PRODUCTIVE RETAIL LAWS

PRODUCTIVE RETAIL LAWS

R

^EGULATORY

R

EFORMS AND THEIR

E

^{FFECT ON}

P

RODUCTIVITY

G

ROWTH IN

R

ETAIL

D

ISTRIBUTION

Tristan Kohl

December 2005

PRODUCTIVE RETAIL LAWS

R

^EGULATORY

R

EFORMS AND THEIR

E

^{FFECT ON}

P

RODUCTIVITY

G

ROWTH IN

R

ETAIL

D

ISTRIBUTION

Tristan Kohl 1257498

December 2005

Doctoraal Thesis

International Economics and Business

Faculty of Economics Supervisors:

University of Groningen prof. dr. Bart van Ark

The Netherlands dr. Lourens Broersma

ABSTRACT

This explorative study examines the relationship between regulatory reforms and productivity growth in OECD countries’ retail industry over the period of 1995-2002. First, price-cost averages are used as indicators of regulatory reform and a

∩

-shaped relationship is sought between regulation and productivity performance. The data do not provide much support for such a relationship. Second, potential linear and nonlinear relationships between regulation and productivity growth are examined. Recent data from the OECD International Regulation Database are used as indicators of specific retail laws. Negative linear relations are found for administrative burdens and restrictions on opening hours. Longer time series and in-depth data on the nature of regulations are required to allow for more thorough analyses.

ACKNOWLEDGEMENTS

While on exchange at McGill University in Montreal, Canada, I was faced with the question as to what I would write my thesis on upon my return to Groningen. The variety of courses I had taken and taught by then did not make the decision easy.

Fortunately, Jennifer Hunt’s labour economics course at McGill provided the spark I needed to start the fire – most likely without her even knowing it. The theoretical problems and statistics on worldwide labour productivity, wages, and (labour market) regulations she shared during the lectures showed me that there was much more to be said about the interaction between the economic and legal disciplines than I had initially thought possible. Since I am a student in both fields, the combination of labour productivity and regulation formed the theme – challenging as it is – that I was looking for.

I would like to gratefully acknowledge those who have so enthusiastically contributed to this paper. Bart van Ark stimulated my venture into the world of retail distribution. His creativity and expertise, combined with Lourens Broersma’s hands-on experience, helped me a long way in making this project a success. Paul Conway at the OECD and Gerard Ypma at the GGDC shared their data with me on several occasions and made working with datasets just that extra bit more enjoyable.

Marcel Timmer provided helpful comments on an earlier draft. Finally, I am thankful to Jesus Christ, my parents, Eugenie, and numerous friends and colleagues for providing vision, advice and encouragement. I sincerely appreciate all your help.

Tristan Kohl

TABLE OF CONTENTS

OVERVIEW OF ABBREVIATIONS...ii

OVERVIEW OF FIGURES ... iv

OVERVIEW OF TABLES... v

1. INTRODUCTION... 6

1.1. Purpose of the study ... 6

1.2. Why study retail distribution?... 7

2. LITERATURE REVIEW ... 10

2.1. Introduction to regulation... 10

2.2. Regulation, competition and productivity growth... 11

2.3. Regulation in retail distribution ... 12

2.4. Specific regulations in retail distribution ... 13

2.4.1. Legal/administrative barriers to entry... 14

2.4.2. Regulations on business operation ... 15

3. RESEARCH OBJECTIVES ... 16

4. PRICE-COST AVERAGES AND PRODUCTIVITY GROWTH ... 17

4.1. Introduction ... 17

4.2. Methodological framework... 18

4.2.1. Price-cost margins and price-cost averages ... 18

4.2.2. Models ... 18

4.3. Data sources ... 20

4.4. Results... 22

4.5. Discussion... 24

5. REGULATORY INDICATORS AND PRODUCTIVITY GROWTH ... 26

5.1. Introduction ... 26

5.2. Methodological framework and data sources ... 26

5.3. Descriptions of regulatory indicators ... 28

5.3.1. Administrative burdens (AB)... 28

5.3.2. Market entry requirements (MER) ... 29

5.3.3. Legal monopolies (MON)... 30

5.3.4. Land use policies (LUP) ... 31

5.3.5. Restrictions on opening hours (OH) ... 33

5.3.6. Price controls (PC) ... 34

5.3.7. Direct government control over business enterprises (GC) ... 35

5.4. Results... 36

5.5. Discussion... 38

6. CONCLUSION ... 39

BIBLIOGRAPHY... 42 APPENDIX A: DATA ...Error! Bookmark not defined.

APPENDIX B: INDICATORS OF PRODUCT MARKET REGULATIONError! Bookmark not defined.

OVERVIEW OF ABBREVIATIONS

AB Regulatory indicator: Administrative burdens

AUS Australia

AUT Austria

BE Regulatory indicator: Legal/administrative barriers to entry

BEL Belgium

BO Regulatory indicator: Regulations on business operation

CAN Canada

CZE Czech Republic

DEN Denmark

EC European Commission

ESP Spain

EU European Union

FIN Finland

FRA France

GC Regulatory indicator: Direct government control over business enterprises

GER Germany

GGDC Groningen Growth and Development Centre

GRE Greece

HUN Hungary

IRD International Regulation Database

IRL Ireland

ISL Iceland

ITA Italy

JPN Japan

KOR South Korea

LUP Regulatory indicator: Land use policies

LUX Luxembourg

MER Regulatory indicator: Market entry requirements

MEX Mexico

MON Regulatory indicator: Legal monopolies

NED Netherlands

NOR Norway

NZL New Zealand

OECD Organisation for Economic Co-operation and Development OH Regulatory indicator: Restrictions on opening hours PC Regulatory indicator: Price controls

PCA Price-Cost Average

POL Poland

POR Portugal

STAN Industry Structural Analysis

SUI Switzerland

SVK Slovak Republic

SWE Sweden

TUR Turkey

UK United Kingdom

US United States

OVERVIEW OF FIGURES

Figure 1: Administrative Burdens (AB) ... 29

Figure 2: Market Entry Requirements (MER) ... 30

Figure 3: Legal Monopolies (MON) ... 31

Figure 4: Land Use Policies (LUP) ... 32

Figure 5: Restrictions on Opening Hours (OH) ... 33

Figure 6: Price Controls (PC) ... 34

Figure 7: Direct Government Control over Business Enterprises (GC)... 35

OVERVIEW OF TABLES

Table 1: The Retail Industry's Contribution to Overall Economies, 1995-2002 ... 8

Table 2: Compliance with Assumptions of Regression Analysis... 23

Table 3: Results of Estimating Equation (1) and (2) with Pooled Country-Data ... 24

Table 4: Compliance with Assumptions of Regression Analysis... 37

Table 5: Results of Estimating Equation (3) with Pooled Country-Data... 37 Table A1: Value Added (in millions of national currency)... Error! Bookmark not defined.

Table A2: Number of Persons Engaged (in thousands) ... Error! Bookmark not defined.

Table A3: Labour Productivity per Person Engaged (in chained (1995) units of national currency) ... Error! Bookmark not defined.

Table A4: Labour Costs (in millions of national currency) ... Error! Bookmark not defined.

Table A5: Gross Output (in millions of national currency) ... Error! Bookmark not defined.

Table A6: Gross Fixed Capital Stock (in millions of national currency) ... Error! Bookmark not defined.

Table A7: Sales per Outlet (in thousands of euros)... Error! Bookmark not defined.

Table A8: Price-Cost Averages in Current Prices... Error! Bookmark not defined.

Table A9: Labour Productivity per Hour Worked (in chained (1995) units of national currency) .. Error!

Bookmark not defined.

Table B1: The Indicators of Regulation in Retail Distribution:

Category I – Legal/Administrative Barriers to Entry (BE) ... Error! Bookmark not defined.

Table B2: The Indicators of Regulation in Retail Distribution:

Category II – Regulations on Business Operation (BO) ... Error! Bookmark not defined.

Table B3: Conversion Matrix for Administrative Burdens (AB)... Error! Bookmark not defined.

Table B4: Conversion Matrix for Market Entry Requirements (MER)... Error! Bookmark not defined.

Table B5: Conversion Matrix for Legal Monopolies (MON) ... Error! Bookmark not defined.

Table B6: Conversion Matrix for Land Use Policies (LUP)... Error! Bookmark not defined.

Table B7: Conversion Matrix for Restrictions on Opening Hours (OH)... Error! Bookmark not defined.

Table B8: Conversion Matrix for Price Controls (PC)... Error! Bookmark not defined.

Table B9: Conversion Matrix for Direct Government Control over Business Enterprises (GC) ... Error!

Bookmark not defined.

Table B10: The Indicators of Regulation in Retail Distribution per Category, Type, Year, and Country ... Error! Bookmark not defined.

1. INTRODUCTION

1.1. Purpose of the study

Imagine a shopper’s paradise: large stores offering a broad range of high-quality goods and services at affordable prices. Around the clock business hours ensure full accessibility for early birds and night owls alike. New firms sprout up at a rapid pace to supply an even better range of goods and services.

And all this happens just around the corner from where everybody works and lives, of course.

Behind this buzz of activity lies a complex network of rules and regulations. A few changes to, say, laws stipulating how large stores can be, and the once so idyllic business environment becomes much less appealing to retailers wanting to enter or expand in the market. Just how would such a regulatory change affect the region’s economic performance?

The aim of this study is to determine how various laws in retail distribution affected the global industry’s productivity performance in the late 1990’s and early 2000’s. Seven types of retail laws are identified and related to labour productivity growth in most OECD (Organisation for Economic Co- operation and Development) countries. These laws cover the following domains: administrative burdens, market entry requirements, legal monopolies, land use policies, restrictions on opening hours, price controls, and direct government control over business enterprises.

Various approaches have been employed to better understand this field. One approach is to study the effects of one or more industry-specific laws on a single economy. For example, Skuterud (2000) analyses businesses’ responses to Canada’s deregulation of Sunday shopping hours (see also, e.g., Carree and Nijkamp, 2001; McKinsey, 2002; Burda and Weil, 2004). Although the detailed characteristics of country-specific regulations are highlighted and related to a country’s economy, these studies do not put their findings in an international comparative framework.

Another approach is to examine the aggregate, economy-wide effects of national reforms for a variety of countries. For instance, an OECD-wide study by Nicoletti and Scarpetta (2003) provides evidence that privatisation and entry liberalisation reforms enhanced productivity in manufacturing and service industries (also see, e.g., Conway, Janod and Nicoletti, 2005). This helps to better understand similarities and differences between countries’ legal and economic frameworks, but does not necessarily provide insight at the industry level on specific regulations. Some studies form an important exception to the rule, as they focus on the role of one or more retail laws at a global, industry level (see, e.g. Høj, Kato and Pilat, 1995; Pilat, 1997; Burda, 2000). However, these studies are mostly descriptive due to data limitations.

Broersma and Van Ark (2004) made progress in using quantitative data to relate cross-country deregulation to performance at an industry level for aviation, electricity, retail, road freight, and telecom. However, data limitations prevented studying industry-specific regulations at the time.

Fortunately, a new panel of data enables the relations between various retail laws and labour productivity to be studied at the industry level in the present study. This paper’s contribution to the literature is therefore new, in the sense that it examines the relations between specific retail laws and industry performance using quantitative data at an international comparative level.

1.2. Why study retail distribution?

Since the 1980s, countries belonging to the OECD have implemented structural reforms to free businesses from excessive legal burdens and to promote competition. By deregulating their economies, governments intended to enhance economic agents’ ability to adapt to major economic shocks, improving markets’ and businesses’ flexibility and efficiency (Høj, Kato and Pilat, 1995).

The retail trade industry has often been the focus of these regulatory reform programmes for at least two reasons. Firstly, retail distribution employs a considerable share of the national labour force and is a major contributor to overall production in most OECD countries (Pilat, 1997).

Table 1 displays the average percentage of the labour force working in retail distribution, the average share of value added it contributed to the economy, and the average annual productivity growth rate it experienced in most OECD countries in the period 1995-2002. Compared to other countries, Slovakia’s retail industry employed the smallest proportion of its national labour force (5.5%), whereas Australia’s retail industry employed the largest proportion (11.9%). The industry contributed between an average of 3.4% (Finland) and 7.5% (Greece) to total value added. The annual growth rates of retail labour productivity are most pronounced. Average annual retail labour productivity slightly shrunk in Belgium, but grew by more than 7% in Norway and the United States.

Notice that countries employing a larger proportion of their labour force than others do not necessarily enjoy a higher annual labour productivity growth rate. For example, 11.7% of both the Japanese and Korean labour force worked in retail distribution during 1995-2002, with Japan’s retail value added contributing to a larger portion of total value added (5.3%) than Korea’s (4.6%). Despite these similarities, however, Korea experienced higher productivity growth rates (5.7%) than Japan (0.4%).

Another example is that of the US and EU-15. The US has an only slightly higher employment share in retail distribution (9.6%) than the EU-15 (8.8%), but their growth rates are vastly different. The US experienced an average annual labour productivity growth rate of 7.2%, the EU-15 just 1.6%.

Table 1: The Retail Industry's Contribution to Overall Economies, 1995-2002.

Country Employment,

Average Share (%) Country Value Added,

Average Share (%) Country Average Annual Productivity Growth Rate¹ (%)

AUS 11.9 GRE 7.5 NOR 7.9

JPN 11.7 POL 7.3 US 7.2

KOR 11.7 ITA 6.3 KOR 5.7

UK 11.2 US 5.9 IRL 4.7

HUN 10.2 SVK 5.5 UK 4.3

IRL 10.1 ESP 5.4 AUT 3.7

US 9.6 POR 5.4 CZE 3.7

GRE 9.5 JPN 5.3 FIN 3.6

CAN 9.2 HUN 5.2 CAN 3.4

ESP 9.2 SUI² 5.2 SWE 3.4

SUI² 9.2 CZE 5.0 GRE 2.9

ITA 8.9 AUS 4.8 POL 2.8

EU-15 8.8 UK 4.8 AUS 2.4

GER 8.8 EU-15 4.6 LUX 1.8

NED 8.8 KOR 4.6 NED 1.7

CZE 8.5 FRA 4.5 EU-15 1.6

POR 7.8 AUT 4.3 FRA 1.6

POL 7.4 IRL 4.3 GER 1.6

AUT 7.2 DEN 4.1 SVK 1.3

BEL 7.1 NED 4.1 POR 1.2

DEN 7.1 GER 4.0 ITA 0.7

LUX 7.1 NOR 4.0 JPN 0.4

FRA 7.0 SWE 3.8 ESP 0.2

NOR 6.6 CAN 3.7 SUI² 0.1

FIN 6.2 BEL 3.6 HUN 0.0

SWE 5.9 LUX 3.5 DEN -0.1

SVK 5.5 FIN 3.4 BEL -0.2

Sources: GGDC (2005) and own calculations.

These data illustrate the second reason why the retail industry commands attention. The industry’s importance is characterised by the finding that wholesale and retail industries combined have been responsible for at least half of the United States’ economy-wide productivity lead over the European Union since 1995. Although productivity levels were similar across the Atlantic, the divergence in growth rates has resulted in a productivity gap (Van Ark, Inklaar and McGuckin, 2003;

Gordon, 2004).

The sheer size of the retail industry, along with its contribution to differences in countries’

economic performance, therefore warrants further investigation into the nature of the relationship between industry-specific regulatory reforms and economic growth.

1 Labour productivity is defined as value added per hour worked in order to take part-time employment into account.

2 The data cover 1995-2000 for Switzerland.

In analysing the US-EU productivity gap, McGuckin, Spiegelman and Van Ark (2005) determined that a country’s retail trade performance (or lack thereof) can be attributed to (1) firms’ capacity to exploit new technologies, (2) managers’ ability to implement organisational and process changes, and (3) national regulations. Countries’ legal frameworks and business environments may vary for any number of reasons, including differences in their initial conditions and the extent and pace of deregulation. OECD members’ regulations on retail distribution tend to be quite different (Høj, Kato and Pilat, 1995; Boylaud and Nicoletti, 2001), which hinders innovation and economic growth in service industries (Nicoletti, 2001).

Aside from differences in industry-specific regulations, productivity growth differentials may also arise because of scale advantages, i.e. regulatory reforms could facilitate certain businesses in exploiting economies of scale. For example, suppose there is a country-market with relatively few regulations, such as the United States, and a country which regulates its retail industry more intensively, such as France. In a market with relatively few restrictions, American retailers would be better capable of (re)organising and exploiting their resources than would be the case in France, resulting in higher efficiency and enabling higher productivity growth in the US.

By adopting an internationally comparative approach in investigating the relation between regulatory reform and productivity growth, economists are aided in studying the economic implications of various policy choices affecting the global retail industry. Additionally, policy-makers are provided with a more detailed understanding of their country’s legal and economic position vis-à- vis the rest of the world and how it could be influenced by implementing regulatory reforms.

Such an international approach is particularly insightful in light of regulations that could prevent global retailers from exploiting cross-border economies of scale. Up until 1996, for example, the South- Korean government implemented numerous policies to redirect capital to only a handful of

“strategic” manufacturing industries. Retail distribution was not considered to be a strategic industry and restrictions were even put on domestic chaebols wanting to expand into this industry. Low- productivity shops were also protected from competition with large-scale retailers by banning shopping centres. This made it next to impossible for global firms with modern retailing formats to establish sufficient outlets and achieve critical scale (Lewis, 2004:123-124). Exploiting cross-border economies of scale can therefore be prevented by stringent (cross-border) regulations.

A better understanding of the productivity effects of various regulations does not imply that policy-makers are expected to be able to adjust one policy without having to modify the broader regulatory framework. However, there may be an opportunity to identify key legislations best suited at enhancing countries’ competitiveness and productivity as more detailed data on retail laws become available.

This paper proceeds as follows. Section 2 discusses the literature. Section 3 provides an overview of the research objectives. Section 4 re-examines some of the findings reported in Broersma and Van Ark (2004) using a replicated and larger dataset. Section 5 studies the relation between various retail laws and labour productivity growth. Section 6 concludes.

2. LITERATURE REVIEW

2.1. Introduction to regulation

The present study focuses on product market regulation in retail distribution. The product market is the market where goods and services are sold. For retail distribution, laws in this field mainly encompass administrative and economic regulations. Properly designed and implemented product market regulations are aimed at enhancing market economies’ allocative and productive efficiency.

This is achieved by altering economic agents’ entry and exit decisions, their choice of inputs, the types and quantities of output produced, and prices (OECD, 1996). Labour market regulations are not considered for the purpose of this paper, albeit that they do form an important field for further study (see, e.g. Nicoletti, Bassanini, Ernst, Jean, Santiago and Swaim, 2001; Gust and Marquez, 2004).

Product market regulation is a means of government intervention to promote or discourage certain forms of economic behaviour. A variety of laws can create certainty about contracts and induce efficient, competitive markets. On the other hand, they may deter monopolist behaviour (e.g. EC Merger Law), restrict commercial opening times, and protect small mom-and-pop stores from being dominated by large-scale retail outlets.

These objectives may be complementary. For instance, restrictions on monopoly and cartel behaviour ensure that a certain level of competitiveness is maintained in most markets. However, laws’ objectives may also have counterproductive economic effects. For example, zoning laws may initially be useful in planning, ordering and administrating the use of land. But, if these laws (excessively) protect certain interests – those of owners of small stores, or the preservation of old buildings, for instance – they can reduce competition laws’ ability to induce efficient, competitive market conditions. Exactly why this would be the case is described in greater detail in section 2.2. The main point here is that regulations may be beneficial up to a certain extent, before they start having undesirable side-effects.

2.2. Regulation, competition and productivity growth

Capital and labour market regulations have long been thought to be the main legal barriers to competition. However, industry-level (product market) laws can be argued to be equally, if not more, important (Lewis, 2004). Instead of focussing only on the overall, economy-wide picture, industry- level analyses should be performed to identify fundamental trends, patterns and issues.

Metaphorically, the overall (economy-wide) picture will only be better understood by studying the individual (industry-level) pieces of the puzzle.

For example, Lewis (2004:xxiv, 9-13) argues that Thatcher’s government was believed to have implemented capital and labour market reforms in order for the UK economy to be similar to that of the US, at least from a regulatory point of view. However, product market competition remained much more distorted in the UK than was the case in the US. An important reason was the difference in industry-level restrictions. The UK may have reformed its capital and labour markets, but strict land use policies in the UK’s retail distribution continued to prevent businesses from creating larger outlets and reaping the benefits of economies of scale and scope.

Neoclassical economic theory argues that profit-maximising firms facing unrestricted, competitive conditions will choose an efficient, optimal production point on their production possibility frontier.

The case of the US illustrates how flexible land use policies (also known as zoning laws) aided Wal- Mart in implementing its “big-box” retailing concept in the 1990s. Wal-Mart’s stores carry a broad assortment of goods. These “one-stop shops” are large and diverse, which enables them to exploit economies of scale and scope and underprice competitors. As price competition increased, its rivals were forced to either catch up with Wal-Mart’s innovation (such as Target, Sears and K-mart) or exit the market (such as mom-and-pop stores, which do not have the resources to compete effectively).

Overall, these developments have increased competition and efficiency in the US retail industry (Lewis, 2004:31, 93-94).

However, the “big-box” retailing concept and its competitive consequences could not have been implemented, had zoning laws been in favour of protecting small retailers by banning large-scale outlets. Thus, firms’ incentives and opportunities change if they are subject to the burdens of excessive legal restraints and will consequently choose an inefficient level of output (Winston, 1998). As a result, firms will not economise on resources or exploit economies of scale and scope, which limits competitive behaviour. Thus, inappropriate regulations impose higher costs and inefficiencies on firms, the sectors in which they operate, and economies at large. Fortunately, prudent regulatory reforms may facilitate such firms in reaching less inefficient or efficient production levels, thereby allowing them to become more competitive (Blöndal and Pilat, 1997).

Research on product market regulatory reforms at a cross-country level confirms that burdensome administrative environments go hand in hand with economic regulations that restrict competition (Nicoletti, Scarpetta and Boylaud, 2000). High administrative burdens and strict market entry requirements, for example, dampen entrepreneurial activity and reduce incumbents’ risk of having to face competition from new entrants. Although reforms to lower regulatory impediments to competition between 1998 and 2003 had some success, most OECD countries continued to have some set of regulations that persistently hinders competition (Conway et al., 2005). Thus, regulatory reforms have not yet succeeded in creating a level, competitive global playing field.

Empirical work further supports the notion that increased product-market competition is associated with higher innovation and productivity growth, as firms seek ways to develop and maintain their competitive advantage by increased technological diffusion and innovation (Nickell, 1996; Blundell, Griffith and Van Reenen, 1999; Bassanini and Ernst, 2002). This is illustrated by the case of Wal-Mart and the US retailing industry in the 1990s.

Nevertheless, expected gains from innovation and firms’ resources to invest in productivity growth may decrease if markets become too competitive. For example, retailers engaged in severe price competition must undercut their rivals in order to survive. In a situation where consumer prices are lower than the underpriced goods’ costs, only the firm that has the larger resource base will be able to outlive its competitors and retain the (largest) market share. However, this is a severely costly strategy and can deplete firms’ resources. The result is that firms do not have sufficient means left to adequately invest in innovation and productivity growth. This is known as the Schumpeterian effect, which renders the aforementioned positive relation between competition, innovation and productivity negative. Combining these theoretical and empirical findings gives rise to a

∩

-shaped relationship between product-market competition and innovation (Aghion, Bloom, Blundell, Griffith and Howitt, 2002; Rogers, 2002).

2.3. Regulation in retail distribution

Several studies have been geared at investigating industry-specific regulations and their effects on a limited set of national retail industries (see e.g. Burda, 2000; Skuterud, 2000; Carree and Nijkamp, 2001;

McKinsey, 2002). All these studies generally conclude that fewer and/or more flexible regulations result in higher output, lower prices, more services to consumers, higher consumer welfare, increased employment, and higher productivity. However, cross-country evidence on the nature of the relationship between competition-enhancing regulatory reforms and performance in retail distribution appears to be somewhat limited and mixed.

One study (OECD, 1996) concludes that worldwide deregulation has resulted in increased innovation in many sectors, including utilities, telecommunications, financial services and retail distribution. Inklaar, O’Mahony and Timmer (2003) report that the retail industry intensively uses ICT technology in its production process and that it is one of the main contributors to the US’ retail productivity lead over the EU. Nicoletti and Scarpetta (2003) confirm that regulations affect technology adoption more strongly in competitive markets, such as retail trade and financial intermediation, than in natural monopoly markets, such as transportation and telecommunications.

Their study also shows that privatisation and entry liberalisation in services enhances competition, which positively affects productivity.

An illustration of the above findings is found, again, in Wal-Mart. Lewis (2004:93-94) describes that an important contribution to the company’s success was made by the computerisation of its inventory management system. The immediate benefit was that the system performed much more efficiently, saving resources and aiding the firm to be better at maintaining a steady inventory than its rivals. Market conditions allowed rivals to adopt this innovative approach, though not all managed to do so (as) successfully.

Nevertheless, recall that regulations could have negative economical consequences, i.e. regulation is economically beneficial up to a certain point, but not beyond. A recent cross-country, industry-level study by Broersma and Van Ark (2004) suggests that such a

∩

-shaped relationship exists in retail trade and utility services. Such a relationship implies that there is an optimal level of (industry-level) regulation at which productivity growth is maximised, but that productivity growth will be compromised if regulations are either too strict or too flexible vis-à-vis that optimal level. However, their measure of regulation, price-cost averages, does not provide insight as to which regulations in particular are responsible for this finding.

2.4. Specific regulations in retail distribution

Boylaud and Nicoletti (2001) categorise the major regulations that affect retail distribution as:

(1) legal/administrative barriers to entry and (2) regulations on business operation. Legal/

administrative barriers to entry include four specific retail laws. These are: administrative burdens, market entry requirements, legal monopolies, and land use policies. Regulations on business operation include restrictions on store opening hours, price controls, and direct government control over business enterprises. How these seven retail laws may affect productivity growth is discussed below.

2.4.1. Legal/administrative barriers to entry

The retail industry is characterised as a competitive industry with a large number of firms and high entry and exit rates. However, the structure varies widely across countries due to, amongst others, barriers to entry (Pilat, 1997). Entry into the retail industry may be regulated and restricted by means of administrative burdens, market entry requirements, the presence of legal monopolies, and land use policies.

High administrative burdens and market entry requirements restrict competition and hamper innovation and growth by increasing (innovative) entrants’ costs of formalities. Endorsing local or national monopolies to sell certain goods, such as alcohol, pharmaceuticals and tobacco, also creates an artificial barrier to entry. Entrepreneurs are restrained in setting up sole-proprietor firms and corporations and are deterred in creating new enterprises and forms of competition, such as e- commerce, in an otherwise dynamic, rapidly changing industry. This generally hinders the industry’s modernisation (Boylaud, 2000) and restricts competition, innovation and growth (Bassanini, Scarpetta and Vasco, 2000). Hence, heavier administrative burdens, stricter market entry requirements, and legal monopolies are expected to result in lower productivity growth (hypotheses 2a, 2b, and 2c).

Decisions regarding retail outlets’ size and location are highly dependent on land use policies.

Examples of how different the effects can be have already been provided on the UK and US earlier in this section. Recapitulating, rules are relatively flexible in the United States and thereby open the market to large hypermarkets, which can exploit economies of scale and scope, resulting in lower prices, increased competition, more innovation, and higher productivity. These firms tend to innovate more than small independent firms (Reardon, Hasty and Coe, 1996; Pilat, 1997), and modern retailing formats are almost twice as productive as their traditional counterparts (Lewis, 2004:31). This leads to higher productivity growth for countries that have flexible policies favourable to large retailers.

However, firms will be restricted in exploiting economies of scale and scope in markets where land use policies set limitations on retail outlets’ floor space and location. Strict land use policies may also cause the property values of incumbent, inefficient firms to be artificially inflated, increasing barriers to entry for innovative entrants, yet not prompting incumbent firms to exit. Such restrictions also limit incumbents’ ability to expand their current operation surface in order to exploit economies of scale and derive maximum efficiency from optimal labour and delivery scheduling. Therefore, stricter land use laws are expected to be associated with lower productivity growth (hypothesis 2d).

2.4.2. Regulations on business operation

Once a firm has entered the retail industry, it is subject to at least three distinct policies that regulate its operations. These are restrictions on shop opening hours, price controls, and direct government control over business enterprises.

The first regulation on business operation concerns the restriction of shop opening hours. These laws limit productivity for a number of reasons. Consumers may not able to obtain the products they are looking for due to limited hours available for shopping, resulting in a loss of value-added (McGuckin et al., 2005). Additionally, consumers may be forced to make several (inconvenient) short stops at local stores throughout the week instead of having sufficient time to allow for one, long visit to a one-stop retailer, especially if all senior members of the household have full-time jobs. Such time constraints make it more difficult for consumers to find sufficient time to purchase more than only the

“bare essentials”, hampering retail sales, productivity and consumer welfare (Pilat, 1997). Restrictions on opening hours are therefore expected to result in lower productivity growth (hypothesis 2e).

Price controls form the second type of regulation on retailers’ operations. Documentation on the industry-wide productivity effects of price controls appears to be scarce, though it suggests that price controls restrict competition in competitive industries, such as road freight and retail distribution (Nicoletti, Scarpetta and Boylaud, 2000). Intuitively, higher prices can be charged if price controls are abolished. Yet competitors can capture the market by undercutting their rivals, causing margins to decline. This effectively enhances competition and further enforces the need for modernisation to stay in the market. This results in more innovative (and productive) means of doing business (Nicoletti, 2001). Hence, fewer price controls may lead to more productive industries (hypothesis 2f).

The final retail law investigated in this paper is direct government control over business

enterprises. Governments can create a framework that is conducive for economic growth by (1) facilitating a legal and physical infrastructure that allows the market economy to function, and (2) providing a limited set of public goods. More regulation, in this case, is better than none. Yet should a government become directly involved with controlling a business enterprise and thereby move beyond these core tasks, it is likely to hamper economic growth. This is because government officials and politicians are not primarily qualified to make strategic business decisions, where the highly competitive, dynamic marketplace continually requires innovative approaches to doing business (Gwartney, Lawson and Holcombe, 1998). Hence, only minimalist government control is expected to clear the way for productivity growth (hypothesis 2g).

3. RESEARCH OBJECTIVES

The objective of this study is to determine how various regulations affected OECD economies’

productivity growth rates in retail distribution during the late 1990s and early 2000s. Did all types of regulation have the same impact on productivity growth, or did they vary in importance? Do differences in countries’ regulatory framework reforms explain the resulting differences in performance sufficiently, or should other factors, such as economies of scale, be taken into account?

Various linear and nonlinear models are tested in an attempt to provide answers.

Section 4 confronts the model tested in Broersma and Van Ark (2004) with a new panel of data.

Using price-cost averages (PCAs) as countries’ indicators of regulatory regimes and a somewhat larger dataset, a standard Cobb-Douglas production function is tested to determine whether a

∩

^-

shaped relationship exists between regulation and productivity growth at an overall, industry level.

The following hypothesis applies:

Hypothesis 1: The relation between regulation and productivity growth is

∩

^-shaped.

Section 5 studies the relationship between productivity growth and seven types of industry- specific regulations, using the latest data from the OECD’s International Regulation Database (IRD).

The literature review in section 2 suggests that these relations are linear. However, the analyses will also allow for non-linear relationships because of the exploratory nature of this paper. This will also be helpful in comparing the results of sections 4 and 5 if non-linear relationships are found.

Hypothesis 2a: The relation between administrative burdens (AB) and productivity growth is negative.

Hypothesis 2b: The relation between market entry requirements (MER) and productivity growth is negative.

Hypothesis 2c: The relation between legal monopolies (MON) and productivity growth is negative.

Hypothesis 2d: The relation between land use policies (LUP) and productivity growth is negative.

Hypothesis 2e: The relation between restrictions on opening hours (OH) and productivity growth is negative.

Hypothesis 2f: The relation between price controls (PC) and productivity growth is negative.

Hypothesis 2g: The relation between direct government control over business enterprises (GC) and productivity growth is negative.

Recall that administrative burdens, market entry requirements, legal monopolies and land use policies are all examples of legal/administrative barriers to entry. Indicators for this separate type of regulation are constructed by means of factor analysis. The same is done for restrictions on opening hours, price controls, and direct government control over business enterprises, which are examples of regulations on business operation. Two more hypotheses are consequently tested to examine the relation between regulatory barriers to entry/restrictions on business operations and productivity growth, respectively.

Hypothesis 3a: The relation between legal/administrative barriers to entry (BE) and productivity growth is negative.

Hypothesis 3b: The relation between regulations on business operation (BO) and productivity growth is negative.

Section 4 and 5 each provide a discussion of their respective findings; section 6 concludes.

4. PRICE-COST AVERAGES AND PRODUCTIVITY GROWTH

4.1. Introduction

In their study on the impact of regulation on productivity growth at an industry level, Broersma and Van Ark (2004) [BvA] identify a

∩

-shaped relationship between regulation and performance in retail distribution and utility services. Such a relationship implies that each country has a specific degree of regulation at which performance is optimal.

This means that (1) too much government intervention places unnecessary burdens on economic agents and limits performance, and (2) retailers require a certain level of government involvement and order to induce proper and efficient business operation. In theory, a

∩

-shaped relationship would require governments to deregulate, or even re-regulate, up until the optimal point of regulation is obtained.

This section re-examines the BvA-model using a somewhat larger data panel. The methodological framework is introduced, followed by a description of the data sources and a presentation and discussion of the results.

4.2. Methodological framework

4.2.1. Price-cost margins and price-cost averages

At the time of the BvA study, OECD IRD indicators on regulation in retail distribution were only available for 1998, preventing time-series analysis. However, price-cost margins were argued to be an acceptable indicator of market regulation, as most regulatory reforms are aimed at opening up markets for more competition. The lower a firm’s market power (i.e. the more competitive the market), the smaller the difference between a product’s market price,

p

, and its marginal costs,

mc

, thereby decreasing firms’ price-cost margins,

( ^{p − mc p} )

. Additionally, price-cost margins could be constructed to reflect regulation on an annual basis, thereby enabling time-series analysis of regulatory reform.

Despite the price-cost margin’s theoretical appeal, marginal costs cannot be observed.

Alternatively, price-cost averages were selected as an empirically viable proxy of market concentration. Following Aghion et al. (2002), the price-cost average is the ratio of operating surplus to production,

os pq

, where operating surplus is defined as value added minus labour costs,

va − lc

^.

Value added is the difference between gross output and intermediate costs, such as purchasing costs and operating costs that are not attributable to labour or capital. Subtracting labour costs from value added implies that operating surplus measures the return to capital, or profit. As a country’s retail industry becomes more competitive, the price-cost margin declines and profitability decreases, which lowers the value of the price-cost average.

4.2.2. Models

Using a standard Cobb-Douglas production function with constant returns to scale, the data are pooled for all countries to maximise the number of observations and analysed with equation (1),

( ) ^α ( ) ^ε

α

α α

α α

+ Δ

× +

Δ

× +

+

⎟ +

⎟

⎠

⎞

⎜ ⎜

⎝ Δ ⎛ +

⎟ =

⎟

⎠

⎞

⎜ ⎜

⎝ Δ ⎛

−

−

−

−

t c t

c t

c t

c

t c t

c t

c t c t

c t c

L PCA

K PCA

PCA L PCA

K L

Y

, 2

, 5 , 2

, 4

2 2 , 3 2 , 2 ,

, 1

0 ,

,

log log

log log

(1)

where index

c

refers to country and index

t

to time,

Y

represents value added,

L

total employment,

K

real capital stock,

PCA

the price-cost average, and

ε

is the error term. The price-cost average is lagged with two periods (years) to avoid simultaneity bias, as value added is incorporated in both the left- and right-hand variables.

Equation (1) allows for a direct effect of deregulation on productivity growth, but also for an indirect effect on capital growth (first cross-term) and on employment growth (second cross-term).

Regulations may prevent businesses from adopting certain innovative technologies, thereby slowing capital growth (Nicoletti and Scarpetta, 2003), or reduce firms’ need to employ more skilled, productive labourers, thereby limiting labour growth. This means that if either capital or labour growth are positive, negative values for

α

₄ ^and

α

5 imply that deregulation enhances performance.

Equation (1) is the original BvA specification. However, it does not consider the role of economies of scale in determining the relationship between regulation and productivity growth. An additional equation, (2), is introduced to control for outlets’ scale,

( )

( ) ^{α α ε}

α

α α

α α

α

+

× +

+ Δ

× +

Δ

× +

+

⎟ +

⎟

⎠

⎞

⎜ ⎜

⎝ Δ ⎛ +

⎟ =

⎟

⎠

⎞

⎜ ⎜

⎝ Δ ⎛

−

−

−

−

−

c t

c c

t c t

c

t c t

c t

c t

c t

c t c t

c t c

SCALE PCA

SCALE L

PCA

K PCA

PCA L PCA

K L

Y

2 , 7 6

, 2

, 5

, 2

, 4 2

2 , 3 2 , 2 ,

, 1

0 ,

,

log

log log

log

(2)

where

SCALE

controls for each country’s average outlet size. All other symbols are equal to what they represent in equation (1). Positive values are expected for

α

₆, as firms capable of exploiting economies of scale (e.g. supermarkets) can be more productive than small-scale retailers (e.g. mom- and-pop stores).

The new cross-term allows for an interaction between deregulation (i.e. a lower price-cost average) and scale effects (measured as sales per outlet). Limited/no deregulation (i.e. a higher price-cost average) does not provide retailers new opportunities to exploit economies of scale, which hinders them from experiencing high sales per outlet and productivity growth. Hence, negative values for

α

₇

indicate that deregulation enhances productivity growth. Overall, there will be a

∩

^-shaped

relationship between regulation and performance if

α

₃

< 0

in either equation.

Both equations are estimated using country-data covering 1995-2001. The observations are pooled, yielding sufficient observations to perform regression analysis. Dataset 1 represents the ten countries used in the BvA study. Dataset 2 adds two more countries to the analysis. Unfortunately, none of the models allow for country-fixed effects because of a limited number of degrees of freedom.

4.3. Data sources

The original set of countries studied in Broersma and Van Ark (2004), or dataset 1, includes Austria, Belgium, Finland, France, Germany, Italy, the Netherlands, Sweden, the United Kingdom and the United States, covering the period 1995-2001 and thereby yielding a maximum of 70 observations.

Further data mining has added Denmark and Greece to the additional set of countries that is also analysed in this paper (dataset 2). The timeframe of analysis remains unchanged due to data limitations, resulting in a maximum of 84 observations. Datasets 1 and 2 are based on numerous sources. Their construction is addressed in this subsection.

Data on value added in current prices (appendix table A1) and total persons engaged (table A2) were obtained from the GGDC 60-Industry Database (GGDC, 2005) and O’Mahony and Van Ark (2003) CD-ROM. The GGDC database is based on national accounts (as compiled in the OECD STAN Database) and industrial and business surveys. Labour productivity per person engaged was calculated by taking the ratio of value added and total persons engaged (see appendix table A3 for the data). The GGDC makes a distinction between total persons employed and total persons engaged. The latter includes independent workers and is used in this study because they are also affected by regulations and contribute to the industry’s value added. Moreover, data on total persons employed were rather limited. Note, however, that the methodologies used in GGDC (2005) and O’Mahony and Van Ark (2003) are not entirely comparable. This may result in higher values for the number of persons employed than for the number of persons engaged, which is a limitation to the data’s quality.

Total labour costs (table A4) were also derived from the abovementioned sources. Labour costs were calculated, when necessary, by multiplying compensation per employee by the number of persons engaged. Note that this is not the same as multiplying by the number of persons employed.

However, this was the only feasible solution, as data on total persons employed and compensation per employee were not available for all countries involved.

Data on production in current prices (table A5) were derived from the OECD STAN Database (OECD, 2005a). Unfortunately, these data were only available at an aggregate level (wholesale and retail combined) for Belgium, Sweden, and the US. Corrections were made in the following manner.

First, disaggregated annual data on value added were obtained from the GGDC 60-Industry Database (GGDC, 2005). The annual contribution of each country’s retail industry to the combined value added of the wholesale and retail industries was then calculated. These shares were multiplied by the total production data to obtain an estimate of retail production in current prices.

Data on real capital stocks (table A6) are not readily available and therefore need to be calculated using a long range of time-series investment data and equation (3),

t c t c t c t

c

K I

K

_,

= ( 1 − δ

_,−1

)

_,−1

+

_, ⁽³⁾

where index

c

refers to country and index

t

to time,

K

represents real capital stock,

δ

the annual depreciation rate and

I

gross fixed capital formation (investment).

A series of investment data covering 1979-2001 was needed to obtain appropriate estimations of real capital stocks for the period 1995-2001. Gross fixed capital formation data were derived from the OECD STAN Database (OECD, 2005a) and Broersma and Van Ark (2004), and deflators from the GGDC Total Economy Growth Accounting Database (Timmer, Ypma and Van Ark, 2003). This information was supplemented, where necessary, with data from national statistics offices.

The calculations proceeded as follows. First, the deflators were used to convert gross fixed capital formation to constant prices. Equation (3) was consequently used to calculate gross fixed capital stock.

Note that the deflators and depreciation data are unweighted annual averages of France, Germany, the Netherlands, the UK and the US, which constitute (almost) half the countries listed in datasets 1 and 2. Such a measure is not perfect, but acceptable in light of data availability.

The scale variable (table A7) controls for the size of each country’s average outlet. This outlet-level scale variable is based on the average sales per outlet. Data on total retail sales in 1997 were obtained from various national statistical records. Time-series data were subsequently constructed as follows.

First, note that retailing distribution’s gross output, or production, is the difference between its sales and purchasing costs (see Timmer, Inklaar and Van Ark, 2005, for a discussion). Assuming that these value margins (i.e. the difference between the price received for a good and price that must be paid to replace it) remain constant, the trend in production data (as described above) can then be used to construct total sales for the timeframe of 1995-2001.

The price-cost averages (PCAs) were calculated by subtracting labour costs (table A4) from value added (table A1). The difference is operating surplus, which is consequently divided by gross output (table A5) to yield price-cost averages. Their values are displayed in table A8.

4.4. Results

Four assumptions must be met prior to performing regression analysis. First, each variable must be normally distributed (the Kolmogorov-Smirnov and Shapiro-Wilk statistics must yield p-values greater than 0.05). Second, the residuals must be normally distributed with a mean of 0 (the Kolmogorov-Smirnov and Shapiro-Wilk statistics must yield p-values greater than 0.05). Third, the variance of residuals must be constant (approximated by graphic analysis). Fourth, the residuals may not be correlated (the Durbin-Watson statistic should range between 1.5 and 2.5).

Details on the extent to which the assumptions of regression analysis are met, are displayed in table 2. Almost none of the variables are normally distributed. This is to be expected with the rather limited number of observations used in this study. The other assumptions for regression analysis are met: the residuals are normally distributed with a mean of 0, show constant variance and are not correlated. In sum, not all assumptions are met to perform regression analysis. The regression results should therefore be interpreted with some caution.

The results of estimating equations (1) and (2) are presented in table 3. The second column displays the results reported in the original study by Broersma and Van Ark (2004). Dataset 1 repeats the same regression for the same set of countries and timeframe. However, none of the coefficients are significant at the 5% level and equation (1) only explains 1% of the variance in the data when taking the number of variables and observations into account. The model is not significant with a p-value greater than 0.05.

Two countries – Denmark and Greece – were added to those in dataset 1 to construct dataset 2.

Equation (1) then explains 5% of the variance in the data and the model is not significant. However, it does provide weak support for the notion that deregulation stimulates labour growth and therefore labour productivity growth (

α

₅

< 0

and is significant at the 10% level).

Equation (2) controls for scale effects. The role of scale effects is positive and significant for both datasets (

α

₆

> 0

). Both also affirm the notion that deregulation enhances productivity indirectly through labour and scale effects, albeit that the results are not statistically significant. The coefficients reported for

α

₃ are negative when testing with equation (1), but positive when controlling for scale effects. None of these values are statistically significant.

Table 2: Compliance with Assumptions of Regression Analysis³.

Dataset 1 Dataset 2

Equation (1) Equation (2) Equation (1) Equation (2) Δlog(Y/L) 0.15

(0.00)

0.15 (0.00)

0.16 (0.00)

0.14 (0.00) Δlog(K/L) 0.08

(0.20^†)

0.08 (0.20^†)

0.21 (0.00)

0.16 (0.00)

PCA 0.12

(0.02)

0.12 (0.02)

0.08 (0.20^†)

0.10 (0.07)

PCA² 0.21

(0.00)

0.21 (0.00)

0.17 (0.00)

0.19 (0.00)

Δlog(K) 0.09

(0.20^†)

0.09 (0.20^†)

0.23 (0.00)

0.19 (0.00) PCA × Δlog(K) 0.17

(0.00)

0.17 (0.00)

0.23 (0.00)

0.21 (0.00)

Δlog(L) 0.14

(0.00)

0.14 (0.00)

0.11 (0.02)

0.12 (0.01) PCA × Δlog(L) 0.19

(0.00)

0.19 (0.00)

0.16 (0.00)

0.18 (0.00)

SCALE - 0.14

(0.00) - 0.16

(0.00)

PCA × SCALE - 0.10

(0.08) - 0.09

(0.20^†) Standardised Residual 0.10

(0.10)

0.11 (0.07)

0.11 (0.03)

0.08 (0.20^†) Constant Variance Sufficient Sufficient Sufficient Sufficient

Durbin-Watson

statistic 2.07 2.09 1.86 1.95

Notes: p-values are between parentheses. ^† indicates the lower bound of the true significance.

Values are bold when they meet the relevant assumption. The Kolmogorov-Smirnov statistic applies Lilliefors significance correction.

3 The table displays Kolmogorov-Smirnov statistics for the dependent variable, the independent variables and the standardised residual. Although both the Kolmogorov-Smirnov and Shapiro-Wilk tests may be used to examine whether variables are normally distributed, only the former is used. This is because it is more reliable

Table 3: Results of Estimating Equation (1) and (2) with Pooled Country-Data.

Dataset 1:

10 Countries

Dataset 2:

12 Countries BvA (2004)

(1) (2) (1) (2)

Constant

(

α

₀^{) (-2.92) -0.11* (1.19) 0.01 (-0.92) -0.02 (1.19) 0.01 (-1.02) -0.01}

K/L

(

α

₁^{) (2.72) 0.56* (-0.65) -0.20 (-0.55) -0.20 (-0.08) -0.02 (-0.91) -0.27}

PCA

(

α

₂^{) (3.49) 1.21* (0.88) 0.05 (0.26) 0.04 (0.80) 0.05 (0.20) 0.02}

PCA²

(

α

₃^{) (-3.20) -2.17* (-1.59) -0.21 (0.34) 0.06 (-1.12) -0.14 (0.28) 0.05}

PCA × K

(

α

₄^{) (-3.59) -3.24* (0.94) 1.33 (0.08) 0.16 (0.00) 0.00 (0.67) 0.72}

PCA × L

(

α

₅⁾ (-0.32) ^-0.28

-1.92 (-1.22)

-1.86 (-1.07)

-2.51^‡ (-1.67)

-2.86^‡ (-1.91) SCALE

(

α

₆^{) - - 0.04}

†

(2.04) - 0.04*

(2.89) PCA × SCALE

(

α

₇^{) - -} (-0.94) ^-0.12 - -0.10 (-1.41)

R² 0.28 0.09 0.22 0.12 0.23

Adjusted R² 0.22 0.01 0.12 0.05 0.14 F-statistic

(p-value)

4.62 (0.00)

1.10 (0.37)

2.27 (0.04)

1.83 (0.12)

2.66 (0.02)

Observations 65 65 65 76 71

Notes: * means the coefficient is significant at 1%, ^† at 5%, and ^‡ at 10%.

t-values are displayed between parentheses. Significant values are bold.

4.5. Discussion

Table 2 shows that most of the variables are not normally distributed. Caution is therefore advised when interpreting the regression results.

Re-examining the original set of countries with equation (1) does not yield results that are entirely comparable to those in the BvA study. Although equation (1) reveals negative values for

α

₃ ^{in both}

datasets, none of the values are statistically significant. Controlling for scale effects renders

α

₃

statistically insignificant and incorrectly signed. Hence, neither dataset provides empirical support for a

∩

-shaped relationship between regulation and performance, regardless of whether economies of scale are accounted for.

The reason why the results differ from those in BvA is mainly due to a difference in the construction of the dependent variable. BvA base their price-cost averages on the labour costs of only employed persons, and not the self-employed. However, persons engaged is the preferred measure because it includes all who work in the industry and are affected by regulatory reforms. Moreover, labour costs that exclude self-employed workers’ labour compensation underestimate the true labour costs and thereby overestimate operating surplus and the price-cost average.

The scale variable is statistically significant for both datasets and correctly signed. It can thus be concluded that higher average sales per outlet (representing economies of scale) have a direct effect on labour productivity growth in retail distribution. The indirect effect of deregulation on average sales per outlet, and thereby labour productivity growth, is correctly signed but not significant. All in all, hypothesis 1 cannot be accepted.

There are several limitations to this model and its results.

First, the regressions do not meet all assumptions required to perform regression analysis, which causes the results to be biased. More observations are needed to overcome this obstacle.

Second, data on capital stocks are based on own calculations, which rely on an overly simplified model (equation (3)) and assumptions. Although it is a theoretically appealing solution for data limitations, the quality and accuracy of the data must be improved.

Third, datasets based on numerous sources may contain data that are not all based on the same methodology. This implies, for example, that not all the labour productivity data used in the datasets will necessarily measure labour productivity in exactly the same way.

Fourth, price-cost averages are only an approximation of price-cost margins, but certainly not a substitute measure. Using the former introduces further bias in the results, as the dependent variable does not accurately measure the extent of regulation, or, for that matter, price-cost margins.

Fifth, value margins are assumed to remain constant in order to construct the scale variable.

However, this defies the notion that deregulation should lower price-cost averages. This is because price-cost averages are closely related to value margins, as both reflect the difference between the price received and price paid for goods and services.

Sixth, the scope of regulation and the economic domains it affects becomes limited if it is considered to (only) boil down to changes in competitive behaviour, which in turn affects price-cost margins. That is, price-cost margins are but a crude proxy for regulations.

Finally, price-cost margins do not allow specific regulations to be analysed individually. This means that no distinction can be made between different types of regulations and their unique effects on productivity growth. An attempt to solve this problem is made in the next section, using data on specific retail laws.

5. REGULATORY INDICATORS AND PRODUCTIVITY GROWTH

5.1. Introduction

Most studies on the relationship between regulation and performance focus on total measures of regulation, i.e. the entire set of laws pertaining to certain countries and, occasionally, industries.

Although detailed regulations constitute only a part of a larger regulatory framework, understanding the properties of specific laws may help improve the accuracy with which regulatory reforms take place to enhance economic performance.

Previous studies, certainly at an international level, were restricted to mostly qualitative research due to data limitations. Although these problems largely remain, the effects of various industry- specific regulations on productivity growth can be examined in this section using new time-series data on regulatory indicators in retail distribution.

The seven types of regulation are (1) administrative burdens (AB), (2) market entry requirements (MER), (3) legal monopolies (MON), (4) land use policies (LUP), (5) restrictions on opening hours (OH), (6) price controls (PC), and (7) direct government control over business enterprises (GC).

This section is organised as follows. The empirical framework and data sources are presented first, followed by a brief impression of countries that have the most flexible and strictest industry-specific laws. The section concludes with an overview and discussion of the empirical results.

5.2. Methodological framework and data sources

The relation between regulation and productivity growth is estimated using pooled country-data and equation (3),

ε α

α α

α

α + + + + × +

⎟ =

⎟

⎠

⎞

⎜ ⎜

⎝

Δ ⎛

rct rct ct rct

t c

t

c

REG REG SCALE TIME REG

H Y

, , 4

, 3

2 , , 2 , , 1 0 ,

log

, (3)

where index

c

refers to country, index

t

to time, index

r

to one of the seven specific types of regulation,

Y

represents value added,

H

hours worked,

REG

the regulatory indicator,

SCALE

controls for sales per outlet (i.e. scale effects),

TIME

is a dummy variable where 1 represents the period 1999-2002, and

ε

is the error term. Contrary to equations (1) and (2), equation (3) does not include variables for labour or capital growth due to limited data and degrees of freedom.