from Bangladesh, Brazil, China, Honduras, India,
Nicaragua, Pakistan and Peru to carry out probit estimations of whether the probability
that a randomly chosen firm in a particular city exports is connected to business climate
indicators (controlling for country and a number of other variables). They include `losses
from power outages' as their infrastructure indicator and find it to have a negative and
significant impact on the probability of exporting in all their specifications.
Datta (2008) uses Enterprise Survey data from India to investigate the effects of a
highway improvement program on the production efficiency of firms. Datta's paper is
particularly interesting for the way he exploits panel data to avoid the reverse causality
problem stemming from the fact that better economic performance may attract more
infrastructure, rather than more infrastructure causing firms' efficiency to improve. Datta
argues that if
`the precise route of the highway was not manipulated to include some
intermediate areas (counties, districts, cities) and exclude others based on factors
correlated with the outcomes of interest, then the highway construction can be
17
treated as exogenous to the areas that the highway runs through...This allows for
a difference-in-difference estimation strategy, where changes in relevant
outcomes for affected firms are compared to the corresponding outcomes for
firms whose location precluded their directly benefiting from the highway
program'56
Datta argues that since the highway improvement program in question used the most
direct routes between its destinations, and because no `opting out' was possible and no
realignments carried out, the areas in between the destinations can indeed be viewed as a
quasi-random selection of locations with existing highways to which the upgrade
`treatment' was applied. Datta finds that firms that profited from the upgrade held
significantly lower inventories, became less likely to report transportation as a major or
severe problem, and showed a greater propensity to change suppliers between the two
years (suggesting they found more suitable ones). This is interpreted as evidence that
`improved highways facilitated productive choices', `eased the extent to...which
transportation infrastructure constrains firms' and allowed them to `produce more
efficiently'57.
Papers that find no significant effects of infrastructure indicators on firm performance are
in the minority, and generally have very specific samples or clear methodological
limitations. For instance, Commander and Svejnar (2007) use a sample of BEEPS
surveys (round II and III) to regress firm revenues on a number of controls and subjective
business climate variables, including a composite `infrastructure' one based on the
perceptions question reproduced in Box 1. They do find that perceived infrastructure
constraints have a negative and significant effect on firm revenue � but only without
controls for country fixed effects. The authors conclude quite generally that only country
effects (due partly to differences in infrastructure, partly to other unobserved
heterogeneities) have an impact, while within-country differences in infrastructure do not.
However, this seems like a premature conclusion given the significant within-country
effects found in many other studies, that the sample is limited to Eastern Europe and
Central Asia and only subjective indicators were used.
Fisman and Svensson (2005) use an Ugandan dataset to test whether firm sales growth is
explained by corruption and taxation, controlling for a number of variables including an
composite index of `public services' (electricity, water, telephone, waste disposal, paved
roads). Although this index has the expected positive sign (a higher index number
standing for better infrastructure), it is not individually significant.
Competition and Regulation
The view that competition and entry should promote efficiency and prosperity `has now
become...common wisdom worldwide.'58 Generally speaking, this view would lead us to
expect a positive effect of competition on firm performance, and a negative effect of
(excessive) regulation. Studies based on business climate survey data have predominantly
56Datta, S. 2008: 2-3. [Italics are the author's]
57Ibid.: 4 and 15.
58Aghion and Griffith. 2005: 1.
18
focused on using existing local or cross-country differences in regulatory outcomes to
explain firm performance. There is still a lack of panel analyses which could estimate the
impact of changes (i.e. reforms) in the same regulatory framework(s) over time. This is
partly due to the still small number of Enterprise Survey rounds, which restricts the size
of available panel datasets..
Carlin et al.'s (2006) examination of descriptive statistics based on subjective Enterprise
Survey variables from some 60 countries shows that anti-competitive practices are ranked
as of greater than average importance in all of their country groups. Alan Gelb et al.
(2007) look at two types of (subjective) indicators of regulation � tax administration and
labor regulations. The results yielded by their subjective Enterprise Survey data are quite
intuitive. Like corruption and crime, tax administration is perceived as a problem
primarily in developing countries in the middle income range.59 As one moves further up
in the income level, labor regulations are more often perceived as severe constraint. Gelb
et al. argue that policies `become more serious determinants of the business climate at
this stage, largely because the state has stronger capacity to implement them.'60 The
World Development Report 2005 cites some evidence that larger firms spend more time
dealing with officials and are inspected more often61:
Bastos and Nasir's (2004) analysis of BEEPS data regresses productivity on an aggregate
`competition' variable based on four subjective and one objective indicator62. They find a
strongly positive and significant impact of competition on productivity. Indeed,
competition is shown to explain a far larger part of the variation in firm performance than
their `rent predation' and `infrastructure' variables. They conclude that this
`finding suggests that the relatively quick steps governments can take to
increase competition will have a big payoff in firm performance � even as
the slow, expensive process of upgrading infrastructure takes place. It also
59Gelb, A., V. Ramachandran, M. Kedia-Shah and G. Turner. 2007: 13-14.
60Ibid.: 15.
61World Bank. 2004: 100.
62i) & ii) Importance of domestic competition for decisions to: a. introduce new products, b. reduce costs; iii) & iv)
Importance of foreign competition for decisions to a. introduce new products, b. to reduce costs; v) Number of
competitors in main product line;
19
indicates that high levels of fixed investment...will not be enough to spur
growth'63
As stated above, however, their conclusions must be qualified by possible problems due
to simultaneity bias and the failure to check for robustness of their results with country
dummies.
Commander and Svejnar (2007) include an objective measure of competition � `more
than 3 competitors' � in their regressions on otherwise subjective business climate
indicators, which are not significant once country effects are controlled for. However, the
competition variable is shown to have a very robust positive and significant impact on
firm revenue even if country effects are controlled for.
Escribano and Guasch (2005) do not have a real competition variable, but do check for an
impact of regulation on productivity with the variable: `Number of days spent in
Inspection and regulation related work'. It is shown to impact productivity negatively in a
significant way in almost all specifications. The authors ascribe some 12% of the
variation in productivity to their combined `Red tape, corruption and crime' variables,
making it the most important set of variables. Beck, Demirg��-Kunt and Maksimovic's
(2005) analysis of 54 WBES surveys includes a subjective `degree of legal obstacles'
variable, which is found to be negative and significant. Hallward-Driemeier, Wallsten
and Xu's (2006) examination of China includes the regressor `city-industry mean of the
share of senior managers' time in dealing with regulatory requirements', which has a
negative and significant impact on sales and employment growth, but is not a significant
explanatory variable for firm productivity.
Hallward-Driemeier and Aterido (2007) highlight that the impact of the regulatory
environment is not necessarily only negative. Regulations can have positive sides as well,
especially if they are consistently enforced. Hallward-Driemeier and Aterido find that
consistent enforcement of regulations has a clear positive association with employment
growth in most of the developing world, though it is insignificant for Africa. The variable
indicating management's time spent dealing with the authorities has an ambiguous
impact. In the full sample, it is positive in general but less so in Africa. The authors argue
that while there seem to be some benefits associated with accessing public services, at
`about 15 percent of management time, the marginal impact of additional interactions
with the government is negative'64. On the other hand, pure red tape such as unnecessary
delays in customs have significantly negative effects.
Aterido et al. (2007) use the same two indicators representing the regulatory framework:
the relatively objective `% of management's time dealing with government regulations'
and the subjective ranking of the statement `officials interpretation of regulations is
consistent'.65 The authors find that consistency of enforcement has a positive and
significant effect in general, which is particularly marked for small firms. The authors
63Bastos, F. and J. Nasir. 2004: 24.
64Hallward-Driemeier, M., S. Wallsten and L.C. Xu. 2006: 9.
65See Tables 6,7, 8 and 9.
20
also obtain a generally positive effect of managements' time spent dealing with
authorities, which they interpret as representing the benefit from obtaining public goods.
As before, a quadratic version of the indicator suggests that the benefits are offset as the
overall time mangers spend with officials rises beyond a certain point.66 There is also
some evidence that large firms profit less from management time spent dealing with
authorities67, which may partly be due to the fact that the average time large enterprises
spend dealing with officials is longer than for small firms.
Financial Constraints
Examining subjective indicators, Carlin et al. (2006) find the cost of finance ranked
above average in severity in all of their country groups68. In particular, the cost of finance
is the highest ranked constraint in the African country group69. Alan Gelb et al.'s (2007)
study of subjective perceptions data shows that the perceived severity of `access to
finance' constraint declines with country income level.70 The World Development Report
2005 includes the following graph, based on a similar sample, which also indicates a
higher reported severity of the financial constraint in poorer countries71:
Within countries, descriptive statistics of subjective data indicate that access to finance is
particularly problematic for less productive firms.72 Size also seems to influence the
ability of obtaining credit from banks. Using 54 datasets from the World Business
Environment Surveys (WBES), Beck, Demirg��-Kunt and Maksimovic (2005) regress a
subjective firm level indicator of financial access73 on firm size and a number of specific
country-level institutional effects. It is found that even after controlling for a country's
66Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 21.
67Ibid.: 38. Table 6, Column 1.
68Africa (10), South Asia (5), East Asia (7), Latin America and the Caribbean (7), OECD Europe (6), Central and
Eastern Europe (8), South Eastern Europe (8) and the CIS (11)
69South Africa is an exception; the constraints ranked most highly there are labor regulation, skill shortages,
macroeconomic stability and crime
70Gelb, A., V. Ramachandran, V., M. Kedia-Shah and G. Turner. 2007: 13.
71World Bank. 2004: 115.
72Carlin, W., M.E. Schaffer and P. Seabright. 2006: 6 (Figure 1b).
73`How problematic is financing for the operation and growth of your business: no obstacle (1), a minor obstacle (2), a
moderate obstacle (3), or a major obstacle (4);
21
institutions, smaller firms report significantly higher financial obstacles than large
firms.74 Likewise, Aterido et al (2007) find, based on objective Enterprise Survey data,
that smaller firms have significantly less access to different forms of finance even when
controlling for age, export status, ownership and industry.75 In line with this, the business
climate survey data indicates that small firms tend to finance a much smaller share of
their investments with formal credits.76 Bigsten et al. (2003) confirm that in their sample
of African countries, close to two-thirds of micro firms are credit constrained, but only 10
percent of large firms. The authors also find that regressions controlling for other
important factors such as expected profitability and indebtedness, `the likelihood of a
successful loan application varies with firm size' in the same way.77
Most of the studies employing regression analysis to examine the relationship between
firm performance and the business climate include indicators representing measures of
financial access. The most pertinent results will be outlined below, keeping in mind the
general methodological criticisms of key papers already mentioned in the context of
infrastructure indicators.
Beck, Demirg��-Kunt and Maksimovic (2005) regress firm sales growth on a number of
controls as well as one summary78 and 11 specific79 subjective indicators of financial
obstacles. Entered alongside the legal and corruption summary variables, the financial
obstacles main indicator is found to have a negative and significant effect on firm growth.
The authors also find 6 of the 11 specific financial constraints indicators having a
negative and significant impact, however, because each is entered individually, it is likely
that omitted variable bias distorts these results. It should also be noted that Beck et al. do
not calculate location-industry averages, rendering their estimates vulnerable to reverse
causality at the firm level, as does their reliance on subjective indicators.
Aterido et al.'s (2007) main objective financial constraint indicator is `percentage of
investments financed externally' and they take greater care than Beck et al. (2005) to
reduce endogeneity at the firm level.80 Attempting to explain employment growth, they
find that in general, a higher share of investments financed externally is associated with
greater employment growth. Hallward-Driemeier and Aterido (2007) also find a
significantly positive impact of their financial access variable `share of investment
financed externally' on firms of all sizes:
74Beck, T., A. Demirg��-Kunt and V. Maksimovic. 2005: 150.
75Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 36 (Table 4) and 8.
76World Bank. 2004: 116.
77Bigsten, A. and M. S�derbom. 2005: 11.
78`How problematic is financing for the operation and growth of your business: no obstacle (1), a minor obstacle (2), a
moderate obstacle (3), or a major obstacle (4);
79i) `Are collateral requirements of bans/financial institutions no obstacle (1), a minor obstacle (2), a moderate obstacle
(3), or a major obstacle (4)?'; ii) `Is bank paperwork/bureaucracy no obstacle....?'; iii) `Are high interest rates no
obstacle...?'; iv) `Is the need of special connections with banks/financial institutions no obstacle...?'; v) `Is banks' lack
of money to lend no obstacle...?'; vi) `Is the access to foreign banks no obstacle...?'; vii) `Is the access to nonbank
equity/investors/partners no obstacle...?'; viii) `Is the access to specialized export finance no obstacle...?'; ix) `Is the
access to ease finance for equipment no obstacle...?'; x) `Is inadequate credit/financial information no obstacle...?'; xi)
`Is the access to long term finance no obstacle...?';
80Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 20.
22
`A 10 percent increase in the share of investments financed through bank loans
(equivalent to doubling the average share) is associated with a 3 percent increase
in employment growth. This result is robust to alternative measures of finance,
including formal bank financing of investment to trade credit among firms.'81
By contrast, Commander and Svejnar (2007) cannot find a significant effect of their
subjective `cost of finance' variable on firm revenue in their dataset from Eastern Europe
and Central Asia.82 Dollar et al. (2005) find no significant effect of their financial
indicator `access to overdraft facility' on productivity of firms in the garment industry,
but in an expanded sample they do find a significant and strongly positive impact of the
variable on annual sales growth. Dollar et al. (2006) find a relatively robust positive
relationship between `access to overdraft' and the probability that a firm is an exporter.83
Hallward-Driemeir, Wallsten and Xu's (2006) study of the Chinese business climate
yields no significant link between a variety of firm performance indicators and bank
access. As pointed out above, this result may largely be due to the peculiar nature of the
Chinese state owned banking sector which tends to be relatively inefficient and
subsidizes unsuccessful enterprises for political reasons. Thus, it is not particularly
surprising that access to finance has no systematic impact on variations in firm
performance in China.84
Escribano and Guasch (2005) do not include a variable indicating availability of credit to
the firm in their equations. However, they do include one dummy indicating whether the
firm is a publicly listed company, and another dummy that represents whether the firm is
externally audited or not. Both indicators are significantly positively related to firm
productivity (between 11.5 and 17 percent).
In a large cross country sample, Carlin et al. (2006) find the coefficient of their subjective
indicator of `cost of finance' to be negative in both between and within-country
regressions, and also significant in the latter. However, they argue that this result is not
primarily due to financial constraints impacting productivity, but rather the fact that
inherently less productive firms are rationally denied credit by financial managers (and
complain about it). This reasoning makes intuitive sense, but Carlin et al.'s actual results
are undermined by weaknesses in their methodology. For instance, if the above
endogeneity bias is suspected as a problem, the authors should also have tested the
relationship between firm productivity and the average cost of finance in the firms'
location and industry (rather than for the firm itself), thus alleviating the firm-level
endogeneity. Moreover, the endogeneity mechanism they highlight seems more likely to
impact subjective data and seems less relevant for objective data on which much of the
significant relationships found in the literature are based. Finally, their method of
regression on one indicator at a time is also sub-optimal because it is likely to cause
omitted variable bias. While the finance-endogeneity effect they highlight is interesting
and should be kept in mind when interpreting results of financial indicators, their
81Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 8.
82Except when the variable is entered separately from the other IC variables, which renders it vulnerable to omitted
variable bias
83Dollar, D., M. Hallward-Driemeier and T. Mengistae. 2005. 1507.
84Hallward-Driemeier, M., S. Wallsten and L.C. Xu. 2006: 645.
23
particular econometric results to not by themselves undermine the significant
relationships between financial availability and firm performance found in other papers,
which are more pro-active in countering endogeneity problems.
Corruption and Crime
Carlin et al.'s summary of subjective indicators from a large number of enterprise surveys
identifies crime and corruption as problems reported primarily in less developed
countries. Of the two, corruption is more commonly perceived as problematic:
`Crime and/or corruption show up as important constraints in all groups of
countries except the OECD: crime in only one-quarter of countries and
corruption in 70%.'
The analysis of subjective firm level data by Gelb et al. (2007) shows that concern about
corruption and crime tends to peak in the middle of developing countries' income range.
The authors interpret this as showing that once economies overcome utmost poverty and
the most basic limitations related to infrastructure, finance and macroeconomic stability,
problems of low administrative and bureaucratic capacity come to the forefront of firms'
concerns.
Recent studies that examine the relationship between firm performance and business
climate indicators generally find significant effects for corruption and crime indicators.
Fisman and Svensson (2005) use their Ugandan firm-level dataset for a study focused on
corruption and its effect on growth. Their OLS and IV regressions of sales growth on a
corruption indicator and a variety of controls show a
`strong, robust, and negative relationship between bribery rates and the short-run
growth rates of Ugandan firms, and [...] the effect is much larger than the retarding
effect of taxation.'
Escribano and Guasch's (2005) study of productivity in Guatemala, Honduras and
Nicaragua includes the explanatory variables `payments to deal with bureaucracy faster
as % of sales' and `number of criminal attempts suffered'. The coefficient for the number
of crimes suffered shows the expected negative sign and is significant. However, the size
of bribe payments has a robust positive relation with productivity. This may mean that
firms that can afford paying (more) bribes will tend to be more productive in the first
place and/or reap productivity advantages from their payments. However, in terms of
policy implications it certainly does not imply that the incidence of corruption in the three
countries should be seen as positive for productivity in general. Rather, the authors argue,
`it is clear that there is room for improvement in the administrative procedures followed
in the three countries...so that no more arbitrary administrative gains in productivity
[arise] from bribes of firms.'85 Still, the difference in the direction of the sign of the
corruption variable in the two studies is somewhat puzzling, and further research would
be required to reveal the source of the difference (which could be genuine cross-country
85Escribano, A. and J. Guasch. 2005: 54.
24
variation in the mechanisms of corruption, or related to the somewhat different regression
specifications).
An interesting result with respect to the sign of the coefficient of corruption indicators is
provided by Aterido et al.'s (2007) analysis of employment growth in a very large sample
of some 80 Enterprise Surveys. They find a significantly negative effect of their bribe
dummy, as well as alternative corruption indicators, on the growth of small, medium and
large firms. However, the coefficient is positive for micro-firms.86 This probably
indicates that micro-firms find it easier to escape the attention of corrupt officials and
therefore tend to grow faster relative to larger firms if the industry-location averages of
corruption are higher. In their study of the Chinese business climate, Hallward-Driemeier,
Wallsten and Xu find that objectively measured corruption87 matters `a great deal' for
sales growth. Reducing `the mean score of corruption by one standard deviation...has a
positive effect on sales growth by...6 percentage points'.88 However, no significant effect
of corruption can be shown for other firm performance indicators such as productivity
and employment growth. In Beck, Demirg��-Kunt and Maksimovic's (2005) main
regression of sales growth on a number of controls and three summary subjective
business climate indicators, corruption obstacles, unlike financing and legal ones, are not
significant, although the coefficient does have the expected negative sign. The authors
suspect this to be due to multicolinearity, in the sense that the `impact of corruption on
firm growth is captured by the financial and legal obstacles.'89 In an Eastern European
and Central Asian sample, Bastos and Nasir (2004) also find a significantly negative
effect of their `rent predation' aggregate variable, which measures a combination of
corruption and regulation. However, the rent predation variable explains less variation of
productivity than the infrastructure and competition measures.
Section 3. Lessons and Ways Forward
In the previous section, we have reviewed the results of firm-level studies that relate
enterprise performance to various objective and subjective business climate indicators,
along with a series of controls for variables such as firm characteristics, industry and
country effects.
As has been seen, these studies have provided new evidence for one of the central
assertions of the 2005 World Development Report, namely, that a good business climate
`drives growth by encouraging investment and higher productivity'90 At least four
elements of the investment climate � infrastructure, finance, corruption and crime, and
competition and regulation � have been shown to significantly impact firm performance.
86Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 22.
87Their variable is the city-industry share of the corruption score, which is constructed as the principal component of
two variables: the ratio of bribes to sales plus the share of contract value used as bribe to get a business contract
88Hallward-Driemeier, M., S. Wallsten and L.C. Xu. 2006: 644.
89Beck, T., A. Demirg��-Kunt and V. Maksimovic. 2005: 151-153.
90World Bank. 2004: 2 and 8.
25
Even as problems remain, the firm-level studies have already improved on the macro-
literature in a number of respects. Numerous firm-level papers have now shown that
important within-country heterogeneity exists. Variation in local business climate does
indeed matter for explaining differences in firm performance. Much the same point is
made by single-country, regional business climate studies such as the one on China by
Hallward-Driemeier et al. (2006). Moreover, the much larger sample sizes made possible
by going to this disaggregated level allow for more robust results than in the macro-
studies. The information obtained from the business climate surveys is also much more
detailed and practical than aggregate indicators, allowing, for instance, insights about the
variation of business climate effects across regions and different types of firms.
Building on this, a rich research program becomes possible. Below, we start by outlining
the econometric issues and limitations of the current literature, and summarize the lessons
derived from them. We then highlight what in our view are the most promising areas for
future research. Finally, we open the debate on potential improvements in the design of
existing survey questions.
1. Econometric Lessons from the Current Literature
The standard approach in the current literature based on enterprise survey data has been
to use regression analysis to identify which � if any � business climate indicators
determine firm performance and to what extent. Almost universally, the basic
specification of these regressions has been:
Firm Performance = 1 + 2(IC Indicators) + 3(Firm Characteristics) + 4(Additional Controls) +
When interpreting results from these regressions, it is important to keep some basic
characteristics and limitations of the approach in mind. Significant coefficients of the
explanatory variables are only obtained if there is variation in these variables. Thus, the
results presented above efficiently pinpoint existing bottlenecks that explain observed
variations in firm performance, but they are less useful for identifying universal
problems. For instance, Hallward-Driemeier, Wallsten and Xu (2006) find that access to
banking services is not a significant determinant of firm performance in China. However,
this does not mean that increasing the availability and efficiency of financial services is
unimportant for improving Chinese productivity. As the authors point out, `it only means
that the state-owned banking sector has not contributed significantly to regional firm
growth'91. In other words, the fact that Chinese state-owned banking has not had a
systematic impact on firm performance means that it does not show up as a determinant
of actual variation therein. But the common lack of efficient banking services may still
be responsible for sub-optimal levels of firm performance throughout China. This
methodological issue is particularly relevant for studies with small samples, because
expanding the number of observations will tend to introduce more variation and thus
allow more general statements.
91Hallward-Driemeier, M., S. Wallsten and L.C. Xu. 2006: 645.
26
A related issue is that of `camels and hippos' raised by Hausmann and Velasco (2005)
and discussed in Gelb et al. (2007) and others. All results are necessarily based on the
answers of existing firms that were interviewed. However, if one only interviews those
present (`camels in the desert'), one may miss the crucial constraint (`water') of those
who have not entered (`no hippos in the desert'). In other words, a self-selected sample
may imply a lack of variation in the explanatory variables that prevents us from noticing
a critical constraint. However, in their study of perceptions data, Gelb, Ramachandran,
Shah and Turner (2007) argue that such self-selection is hardly ever complete (e.g.
hippos can be expected to live in a water hole at the edge of the desert), and that firms
that choose to enter in spite of serious constraints (which may force them into costly
evasive actions), will perceive them as particularly serious and thus introduce
econometrically significant variation.92 Still, as it stands it is important to recognize that
the econometric model above only informs us about the effect of constraints on the
sample of existing firms. It is sometimes argued that the more interesting issue is rather
the underlying industrial structure (e.g. the camel/hippo ratio in the desert) which should
give away the most important constraint (i.e. the absence of hippos indicates that the
main constraint is the lack of water). This, however, could only be addressed with
completely different models such as "entry" models.
Another general methodological problem is that of multicolinearity. If regressors are
correlated with each other, estimates will be inefficient and, as Bastos and Nasir (2004)
point out, it may be impossible to `know the importance of any one particular indicator
since it may be serving as a proxy for other, more relevant variables'93. This is a
particular problem with the business climate data, as many indicators are closely related.
For instance, the prevalence of e-mail usage may largely move with the quality of
electricity supply. This counsels caution when interpreting very specific indicators, and
emphasizes the importance of choosing a good regression specification. To some extent,
variables such as `prevalence of e-mail' should be seen as proxies for broader
infrastructure factors. Bastos and Nasir's (2004) solution is to explicitly aggregate a
number of specific indicators into broader measures (`infrastructure, `competition' etc.),
in order to get clearer results at the loss of some (presumably misleading) detail.
Endogeneity � the correlation between the explanatory variables and the error term � is
even more serious than multicolinearity, because it causes not only inefficiency and
interpretative difficulties, but bias and inconsistency of the estimates. The presence of
endogeneity undermines the validity of estimated relationships between business climate
indicators and firm performance.
It is unrealistic to assume that firm level business climate indicators are exogenous for a
number of reasons. First, a major endogeneity problem arises if relevant explanatory
variables are mistakenly omitted from the regression equation and also correlated with
relevant included regressors. If this is the case, the estimated parameters of the included
regressors will pick up some of the impact on the dependent variable of the omitted
variables with which they are correlated. This will distort the estimates of the parameters
92Gelb, A., V. Ramachandran, M. Kedia-Shah and G. Turner. 2007: 27-29.
93Bastos, F. and J. Nasir. 2004: 10.
27
of the included regressors, because they will now capture both their own effect and part
of that of the correlated omitted variables.
Second, better subjective and objective investment climate indicators may be associated
with better performing firms not because they cause such firms to be more productive,
but on the contrary, because `an inherently more efficient firm can work within the
exogenously given environment to reduce inspections, power losses or days for customs
clearance or phone lines.'94 Similarly, not only may better suited business environments
cause firms to be more efficient, but inherently more efficient firms may also be more
likely to have the necessary resources to identify and (re-)locate to better suited
environments. At the aggregate level, inherently more prosperous regions may have
greater political clout to obtain infrastructure and other business climate improvements
from government. If one cannot fully control for these reverse causality factors, estimates
of the effect of the investment climate on firm performance will be biased.
To limit the endogeneity bias problem, the current firm level business climate literature
suggests the following measures:
o Regressions on single business climate indicators are very likely to produce biased
and inconsistent parameter estimates due to omitted variables. A sufficiently broad
array of indicators and controls should therefore be used in regression equations. The
selection of regressors should go from general to specific.95
o Objective indicators are generally preferable to subjective ones as explanatory
variables, because they are less vulnerable to measurement error and reverse
causality. 96
o Using location-industry or industry averages instead of (or as instruments for) the
firm-level objective indicators can help alleviate endogeneity due to reverse
causality.97 The idea is that while better region-industry investment climate indicators
should explain variation in firm performance, individual firm performance has
virtually no impact on the average-indicator. This alleviates direct reverse causality.
o Country-level effects should be controlled for, either with country dummies or more
specific country-effects variables, to avoid a contamination of the IC coefficients with
correlated but unobserved country level effects on firm performance.
o A simple two-step estimation procedure that estimates TFP as the residual of a
production function and then attempts to explain TFP with IC indicators is potentially
vulnerable to simultaneity bias. The problem is that in most cases the inputs of the
production function will be correlated with the investment climate indicators, because
94Dollar, D., M. Hallward-Driemeier and T. Mengistae. 2005: 9.
95Carlin, W., M.E. Schaffer and P. Seabright. 2006: 36.
96Bertrand, M.; and S. Mullainathan. 2001: 71. Also: Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007: 12.
97See for instance: Hallward-Driemeier, Wallsten and Xu, 2006. Dollar, Hallward-Driemeier and Mengistae, 2006.
Escribano and Guasch, 2005. Commander and Svejnar, 2007. These authors use location-industry averages that exclude
the respective firm.
28
the investment climate influences not only productivity per se, but also input choices
of firms. This means that in the production function regression, the error term (i.e.
TFP) is likely to be correlated with the regressors (labor, capital etc.) via the
investment climate, leading to bias. If possible, this approach should thus be avoided.
Escribano and Guasch (2005) suggest alternative procedures.
o In the absence of panel data, an approach similar to Miguel, Gertler and Levine
(2005) might be useful to alleviate some endogeneity problems. They try to explain
industrialization (measured as the growth in manufacturing employment) at the
district level in Indonesia over a 10 years period with social capital indicators at the
beginning of the period (but find no effect!). A similar approach could be taken with
indicators from ICA surveys.
Finally, note that a complete assessment of the results in the literature we have surveyed
so far would also require that we look more closely at the issue of the quality and
relevance of the performance proxies used as dependent variables (productivity or profit,
or sales growth etc.). While entering into the details of the literature on this topic would
take us beyond the scope of this paper, let us note that in general measures of firm-level
productivity are much more likely to run into problems and generate biases, as the very
construction process of these variables make them likely to be correlated with policy
shocks and managerial decisions (Katayama, Lu and Tybout, 2006). This is not to say
that alternative proxies (e.g. profit, sales or employment growth) are completely free of
these problems (see Del Mel, McKenzie and Woodruff, forthcoming) but in many cases
they appear to be preferable.
2. The Research Agenda
There remain a number of areas in which additional research could bring interesting
results. At the theoretical level, we need to develop a better understanding of the link
between firms' choices and the business climate, especially in developing countries. That
means modeling decisions about investment, R&D, employment and so on, which hinge
on the type of constraints revealed by the existing surveys (things like credit constraints,
infrastructure bottlenecks, level of competition in goods and labor markets, volatility of
macroeconomic conditions, entry costs, commitment and enforcement problems or
information issues).
The type of modeling used in the literature on firms' choices of formality might be useful
here.98 Together with tools from industrial organization and contract theory, this
approach should provide a good basis to formalize insights on market behavior in
developing countries. Additionally, results could then be used to understand the very
different shapes of firms' distributions we see across countries, for example in terms of
size, productivity or exporting behavior, and guide the empirical applications.
98For instance: Rauch 1991; Straub 2005; De Paula and Sheinkman 2008.
29
At the empirical level, some of the most interesting insights in the firm-level business
climate literature have come from recent studies that look at interactions of business
climate indicators with firm characteristics or with each other. For instance, Hallward-
Driemeier and Aterido (2007) interact IC measures with firm sizes to obtain more
detailed results on the impact of the business environment on the performance of
different types of firms. Honorati and Mengistae (2007) examine the interplay of
regulation, infrastructure, financial constraints and corruption. They obtain some
interesting results, for instance that all three aspects have significant influence on Indian
industrial growth, yet their effect depends on the incidence of corruption. Most existing
firm level studies have not considered these types of interactions, and more work in this
direction could deliver interesting outcomes that lay the groundwork for more precisely
targeted policy recommendations.
A key research goal highlighted by a number of authors is that once more surveys rounds
become available, proper panel data regressions could test for the impact of changes in
the business climate on productivity, factor returns, and growth. 99 For instance, whereas
current microeconomic studies predominantly aim to assess the variation in firm
performance due to local and cross-country variations in existing constraints, panel data
could allow an assessment of the impact of changes (reforms) in the shape of different
constraints on firm performance. However, with only 3 survey rounds available at most,
it is still relatively early for these types of studies.
Even the standard methodological approaches have not yet made full use of the large
Enterprise Survey database. For instance, no best-practice study (properly accounting for
endogeneity) of the relationship between firm productivity and the business climate has
been carried out for the full, up-to-date Enterprise Survey database.100 Likewise, `little
analysis is available on the impact of infrastructure on manufacturing firm productivity'
in Africa.101 More specifically still, there is a lot of scope to carry out detailed country
studies such as that of China by Hallward-Driemeier, Wallsten and Xu (2006) or that of
India by Honorati and Mengistae (2007) and Amin (2007). It is generally easier to
correctly interpret econometric results in single-country studies, because outcomes are
easier to connect to real-life circumstances and complementary data.
As noted above, a recent paper by Durlauf et al. (2008) argued that the effect of
institutions is `likely to be through their influence on proximate growth determinants
(factor accumulation, in this case) rather than through their effects on technological
innovation.' It would be interesting to explore this question further in a micro-context. So
far, only a few papers have used measures of capital (or human capital) accumulation as
dependent variable and there has been no systematic comparison to the results for total
factor productivity.
99See for instance: Dollar, D., M. Hallward-Driemeier and T. Mengistae. 2005: 30.
100This has only been done with employment growth as dependent variable. See: Aterido, R., M.Hallward-Driemeier
and C. Pag�s, 2007.
101Bigsten, A. and M. S�derbom. 2005: 15. However, there has been at least one paper on the relationship between
employment growth and the investment climate in Africa.
30
Future studies should make sure to extensively test the robustness of their results and if
possible improve on the methodology in a more fundamental way. This is because even
the current `best-practice' precautions against endogeneity � such as using location-
industry averages as instruments of the firm-level indicators, regressing on multiple IC
indicators at a time and controlling for the current country, region and industry effects �
leave regressions vulnerable to inconsistency and bias, as Carlin et al. (2006) and others
point out. For instance, as highlighted above, location-industry averages are used as
instruments to alleviate endogeneity stemming from reverse causality. Yet, such
endogeneity effects can persist at a more aggregate level as well, because of policy
endogeneity and endogenous placement decisions of firms. For instance, using industry-
location averages one may find a strong relationship between the performance of firms
and the average quality of telecom services of a specific industry and region. However, as
Carlin et al. (2006) point out, regions `that are prosperous for a variety of other reasons
for which it is not realistically possible to control econometrically also happen to have
higher levels of telecom services.' To counter this effect, a recent paper by Hallward-
Driemeier, Wallsten and Xu (2006) has included additional city information and sector
dummies to at least help `control for those more macro issues that affect both the IC
variable and the firm'.102 Nevertheless, the inability to sufficiently control for all factors
implies that the endogeneity problem is likely to persist to some extent. In light of this, it
is clear that the need arises for more creative instrumental strategies. Again some
examples can be found in the literature, for example in Duflo and Pande (2007), who use
geographical data to instrument for the endogenous placement of infrastructure, or Datta
(2008) and Gibson and Rozelle (2003) who take advantage of the seemingly exogenous
placement of road works in specific contexts to assess their impact.
Finally, it should be noted that the underlying assumption of most of the firm-level
literature is that changes in the business climate which improve firm performance will
translate into broad social benefits. Regression results based on data provided by firm-
managers are thus often straightforwardly translated into policy advice, e.g. to increase
competition and lower regulation. In order to reduce the risk of any negative impacts, it
may be worthwhile to consider possible competing interests when deducing policy
recommendations. For instance, regulations may impact firm productivity negatively but
provide benefits to non-managerial social groups.
3. Improving Questionnaire Design
At a fundamental level, it may also be worthwhile to re-think some of the Enterprise
Survey questions which determine the raw data on which all analyses are based. For
instance, in the era of mobile phones � which are particularly important in many
developing countries � the focus on mainline telephone services is anachronistic and
misleading. With regard to infrastructure indicators, Straub (2008) makes a number of
suggestions for more detailed questions such as firms' access to alternative transport
modes (railways, airports, roads etc.) or the ownership of vehicles.103
102Hallward-Driemeier, M., S. Wallsten and L.C. Xu, 2006: 640.
103Straub, S. 2008: 39.
31
There appear to be many holes in the information provided. In electricity for example,
most information is on quality (outages and cost thereof) but basic information on cost
and availability of service would be needed: average cost of a kilowatt-hour (KwH) of
electricity from the public grid / cost of generators. Similarly, in water, information is
sought on the sources of provision, but it should be complemented with the respective
average unit costs.
In transport, data on the possibility to access different types of services (roads, railroads,
etc.), together with an assessment of their unit cost and quality, and the ownership of
different types of vehicles, would make it possible to assess the significance of the
transport mix chosen by firms. In the case of telecommunications, as mentioned above,
mobile telephony is completely absent from existing surveys. Here again, data on access,
unit cost and quality of service would be necessary. One could also wonder why
questions geared at the use of internet are restricted to the sub-sample of service firms.104
Finally, in all cases, a few key dimensions need to be added. First, information on the
institutional nature of service providers and regulatory arrangements would be crucial
from a policy perspective. Moreover, in a context where the geographical dimension is
increasingly recognized to be important,105 data need to be spatially referenced.
Obviously, the practical task of gathering this type of data (including in particular several
hours spent with directors and managers of firms, who often have imperfect knowledge
about the things they are asked to report) implies a trade-off between being exhaustive
and the quality of the data collected. However, the fact that such exercises are bound to
aim at second best results should not impede that we try to address the most obvious
shortcomings of current surveys.
104Similarly, questions on innovation are restricted to manufacturing firms.
105See Straub, 2008. Also, Gibson and McKenzie, 2007.
32
Bibliography
Acemoglu, D., S. Johnson and J.A. Robinson. 2001. "The Colonial Origins of
Comparative Development: An Empirical Investigation." American Economic Review
91(5): 1369-1401.
Aghion, P. and R. Griffith. 2005. Competition and Growth � Reconciling Theory and
Evidence; Cambridge: MIT Press.
Amin M. 2007. "Labor Regulation and Employment in India's Retail Stores." Policy
Research Paper 4314. World Bank, Washington D.C.
Aterido, R., M. Hallward-Driemeier and C. Pag�s. 2007. "Investment climate and
Employment Growth: The Impact of Access to Finance, Corruption and Regulations
Across Firms." IZA Discussion Paper No. 3138.
Aterido, R. and M. Hallward-Driemeier. 2007. "Impact of Access to Finance, Corruption
and Infrastructure on Employment Growth."
http://www.businessenvironment.org/dyn/be/docs/158/Hallward-Driemeier.pdf,
[07/21/2008].
Banerjee, A. and E. Duflo. 2005. "Growth Theory through the Lens of Development
Economics." in: Philippe A. and S. Durlauf, S. (eds.). 2005. Handbook of Economic
Growth. Elsevier: 473-552.
Bastos, F. and J. Nasir. 2004. "Productivity and the Business climate: What Matters
Most?." Policy Research Working Paper 3335. World Bank, Washington D.C.
Beck, T., A. Demirg��-Kunt and V. Maksimovic. 2005. "Financial and Legal Constraints
to Growth: Does Firm Size Matter?" The Journal of Finance LX(1): 137-177.
Bertrand, M. and S. Mullainathan S. 2001. "Do People Mean What They Say?
Implications for Subjective Survey Data." The American Economic Review 91(2): 67-72.
Bigsten, A. and M. S�derbom. 2006. "What Have We Learned From a Decade of
Manufacturing Enterprise Surveys in Africa?" The World Bank Research Observer.
21(2): 241-265.
Brice�o-Garmendia, C., A. Estache and N. Shafik. 2004. "Infrastructure Services in
Developing Countries: Access, Quality, Costs and Policy Reform." World Bank Policy
Research Working Paper 3468. World Bank, Washington D.C.
Bruhn, M. 2008. "License to Sell: the Effect of Business Registration Reform on
Entrepreneurial Activity in Mexico." Policy Research Working Paper No.4538. World
Bank, Washington D.C.
33
Carlin, W., M.E. Schafferand P. Seabright. 2006. "Where are the Real Bottlenecks? A
Lagrangian Approach to Identifying Constraints on Growth from Subjective Survey
Data." CERT Discussion Paper 2006/04. Centre for Economic Reform and
Transformation, Edinburgh.
Commander, S. and J. Svejnar. 2007. "Do Institutions, Ownership, Exporting and
Competition Explain Firm Performance? Evidence from 26 Transition Countries." IZA
Discussion Paper No. 2637. Institute for the Study of Labor, Bonn.
Datta, S. 2008. "The Impact of Improved Highways on Indian Firms." World Bank,
Washington D.C.
Del Mel, S., D. McKenzie and C. Woodruff. 2007. "Measuring Microenterprise Profits:
Don't Ask how the Sausage is Made." Policy Research Working Paper No.4229. World
Bank, Washington D.C.
De Paula A. and J. Sheinkman. 2008. "The Informal Sector." PIER Working Paper, Penn
State University.
Dollar, D., M. Hallward-Driemeier and T. Mengistae. 2005. "Business climate and Firm
Performance in Developing Economies." Economic Development and Cultural Change
54: 1�31.
___ . 2006. "Investment climate and International Integration." World Development
34(9): 1498-1516.
Durlauf, S.N., A. Kourtellos and C.M. Tan. 2008. "Are Any Growth Theories Robust?"
The Economic Journal 118(527): 329-346.
Escribano, A. and J. Guasch. 2005. "Assessing the Impact of the Business climate on
Productivity Using Firm Level Data: Methodology and the Cases of Guatemala,
Honduras, and Nicaragua." Policy Research Working Paper 3621. World Bank,
Washington D.C.
Fisman, R. and J. Svennson. 2007. "Are Corruption and Taxation Really Harmful to
Growth? Firm Level Evidence." Journal of Development Economics 83: 63�75.
Gelb, A., V. Ramachandran, M. Kedia-Shah and G. Turner. 2007. "What Matters To
African Firms? The Relevance of Perceptions Data." Policy Research Working Paper
4446. World Bank, Washington D.C.
Gibson J. and D. McKenzie. 2007. "Using Global Positioning Systems in Household
Surveys for Better Economics and Better Policy", The World Bank Research Observer
22(2): 217-241.
34
Gibson, J. and S. Rozelle. 2003. "Poverty and Access to Roads in Papua New Guinea."
Economic Development and Cultural Change. 52: 159-185.
Hall, R.E. and C.I. Jones. 1999. "Why Do Some Countries Produce So Much More
Output per Worker than Others?" Quarterly Journal of Economics 114(1): 83-116.
Hallward-Driemeier, M., S. Wallsten and L.C. Xu. 2006. "Ownership, business climate
and firm performance: Evidence from Chinese firms." Economics of Transition 14(4):
629-647.
Hausmann, R. and A. Velasco. 2005. " Slow Growth in Latin America: Common
Outcomes, Common Causes?" Manuscript, October.
Honorati, M. and T. Mengistae. 2007. "Corruption, the Business Environment, and Small
Business Growth in India." Policy Research Working Paper 4338. World Bank,
Washington D.C.
Kaplan, D., E. Piedra and E. Seira. 2007. "Entry regulation and business start-ups:
Evidence from Mexico." Policy Research Working Paper 4322. World Bank, Washington
D.C.
Katayama, H., S. Lu and J. Tybout. 2005. "Firm-level Productivity Studies: Illusions and
a Solution." Mimeo Penn State University.
Knack, S. and P. Keefer. 1995. "Institutions and Economic Performance: Cross-Country
Tests Using Alternative Institutional Measures." Economics & Politics 7(3): 207-227.
Lee, K.S., A. Anas and G. Oh. 1999. "Costs of Infrastructure Deficiencies for
Manufacturing in Nigerian, Indonesian and Thai Cities." Urban Studies 36(12): 2135-
2149.
Miguel, E., P. Gertler and D.I. Levine. 2005. "Does Social Capital Promote
Industrialization? Evidence from a Rapid Industrializer." The Review of Economics and
Statistics 87(4): 754-762.
Mauro, P. 1995. "Corruption and Growth." The Quarterly Journal of Economics 110(3):
681-712.
Olken, B. 2006. "Corruption Perceptions vs. Corruption Reality." NBER Working Paper
12428. National Bureau of Economic Research, Cambridge.
Pande, R. and C. Udry. 2005. "Institutions and Development: A View from Below."
Discussion Paper No. 928, Yale University Economic Growth Center.
35
PICS Implementation Manual (November 2003)
http://iresearch.worldbank.org/InvestmentClimate/Help/pics_manual.pdf [accessed
07/21/2008]
Pierre, G. and S. Scarpetta. 2004. "Employment Regulations through the Eyes of
Employers: Do They Matter and How Do Firms Respond to Them?" IZA Discussion
Paper 1424. Institute for the Study of Labor, Bonn.
Rauch, J. 1991. "Modeling the Informal Sector Formally." Journal of Development
Economics 35(1): pp. 33-47.
Reinikka, R. and J. Svensson. 2002. "Coping with poor public capital" Journal of
Development Economics 69: 51�69.
Romp, W. and J. de Haan. 2005. "Public capital and economic growth: a critical survey."
EIB Papers, 10(1): 40-70.
Stern, N., J.-J. Dethier and F.H. Rogers. 2005. Growth and Empowerment � Making
Development Happen. MIT Press, Cambridge.
Straub, S. 2005. "Informal Sector: the Credit Market Channel." Journal of Development
Economics 78(2): 299-321.
___ . 2008. "Infrastructure and Growth in Developing Countries: Recent Advances and
Research Challenges." Policy Research Working Paper 4460. World Bank, Washington
D.C.
Straub, S., C. Vellutini and M. Warlters. 2008. "Infrastructure and Economic Growth in
East Asia." Policy Research Working Paper 4589. World Bank, Washington D.C.
The World Bank. 2004. World Development Report 2005: A Better Business Climate for
Everyone. World Bank, Washington D.C.
36
Annex: Selection of Papers Analyzing the Relationship between Firm Performance and the Business climate
Paper Dataset Type of Analysis Dependent Independent Variables Results / Criticism
Variable(s)
BEEPS II 1. Estimation of TFP and - ln(TFP) Business climate Variables: Results: The business climate variables have
Bastos and Nasir Extended subsequent regression of TFP the expected signs and are jointly significant.
(2004): on three broad business -Rent predation variable, based on:
(Kyrgyz Republic, climate measures � rent * Amount of unofficial payments Using Kruskal's methodology, competition is
`Productivity and Moldova, Poland, predation, infrastructure and to public officials as % of sales found to explain far more variation in firm-level
Business climate: What Tajikistan, competition � which were * % of senior management time productivity than infrastructure, which in turn
matters most' Uzbekistan) constructed from individual spent in dealing with red tape explains more variation than rent predation.
survey indicators using * Days last year spent on
principal component analysis inspections
Criticism: Authors do not include country
2. Determination of the - Infrastructure variable, based on: dummies (see criticism of Commander and
relative importance of the * Days of interrupted phone serv. Svejnar, 2007). Two step estimation procedure
three business climate * Days of interrupted water serv. vulnerable to simultaneous equation bias as
regressors using the Kruskal * Days of interrupted power serv. outlined by Escribano and Guasch, 2005).
(1987) methodology.
- Competition variable, based on subjective
estimates of importance of domestic /
foreign competition to introduce new
products / reduce costs.
Control Variables: Firm Age, Exports (%
of sales), Foreign Ownership
Beck, Demirg��-Kunt, 54 datasets from 1. To find out whether firm - 32 mostly Business climate variables (subjective on a Results:
Maksimovic (2005): the World size determines perceptions subjective 1 to 4 scale):
Business on financing, legal and business climate 1. `Firms' perception of the financing and
`Financial and Legal Environment corruption constraints to variables in the - Summary financing obstacle corruption obstacles they face relates to firm
Constraints to Growth: Surveys (WBES) doing business, the authors categories - Summary legal obstacle size, with smaller firms reporting significantly
Does Firm Size carry out OLS regressions of `financial', `legal' - Summary corruption obstacle higher obstacles than large firms. In contrast,
Matter' firm-level financing, legal and `corruption'. smaller firms report lower legal obstacles than
and corruption indicators on - 29 subjective and 3 objective IC indicators do larger firms, but these differences are not
firm size, controlling for for more specific constraints within the significant.'
country level financing, legal - Firm growth three summary categories.
and corruption constraints. (percentage 2. When entered individually, all [three
change in firm Controls: Ownership summary] obstacles have a negative and
2. To see whether variations sales over the past (government/foreign); Exporter status; significant effect on firm growth...[entered at
in perceived obstacles can 3 years) number of competitors; industry dummies; the same time] financing and legal obstacles are
37
explain firm sales growth, the country specific dummies (inflation; GDP; both significant and negative, but corruption
authors carry out OLS GDP per capita; GDP growth); loses its significance...'. Entering the 32
regressions of firm growth on individual obstacles each in turn, some of the
IC indicators (each in turn, financing and corruption variables, but none of
not all together), controlling the legal ones are significant. Two of the only
with industry dummies, firm three quasi-objective indicators are significant.
characteristics and country
random effects. 3. The authors find evidence that financial
obstacles have a much greater impact on the
3. To explore how the effects operation and growth of small firms than on that
of IC indicators differ by firm of large firms.
size, a series of size controls
are added as explanatory
variables. Criticism: Very few objective business
climate indicators (only three quasi-objective
ones out of 35). Also, the results on the 32
specific IC variables are based on regressions of
firm growth on each of the variables in turn
Likely omitted variable bias! Authors do not
Dollar, Hallward- Four datasets from 1. GLS and Levinsohn/Petrin For garments Business climate variables: Results: `Business climate matters for the
Driemeier, Mengistae the World Bank production function industry only: level of productivity, wages, profit rates, and the
(2005): Enterprise Surveys estimation of TFP in the - log(custom days export) growth rates of output, employment and capital
Main Database: garments industries of all - TFP - log(custom days import) stock at the firm level � in garments and similar
`Business climate and Bangladesh, countries. TFP is then - Average wage - log (power loss) sectors...'
Firm Performance in China, India and regressed on the logs of a set - Average profit - log (days to get phone)
Developing Pakistan of IC variables and controls. - log (overdraft facility) IC explanatory variables show consistent
Economies' For pooled data joint significance and often individual
2. Regression of factor (all industries): Instrumented by city-sector averages. significance as well.
rewards in garments
industries on the same - Sales growth Control variables (not all in every Results robust to inclusion of country dummies
variables plus firm - Annual growth regression): log of...distance from market; which shows that local business climate
characteristics. The rate of fixed assets distance from port; population; lagged important!
hypothesis is that factor - Annual growth annual sales; lagged age of firm; fixed
rewards will be higher were rate of assets at start of year; last year's
IC is better. employment employment; also country, year and Criticism: Regressions with total factor
industry dummies. productivity potentially vulnerable to
3. Regression of sales growth, simultaneity bias. However, results largely
growth in fixed assets and confirmed with alternative firm performance
growth in employment in all dependent variable.
industries (pooled dataset) on
IC variables and controls.
All regressions carried out for
the full sample, and a sub-
sample of small firms.
38
Escribano and Guasch `Data collected for The paper aims to develop `an - 10 different Business climate Variables: Results:
(2005): ICAs in... appropriate and consistent estimations of firm
Guatemala, econometric methodology to productivity - Red Tape, Corruption and Crime: In the theoretical part of the paper, the authors
`Assessing the Impact Honduras and be used as a benchmark for � No. of days spent in Inspection and highlight that analyses that use a simple two-
of the Business climate Nicaragua' evaluating the impact of IC Regulation related work step procedure to first estimate firm
on Productivity Using variables on productivity at � Fraction of sales undeclared to tax productivity, and then regress this measure on
Firm Level Data: � presumably in the firm level'. The discussed authorities IC variables are likely to suffer from
Methodology and the World Bank econometric methods are then � Payments to deal with bureaucracy simultaneity bias. They propose three different
Cases of Guatemala, Enterprise Surveys applied in the cases of faster (% of sales) methods to estimate productivity.
Honduras and Main Database Guatemala, Honduras and � No. of criminal attempts suffered
Nicaragua' Nicaragua. In the empirical part of the paper, the authors
-Infrastructure: identify four `important categories of business
The econometric analysis � Average duration of power climate (IC) variables...for the case of
consists of a variety of outages (log) Guatemala, Honduras and Nicaragua: (a) red
regressions of productivity � Days to clear customs for tape, corruption and crime; (b) infrastructure; (c)
measures on business climate imports (log) quality, innovation and labor skills; and (d)
indicators and a set of � Shipment losses (% of sales) finance and corporate governance....The
controls. Results are also estimates show consistently the high impact of
analyzed by country, age and � Dummy for internet access business climate on productivity. Overall, it
size of firms. -Quality, Innovation & Labor Skills accounts for over 30 percent of productivity.
� The two most impacting categories are "red
Fraction of computer controlled tape, corruption and crime" and
machinery
� "infrastructure".'
Fraction of total staff engaged in
R & D
� Dummy for ISO quality
certification
� Fraction of total staff with
secondary or higher
� Dummy for Training beyond `on
the job' training
-Finance & Corporate Governance
� Dummy for incorporated
Company
� Dummy for external audit
Instrumented by region-industry averages.
Controls: Age of the firm (log); Share of
imported inputs (fraction); country; firm
size;
39
Fisman and Svensson Uganda Industrial Regression analysis to test for - Sales growth: Business climate variables: Results: Both taxation and bribes are found
(2005): Enterprise Survey a significant impact of [log(sales1997- to have a robust, significantly negative impact
corruption and taxes on firm log(sales1995)]/2 - Reported bribe as share of sales on short-run sales growth; the retarding effect of
`Are Corruption and � presumably in sales growth, controlling for - Reported tax as share of sales bribes is thereby stronger than that of taxes.
Taxation Really World Bank other factors. Foreign ownership has a positive impact on
Harmful to Growth? Enterprise Surveys Instrumented by location-industry avg. sales growth, as does `trade status' at least in
Firm Level Evidence' Main Database one specification.
- Index (0-5) of availability of public
services (elect., water, telephone, waste
disposal, paved roads)
-Index of regulation (log of 1+ percentage
of senior management's time spent dealing
with regulation)
Control Variables: ownership (foreign
>50%); log of firm age; (log of) sales in
1995; trade (firm exports and/or imports);
Dollar, Hallward- Eight datasets Probit regression of export - Indicator Business climate variables: Results: The authors `find that a sound
Driemeier and from the World status on business climate variable of business climate � as reflected in low customs
Mengistae (2006): Bank Enterprise indicators and control whether firm - Three objective IC indicators: days to clearance times, reliable infrastructure, and good
Surveys Main variables (geographic, sector, exports or not clear customs; access to overdraft; losses financial services � makes it more likely that
`Business climate and Database: firm characteristics). from power outages. domestic firms will export, enabling the more
International Bangladesh, productive firms to expand their scale and
Integration' Brazil, China, The aim is to relate business - Indicator - One subjective IC indicator: whether scope.'
Honduras, India, climate to the probability that variable of managers thought government services
Nicaragua, a randomly chosen firm in a whether firm has inefficient. The empirical link is largely robust to the
Pakistan and Peru. particular city exports. foreign ownership inclusion of country dummies (at least joint
Country dummies are used in or not The authors use location averages to significance and some individual significance
some specifications to obtain instrument the variables. remains), showing that local factors matter for
analysis of within-country the IC.
variation. Control variables: distance to international
market; distance to port; population;
population squared; country dummies;
sector dummies; firm size (employment).
40
Carlin, Schaffer and ~60 World Bank 1. Overview of descriptive - TFP (defined for Business climate variables: Results:
Seabright (2006): Enterprise Surveys statistics of subjective IC manufacturing
Main Database indicators firms using TFP 17 subjective indicators of the severity of 1. The descriptive statistics show that physical
`Where are the Real (BEEPS 2002, residuals, or the different business climate constraints on a infrastructure rarely rates highly as a constraint,
Bottlenecks? A 2004 and 2005 and 2. Development of model of firms' self- four/five point scale (see Box 1) problems with licensing and customs affect
Lagrangian Approach PICS from 2000- the firm to predict reported relatively few countries (esp. CIS), crime and/or
to Identifying 2005) relationship between reported technological Control variables: Country dummies; corruption show up as important constraints in
Constraints on Growth constraints and the level) ownership (foreign owned/state owned/new all groups of countries except the OECD, seven
from Subjective characteristics of firms and private owned); big city; dimensions of the business environment that are
Survey Data' ranked as of greater than average importance in
3. Regression of TFP on all country groups: anti-competitive practices,
tax rates and tax administration, access to and
cost of finance, and policy uncertainty and
macroeconomic stability.
2. Regressions yield results largely in line with
model predictions. Between-country regressions
show negative and significant effects of
Telecom, Electricity, Transport, Customs
regulation, Mafia, Land Title and Land Access
indicators. Once country effects are controlled
for, however, customs regulations, transport and
legal system indicators have perverse positive
signs. Authors argue that this is due to
endogeneity bias. Finance has significant
negative impact, as predicted because it does not
have public good characteristics but instead
inherently unproductive firms are rationally
denied credit (and complain about this).
Criticism: Firstly, instead of objective
indicators, the authors use subjective ones which
are particularly vulnerable to the endogeneity
effects they allege. Secondly, their regressions
only use one business climate indicator at a
time, exposing them to omitted variable bias.
Thirdly, the posited relationship between firm
performance and perceived indicator severity
can only be shown for customs regulation and
finance but not for any of the other
disaggregated indicators. When avoiding these
problems, other authors do find non-perverse,
significant effects even when employing country
dummies. Fourthly, Carlin et al. do not seem to
use industry-location averages of their
regressors, which could at least lessen firm-level
endogeneity biases.
41
Hallward-Driemeier, China PIC Survey Regression of four different -Sales growth Business climate variables: Results:
Wallsten, Xu (2006): 2000 (part of firm performance variables on -Investment rate
World Bank largely objective business -Productivity -mean loss of sales due to transport/power - ownership significant, foreign ownership more
`Ownership, business Enterprise Surveys climate indicators (measured -Employment outages so than private domestic
climate and firm Main Database) as city-industry averages) and growth -mean share of labor that uses computers - no evidence that physical infrastructure matters
performance' controls. -mean share of R&D staff in labor significantly, but technological infrastructure
-mean regulatory burden does (expected given that hard infrastructure
-mean corruption such as road and power good in China)
-mean share of non-permanent labor - labor market flexibility weakly significant
-mean bank access - No evidence that average access to finance in a
region and industry affects performance
(expected given inefficiency of Chinese bank
Control variables: ownership (domestic sector)
private/foreign); logs of firm age+1 and -Government regulatory burden and corruption
firm age+1 squared; log lagged sales; log strongly significant
lagged employment; log city population and
GDP per capita; city and industry dummies; As expected, ownership has strong effects on
firm performance. Relative to state ownership,
domestic private ownership is associated with a
higher sales growth rate and investment rate'.
Effect of foreign ownership even larger. There
`is no evidence that physical infrastructure
affects firm performance' but `the impact of
technological infrastructure appears to matter
significantly...Labor market flexibility matters
weakly...We do not find
Criticism: The two-step TFP regression is
vulnerable to Escribano and Guasch
simultaneity bias criticism. However, the
authors also use other firm performance
measures which produce at least approximately
similar results.
Aterido, Hallward- World Bank 1. A descriptive overview of - Firm Business climate variables: Results: `The results indicate significant
Driemeier and Pag�s Enterprise Surveys firm-level employment Employment differences across size categories of firms � both
(2007): Main Database growth and business climate Growth - Finance: in terms of differences in objective conditions
data from over 100 countries, � Firm has overdraft facility faced by firms and in terms of non-linearities in
`Business climate and focusing in particular on [change in the � % of sales sold on credit the impact of these conditions. Low access to
Employment Growth: differences by firm size. enterprise's � % of working capital financed finance, corruption, poorly developed business
The impact of Access permanent externally regulations and infrastructure bottlenecks shift
to Finance, Corruption 2. Regression of employment employment � % of investments financed externally downward the size distribution of employment.
and Regulations growth on IC constraints during the period t Low access to finance and ineffective business
Across Firms' controlling for a variety of and three years -Regulations: regulations reduce the growth of all firms,
firm characteristics (esp. size) before, divided by � Log of days to get an operating especially micro and small firms. Corruption
and other factors. the firm's simple license last 2 years and poor infrastructure create growth
average of � % of management's time dealing bottlenecks for medium and large firms. The
42
permanent with gov't regulation results also reinforce the importance of
workers during the � log of days spent on inspections differentiating the impact across size classes of
same period. The last year firms that allow for the micro firms (less than 10
measure is � log of average days to obtain employees) to be different from `small' firms'
symmetric and imports...last year
bounded by +/-2] � log of average days to get `our estimates suggest that a weak business
exports thru custom last year climate reduces overall employment in the
� log of total days spent on labor business sector...firms may be confined to
inspections last year industries with limited innovation and growth
opportunities. In addition, a larger share of firms
-Corruption may remain informal or semi-informal, reducing
� Firms in comparable activities the capacity of the state of collecting taxes and
bribe to get things done (yes/no) paying for fundamental inputs for development
� % of sales on bribes to get things such as education.'
done by similar firms
� Similar firms give gifts to
officials (yes/no)
� % of government contracts on
bribes by comparable firms
-Infrastructure
� Power outages during the last
year (log days)
� % of sales lost due to power
outages last year
� Log days of no water last year
Authors use country-city-sector-size
averages of these variables.
Controls: Firm size (micro/small
/medium/large); firm age (young/
mature/old); location (large and small
cities); ownership (foreign/government);
exporter; industry; country;
Hallward-Driemeir and World Bank 1. Descriptive overview of - Employment Business climate variables: Results: `Firms in Africa do face greater
Aterido (2007): Enterprise Surveys how the African employment Growth obstacles in terms of finance, infrastructure,
Main Database, growth rate compares to the -share of investments financed with bank public services and governance. ....[However,]
`Impact of Access to with a focus on the rest of the world, and how -Capital Intensity loans the more challenging business environment
Finance, Corruption African datasets specific business climate -days without power conditions [do not translate into lower average
and Infrastructure on contained therein constraints differ across -Change in Capital -management time with officials growth compared to other developing countries,
Employment Growth: regions and different types of Intensity -frequency of bribes to `get things done' but they] are associated with shifting down the
Putting Africa in firms. [firm] size distribution, lowering the relative
Context' Alternative specifications include: days to growth of larger firms, or in some cases
2. For both a sample of clear import customs; consistency of expanding micro-firms'
African countries and one of enforcement of regulations; share of sales
43
other developing countries, on credit; This may be because there are incentives to
the authors regress remain small e.g. because bad transport
employment growth and other Control variables (not all in every infrastructure creates demand-pockets for small
outcome variables on a set of regression): firm size, firm age, export suppliers, being small and informal minimizes
business climate indicators status, foreign ownership, sector controls; contact with corrupt state etc.
and controls. survey dummies, country controls;
Commander and BEEPS II and 1. Authors regress log of firm 1. Log of Firm Firm level IC variables from BEEPS Results:
Svejnar (2007): BEEPS III sales revenues on subjective sales revenues (subjective perceptions of managers on 1-4 - `Overall [the authors] show that country
business climate indicators scale): effects...matter for firm performance but that
`Do Institutions, First cross and a set of controls (OLS differences in the business environment
Ownership, Exporting sectional analyses, and IV) 2. Log of Change - cost of financing constraints observed across firms within
and Competition but authors also in firm sales - tax rates countries do not.'
Explain Firm construct panel 2. Using a constructed panel revenues - custom/foreign trade regulations
Performance? subset with of 1300 firms, the authors - business licencing&permit -Foreign ownership found to have positive effect
Evidence from 26 approx. 1300 firms regress the change of - macroeconomic instability on firm performance, but domestic private
Transition Countries' revenues between 2002 and - functioning of the judiciary ownership not
2005 on the 2002-05 rate of - corruption
change of labor and capital - street crime theft & disorder -Export orientation found to have positive effect
and on the initial 2002 levels - anti-competitive practices only in simple specification, not if authors
of the business environment - infrastructure control for ownership
constraints and structural
variables (OLS) Country level IC variables (in separate Criticism: The authors use only
regressions): subjective, perception based business climate
indicators. This may be sub-optimal. Objective
- 12 Doing Business variables business climate indicators, such as the time
- 10 Heritage Foundation indices of required to clear customs, have been found to be
economic freedom significant even with the inclusion of country
dummies (see for instance Dollar, Hallward-
Control variables (not all in every Driemeier, Mengistae, 2005). Subjective
regression): levels of capital and labor business climate indicators may be sub-optimal
inputs; categories of ownership (privatized; because systematic variations in perceptions in
new private; foreign), export orientation of the cross-country dataset are largely a function
firm; log of exports/sales. Note that main of broad business confidence related to macro
controls are replaced by Instrumental factors such as political and macroeconomic
Variables. stability or the financial system. Such country
level effects are largely captured by country
dummies. Once firm characteristics (such as
size, age etc.) are also controlled for, the
remaining variation in perception based
indicators of specific business climate
constraints may be largely due to quasi-random
factors such as the managers' personality. This
could explain why these subjective indicators do
not show up as significant, although concrete
objective measures of the business climate may
do.
44
Honorati and Indian Firm 1. Descriptive analysis of Annual sales Business climate variables: Results:
Mengistae (2007): Analysis and objective and subjective data. growth
Competitiveness - lagged profitability (finance proxy) 1. The authors find a pattern whereby `the better
`Corruption, the Survey, 2002 and 2. Regressions analysis to - lagged indebtedness (finance proxy) performing states are also better in every
Business Environment, 2005 examine the effects of - indicators of corruption, labor regulation important aspect of their business
and Small Business corruption, labor regulation, and power shortages environment...low-income, low-growth states
Growth in India' access to finance and the have the worst indicators of all institutional
quality of power supply on Controls: industry, state and year dummies; variables except for labor regulation'
the growth of manufacturing initial size;
businesses in India. 2. Regression results show that `average
business growth rate is lower where labor
regulation is greater, power shortages are more
severe, and financial constraints
stronger...[moreover] each of the three factors
on business growth depends on the incidence of
corruption....sales growth is constrained by
cash-flow only in businesses that are not
affected by labor regulation, power shortages or
corruption.' The authors interpret this `as
indication that corruption is a proxy for
something more fundamental than the payments
of bribes, namely, the quality of property rights
institutions in the sense of Acemoglu and
Johnson (2005)'. Their results are consistent
`with the...view that the quality of property
rights institutions exerts more abiding influence
on economic outcomes than the quality of
contracting institutions...'
wb18913
L:\Infrastructure & Invst Climate\Dethier Hirn Straub - Explaining Enterprise Performance - Nov 14 2008.doc
12/05/2008 3:43:00 PM
45