An empirical analysis of productivity growth in a Portuguese retail chain using Malmquist productivity index

ARTICLE IN PRESS

Journal of Retailing and Consumer Services 11 (2004) 269–278

An empirical analysis of productivity growth in a Portuguese retail chain using Malmquist productivity index Carlos Pestana Barros*, Carlos Alves Instituto de Economia e Gestao, Technical University of Lisbon, Rua Miguel Lupi, 20, Lisbon 1249-078, Portugal

Abstract This paper estimates total productivity change and decomposes it into technically efficient change and technological change for a Portuguese retail store chain with data envelopment analysis. The benchmarking procedure implemented is an internal benchmarking, where the stores in the chain are compared against each other. The aim of this procedure is to seek out those best practices that will lead to improved performance throughout the whole chain. We rank the stores according to their total productivity change for the period 1999–2000, concluding that some stores experienced productivity growth while others experienced productivity decrease. Managerial implications arising from the study are considered. r 2003 Elsevier Ltd. All rights reserved. Keywords: Chain retailing; Productivity change; Malmquist index

1. Introduction The Portuguese retail sector is confronted at the present time with several threats that clouds its future independence. These threats are the following: first, the increasing number of major international retailers (in particular, from Spain and France) entering the Portuguese retail market, intensifying the competition and the shrinkage of the margins; second, the small dimension of the Portuguese retailers, which prevents expansion into the European market, lacking the economies of scale that exist for larger operators who can benefit from doing business in several contiguous markets; third, an inadequate state policy that has prevailed in recent years, restricting the growth of hypermarkets and the shrinkage of small grocery stores (Farhangmehr et al., 2000); fourthly, the small dimension of the market and the relative shallowness of disposable income; fifth, the widespread availability of expensive, imported brands of all consumer goods, out of proportion to the market’s average purchasing power. This is currently in sharper focus, owing to the effects of a severe economic downturn; lastly, the existing structural rigidities in the labour market. *Corresponding author. Department of Economics, Technical University of Lisbon, Rua Miguel Lupi, 20, Lisbon 1200-725, Portugal. E-mail address: [email protected] (C.P. Barros). 0969-6989/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0969-6989(03)00053-5

The national retail industry reacts to these threats by attempting to increase the efficient use of inputs. One procedure to improve competitiveness is benchmarking. Benchmarking is the outcome of an investigation into an industry’s best practices, in order that their generalised application might lead to improved performance throughout the whole industry (Walters and Laffy, 1996; Dunne et al., 1992). The efficiency of retail stores is a major theme in contemporary research, i.e. Balakrishan et al. (1994), Athanassopoulos (1995), Kamakura et al. (1996), Thomas et al. (1998) and Donthu and Yoo (1998). Among the benchmarking techniques, data envelopment analysis (DEA), a nonparametric technique, has been used previously, for example in Thomas et al. (1998) and Donthu and Yoo (1998). In the present paper, we analyse the intra-chain comparative efficiency of a major Portuguese retail company, assessing the efficiency of a sample of individual stores by applying a variety of metrics to measure inputs and outputs that combine financial, as well as operational, dimensions. Moreover, we evaluate total productivity with the Malmquist index. The contribution of this paper to literature on the retail sector is based on the application of the Malmquist index to evaluate retail efficiency. The paper is organised as follows. In Section 2, we describe the contextual setting, describing the Portuguese retailing sector in order to shed some light on the

ARTICLE IN PRESS C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

270

threats mentioned above. In Section 3, we survey the existing literature on the topic, with a view to highlighting the contribution that the present paper seeks to make. In Section 4, we explain the theoretical framework supporting the model used. In Sections 5 and 6, we present the data and results. In Section 7, we consider the managerial implications of the study. In Section 8, we put forward the limitations and possible extensions of the study and finally, in Section 9, we make our concluding remarks.

2. Contextual setting On Portugal’s accession to the European Union in 1986, the country’s retail market embarked on a course of profound changes that are reflected in Table 1. These changes have led to the rapid and widespread expansion of hypermarkets and supermarkets throughout the country and to the concomitant decline of small selfservice stores, grocery shops and pure food stores, which were formerly the nation’s leading retailers of foodstuffs and other domestic products. This trend induced a political reaction on the part of the grocery proprietors (i.e. small and medium businesses) whose organisation, the Portuguese Association of Trade, lobbied successfully for the government to delay or slowdown the market-driven evolution from grocery stores to hypermarkets. The result was the laws that restricted the ceiling area in function of the population density (Farhangmehr et al., 2000). These laws remain on the Statute Book today, but the present government has promised their repeal in the future. Today, the Portuguese retail sector is dominated by five commercial groups which account for most of the country’s hypermarkets and supermarkets, as depicted in Table 2. The Portuguese groups among these five ! large retailers are Sonae and Jeronimo Martins;

Competition from overseas exists in the form of the French enterprises, Intermarche! , Auchan and Carrefour. Table 2 presents some characteristics of these groups. Sonae Distribui@*ao is the largest Portuguese retail commercial group, and is part of a conglomerate that has interests in the wood agglomerate industry (in which it is the world leader), media, tourism and construction. In the retail field, Sonae has the large chain of supermarkets, Modelo, including the smaller chain, Modelo Bonjour, in addition to the hypermarket chain, Continente and other, specialised retail stores. ! Jeronimo Martins is also a Portuguese group, with minor partnerships with Unilever. It owns the Pingo Doce supermarket chain and the Feira Nova hypermarkets. The Feira Nova chain was obtained through acquisition. This group also manages the cash-and-carry chain, Recheio. The French group, Intermarche! is organised in a mix of franchising and cooperatives, with about 150 stores owned by individual, independent entrepreneurs. The French Auchan group made its entry into ! Portugal by acquiring the P*ao de A@ucar, a Brazilian hypermarket chain, which had been the first example of such a dimension of retailing in Portugal when its first ! store opened in 1970. The Group P*ao de A@ucar left the market in 1991 with a management buy-out, the chain later being sold on to Auchan in 1997. This is one of the long-established French chains, which has expanded beyond France’s borders to create an empire of hypermarkets and supermarkets across the EU. It sold the discount chain operation, Mini-Pre@o to its rival, Carrefour. Auchan presently has14 hypermarkets. The third French group in the Portuguese market, Carrefour, arrived in 1990 with a joint venture with a Portuguese bank, but later became independent. They manage hypermarkets under their own brand-name, as well as the soft-discount store chains, Dia and Mini-Pre@o.

Table 1 Percentage market share (sales) of different types of stores (1988–1999) Years

Hyper

Supers

Large super

Small super

Self-service

Small grocers

Pure food stores

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

11.7 16.8 21.1 25 30.9 36.2 40.4 42.4 39.9 37.8 37.8 37.2

18.8 19.5 19.1 20.7 21.5 22.5 25.2 28.7 33.4 33.8 39.9 42.4

N/A N/A N/A N/A N/A N/A N/A 14.8 17.9 20.0 21.7 24.1

N/A N/A N/A N/A N/A N/A N/A 13.9 16.5 17.8 18.2 18.3

19.6 17.8 15.6 12.9 11.4 10.1 8.7 7.9 7.3 7.2 6.6 6.3

40.6 38.8 38.6 37.0 31.9 27.0 22.5 18.4 16.0 15.3 14.1 12.7

9.3 7.1 5.6 4.5 4.3 4.3 3.2 2.6 2.4 1.9 1.6 1.4

Source: Anu!ario da Distribui@*ao Portuguesa 2001. N/A=not available.

ARTICLE IN PRESS C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

271

Table 2 The dominant players and their chains in the Portuguese retail sector, 2000 Group

Trade-mark

Sales (Euros)

Square metres

Employees

Earnings

Stocks

Continente Modelo Modelo Bonjour Specialised retail

2 294 470.3 1 157 211.1 788 100.68 49 879.79 299 278.74

309 000 117 000 110 000 13 000 69 000

16 980 6980 5200 460 1810

1 817 971 N/A N/A N/A N/A

5 799 543 N/A N/A N/A N/A

Pingo Doce Feira Nova

1 437 002.4 779 924.16 657 078.24

245 498 142 793 102 705

13 133 9188 3945


3 538 415
2 968 527
569 888

9 476 383 4 217 139 5 259 244

877 884.3

160 795

N/A

N/A

N/A

Jumbo

841 970.85

82 100

5291

2 200 000

37 943 023

Carrefour Dia/Mini Pre@o

654 241.88 329 185.27 325 056.61

119 085 45 772 73 313

3981 2055 1926


637 568
637 568 N/A

14 553 790 14 553 790 N/A

Alisuper Select/Shell Pluricoop Supercompra

151 878.89 46 682.53 43 899.49 33 491.42 27 805.43

21 805 N/A 5959 10 319 5527

1871 576 524 431 340

365 779 99 944 N/A 33 330 232 505

N/A N/A N/A N/A N/A

Sonae

Jeronimo Martins

Intermarch!e Auchan Carrefour

Others

Source: Financial reports of the enterprises in 2000. N/A=not available.

Beyond these five large retail groups, there are a number of small retailers which are prominent at a regional level, such as Alisuper, operating in the Algarve with free service stores; the convenience stores Select, belonging to the Shell petroleum company, operated in gas stations and elsewhere; and Supercompra which is active in the Leiria region. Other representative retailers are the German groups Lidl, which manages 140 hard-discount supermarkets throughout the country, and the cash-and-carry Makro (Metro group) with nine stores. Efficiency is a key element in this segment, because retail chains operate in an oligopolistic market, where small units compete to attract clients in order to be economically viable (Walters and Laffy, 1996). Therefore, the level of competition within the market demands efficiency. Despite the importance of efficiency, there is a paucity of research into the issue.

3. Literature survey Early studies on retailing efficiency focussed on partial aspects of productivity, such as labour productivity (Ratchford and Brown, 1985); other aspects under the control of retail management that affect the efficiency of a store, such as merchandise assortment

(Mahajan et al., 1988), location (Mahajan et al., 1985), pricing (Mahajan, 1991) and promotion (Weitzel et al., 1989), besides aspects beyond the management’s control, such as employment patterns, business cycles and trading area factors (Doutt, 1984; Lusch and Moon, 1984). Intra-chain comparative efficiency in retailing, which is the issue analysed in the present paper, has been addressed by several authors. Balakrishnan et al. (1994) used DEA to evaluate relative spatial efficiency of locations of a network of retail outlets and how the imposition of threshold requirements alters their spatial efficiency. The efficient scores of relative spatial efficiency were obtained with the output, demand variation and the inputs, average distance and demand uncovered. Athanassopoulos (1995) used DEA as a basis for the development of a performance-improvement decision aid system applied to restaurants. The bar area (ft2) and the number of covers were used as adjustable inputs, with market size (potential customers), the number of restaurants in a 1-mile radius, and the number of restaurants in a 3-mile radius comprising the uncontrollable inputs, while food sales (in value) and drinking sales (value) were used as outputs. Kamakura et al. (1996) employed a fuzzy clusterwise translog cost regression to evaluate the 188 branches of a commercial bank in Latin America. The outputs used were the

ARTICLE IN PRESS 272

C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

volume of cash deposits, the volume of other deposits, the volume of funds in transit in the branch and the volume of service charged. Inputs used were labour (total number of man-hours of direct clerks allocated to the branch) and the area (in square meters). Thomas et al. (1998) analysed intra-chain comparative efficiency using 500 domestic retail outlets of a leading specialist retailer in the USA. The inputs and outputs were selected by a Delphi approach on central and regional managers. These authors selected 16 inputs (average number of full-time employees per square foot of selling space times 10,000; the ratio of average number of fulltime to part-time employees; the total annual salaries and wages divided by payroll hours; the average hourly employee tenure in years; the average store managers’ tenure in years; the age of the store in years; the base rent plus other occupancy expenses divided by total selling square footage; dollars of annual operating expenses per store; population per store in the market; the average annual household income in a 2-mile radius; the number of households in a 2-mile radius; the distance in miles to the nearest alternative store; the total average dollars inventory at cost; the average dollar size of transactions; the percentage of annual turnover; and lastly, the dollar shrinkage divided by inventory dollars). The two outputs used were sales and profits. These authors identified store management evaluation and critical success factors. Donthu and Yoo (1998) analysed 24 outlets of a fast-food restaurant chain. The outputs used were sales (value) and customer satisfaction (a 5-point scale). The inputs used were store size, manager tenure, store location (inside a shopping mall versus free-standing) and promotion/give-away expenses.

4. Theoretical framework In this paper, we adopt the efficient frontier approach using the Malmquist productivity index, based on DEA. The Malmquist productivity index allows changes in productivity to be broken down into changes in efficiency and technical change. To set the scene for our productivity measurement, we adopt the framework set in the paper by Fare et al. (1990), Hjalmarsson and Veiderpass (1992) and Price and Weyman-Jones (1996). Fig. 1 shows two observations on the input ðxÞ and output ðyÞ bundles used by a retail store at time t and t þ 1: The objective is to measure the productivity growth between t and t þ 1 in terms of the change from input–output bundle zðtÞ to input–output bundle zðt þ 1Þ: The measurement of the productivity is made through the potential production frontier that is imposed on the production bundle in Fig. 2. The production frontier represents the efficient levels of output ðyÞ that can be

y Z(t+1) y(t+1)

Z(t)

Y(t)

x X(t+1)

X(t)

Fig. 1. Productivity growth in period t and t þ 1:

Frontier (t+1)

y

z(t+1)

y(t+1)

Frontier (t)

z(t)

Y(t)

X(t)

X(t+1)

0

L N P

Q R

x

S

Fig. 2. Production frontier in period t and t þ 1:

produced from a given level of input. If the store is technically efficient in period t; it produces along the frontier the maximum output attainable. Point zðtÞ corresponds to a technically inefficient store, which uses more than the minimal amount of input to make a given level of output. The bundle zðtÞ can be reduced by the horizontal distance ratio=ON/OS in order to make the production technically efficient. The frontier can shift over time. By analogy, the bundle zðt þ 1Þ should be multiplied by the horizontal distance ratio=OR/OQ in order to achieve comparable technical efficiency. Since the frontier has shifted in the meantime, zðt þ 1Þ is technically inefficient in t þ 1: In order for zðt þ 1Þ to be efficient in period t þ 1; it must be reduced by the horizontal distance=OP/OQ. The relative movement of a production observation over time may be because stores are catching up with their own frontier, or because the frontier is shifting up over time. The Malmquist index of productivity growth ðMÞ is the ratio of the two distances in period t and t þ 1: To break down the index in catching up (MC) and shifting up (MF) effects, we rescale M by multiplying it top and bottom by OP/OQ: OR  ON OQ  OS   OP  ON OR ¼  OQ  OS OP

M¼

¼ MCMF:

ð1Þ

ARTICLE IN PRESS C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

We verify that the relative efficiency distances of each observation from its own frontier measure the catchingup effect. The frontier shift effect is measured by the relative distance between the frontiers at period t þ 1: This is the benchmark used by Hjalmarsson and Veiderpass (1992) and Price and Weyman-Jones (1996). An alternative benchmark used by Fare et al. (1990) measures the frontier shift as the relative distance between the frontiers at t and t þ 1 (ON/OL). Formally the Malmquist index is based on the output distance function defined as     t 1 t d ðx ; y Þ  inf y : x ; y eSt ; y T

t

t

ð2Þ

where x denotes a vector of inputs, y are outputs, S is the technology set, and superscript T denotes the technology reference period, usually T ¼ t or T ¼ t þ 1; and 1=y defines the amount by which outputs in year t could have been increased, given the inputs used, if technology for year T had been fully utilised. Caves et al. (1982) showed that productivity movements could be measured by a multi-input, multi-output Malmquist index when input and output data are available in physical units, so that no price index problems arise. They argue that the distance function dðx; yÞ can be used in the construction of the Malmquist index and measured the Malmquist index of change between t and t þ 1 as the ratio d T ðxtþ1 ; ytþ1 Þ=d T ðxt ; yt Þ:

ð3Þ

Fare et al. (1994) proposed to measure the Malmquist index as the geometric mean of such indexes calculated both for year t and t þ 1 reference technologies as Mðxtþ1 ; ytþ1 ; xt ; yt Þ ¼



 d t ðxtþ1 ; ytþ1 d tþ1 ðxtþ1 ; ytþ1 Þ  : d tþ1 ðxt ; yt Þ d t ðxt ; yt Þ ð4Þ

Fare et al. (1994) factor this expression into the product of technical change and efficiency change as Mðxtþ1 ; ytþ1 ; xt ; yt Þ  1=2 d tþ1 ðxtþ1 ; ytþ1 Þ d t ðxtþ1 ; ytþ1 Þ d t ðxt ; yt Þ ¼ : ð5Þ d t ðxt ; yt Þ d tþ1 ðxtþ1 ; ytþ1 Þ d tþ1 ðxt ; yt Þ The ratio outside the brackets is the index of change in technical efficiency (i.e. change in the distance of observed production from the current maximum feasible production) between years t and t þ 1; while the bracketed term is the index of change in technology (or technical change) between two periods evaluated at xt and xtþ1 :

273

The Malmquist index is measured either with the distance function or alternatively, with the reciprocal of the input distance function yðx; yÞ ¼ ½1=dðx; yÞ : This reciprocal of the input distance function yðx; yÞ is the smallest ratio by which an input bundle can be multiplied and still be capable of making a given level of output. The reciprocal distance function is equivalent to the measurement of technical efficiency proposed by Farrel (1957) and is the basis for the efficiency distance ratios used in analysing Fig. 2. When the Farrel measurement of technical efficiency (reciprocal of the input or output distance) is used in constructing the Malmquist index, we obtain productivity growth if M > 1 and productivity regression if Mo1: The Malmquist index (Malmquist, 1953) allows changes in productivity to be broken down into changes in efficiency and technological change. Unlike the econometric stochastic frontier approach, it offers a different rate of technical change for each individual unit, which is more adequate for the purposes of this section, i.e. the analysis of technical change by companies. Moreover, since it is estimated with a nonparametric methodology (DEA), it does not need to impose any functional form on the data, neither to make distributional assumptions for the inefficiency term. This efficiency measurement assumes that the production function of the fully efficient store is known. In practice, this is not the case, and the efficient frontier must be estimated from the sample data. In these conditions, the frontier is relative to the sample considered in the analysis. We developed a Malmquist productivity estimate from mathematical programming models of the frontier production function. For recent surveys, see Fare et al. (1994), Coelli (1996), Coelli et al. (1998) and Thanassoulis (2001).

5. Data To estimate the production frontier, we used crosssection data for the years 1999 and 2000, obtained from one of Portugal’s leading hypermarket and supermarket chains, on 47 of its retail outlets. The outlets that are considered in the analysis are those listed in Table 4. The data are used for internal marketing control purposes. To choose the inputs of the DMUs, we must take into account the distinction between controllable and uncontrollable factors. There are two alternatives for making this distinction; either to exclude the uncontrollable inputs from the analysis and thereby, no allowance is made for factors beyond managerial control, or to include them in the analysis, allowing for variables beyond the control of the management. Including nondiscretionary inputs in the linear programming (LP) model for DEA amounts to an assumption of free disposability of these inputs, which is not necessarily a

ARTICLE IN PRESS 274

C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

Table 3 Characteristics of the inputs and outputs Variables

Units

Range

Mean

Square deviation

Outputs Sales Operational results

Value in KPTE Value in KPTE

1 250 000–12 679 000 126 500–1 597 097

3 739 276 385 677.3

2 059 957.64 256 978.06

Controllable Inputs Number of full-time equivalent employees Cost of labour Number of cash-out points Stock Other costs

Number Value in KPTE Number Value in KPTE Value in KPTE

23–198 72 917–520 764 7–44 73 317–524 990 50 383–405 690

62 188.196 15.63 245 337 125 614.3

31.43 89 638.3 8.13 82 075.8 62 410.07

PTE = Portuguese currency, the escudo, used prior to the adoption of the Euro in 2002 (rate: A1 ¼ 200 468PTE). The values in PTE are measured on a yearly basis. KPTE = thousand PTE.

realistic assumption (Ray, 1991; Worthington and Dollery, 2002). Given the fact that uncontrollable variables demonstrate almost no change in two consecutive periods, we proceeded to analyse the outlets without uncontrollable inputs. Frontier models require the identification of inputs (resources) and outputs (transformation of resources). Several criteria can be used in their selection. The first empirical criterion is the availability of inputs and outputs. It is important for the applicability of the model results, as well as the retail stores’ ‘‘buy in’’ to the process, that the measurements of inputs and outputs are relevant, adequately measurable, that appropriate archival data is available and to bear in mind the dictum, ‘‘more is better’’, in the case of outputs. Usually, the available archival data criterion is the first to be applied, since it encompasses all the other abovementioned criteria. Secondly, the literature survey is a way to ensure the validity of the research and is thus another criterion to take into account. The last criterion for measurement selection is the professional opinion of store managers. These three criteria have been employed in this paper to select the inputs and outputs. We measured inputs by nine variables listed in Table 3, following Walters and Laffy (1996). We ensured the DEA convention that the minimum number of DMUs is greater than three times the number of inputs plus output (47>3(2+5)]. We measured labour by the number of full-time equivalent employees, the number of part-time employees, and the cost of labour. We measured capital by the number of cash-out points and the value of the stock inventory. Moreover, we allowed for other materials in the production function, including other costs (total costs excluding labour costs). We measured output by two indicators: sales and operational results. We verify that the mean outlet of the Portuguese multi-market retail chain in question is characterised as having 15 cash-out points and 245.337 KPTE worth of stock, with 62 full-time employees, earning a monthly salary of 188.196 PTE.

6. Results The Malmquist index can be calculated in several ways (Caves et al., 1982). In the present study, we estimate an output-oriented Malmquist productivity index, based on DEA. Output-oriented efficiency measurements are adequate, if we assume that retail stores behave in a competitive way. In output-oriented models, such as the one adopted in this paper, the DEA seeks to identify technical inefficiency as a proportional increase in output usage. However, it is possible to measure an input-oriented model of technical inefficiency as a proportional decrease in input use. As far as retailers are concerned, market orientation seems to be the natural choice, due to their competitive position in the market. However, since input and output Malmquist indices are equal (Thanassoulis, 2001, p. 182), this specification is more a theoretical issue than a practical one. The DEA allows the estimation of TFP as a Malmquist index. The results are presented in Table 4 with the TFP index broken down into efficient change (diffusion or catch-up component) and technical change (innovation or frontier-shift component). Moreover, we break down efficient change into pure efficient change and scale-efficient change. In Table 4, we verify that the total productivity change (Malmquist index) is higher than 1 for three stores out of 47 stores, denoting that many stores did not experience any gain in total productivity in the period considered. Nevertheless, the mean score is 0.872, which is far from one, signifying that for the majority of the stores, the total productivity suffered a decrease in the period. The change in technical efficiency is defined as the diffusion of best-practice technology in the management of the activity and is attributed to investment planning, technical experience, and management and organisation in the stores. For the period under analysis, we verify that it is higher than one for the majority of the stores,

ARTICLE IN PRESS C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

275

Table 4 Technically efficient change and technological change of Portuguese retail stores: 1999–2000. Stores

MC-technically efficient change

MF-technological change

Pure technical change

Scale-efficient change

Total productivity change

Portalegre Tapada Merces Montijo Albufeira Tomar Olh*ao Beja Fanzeres Santarem Caldas Gaia Felgueiras Torres Novas Porto Alto Vila Real Maia Covilha Marco Canaveses Loule S.J.Madeira Elvas Portim*ao St.Tirso Mafra V.F. Xira Evora Torres Vedras Castelo Branco Ovar Chaves Cartaxo Rio tinto Alcoba@a Pa@os Fereira Silves Guarda Sintra Bragan@a Abrantes Amarante Agueda Vila Conde Rebordosa Braga S.Cosme Vale de Cambra Faro II Mean

1.145 1.019 1.060 1.019 1.020 0.783 1.000 1.055 1.044 1.000 0.920 1.015 1.000 1.008 0.983 0.984 0.955 1.005 1.000 0.946 0.978 1.000 0.933 0.936 0.995 0.952 0.949 0.918 0.921 0.970 0.895 0.932 0.888 0.896 0.948 0.893 0.893 0.953 0.851 0.901 0.844 0.818 0.758 0.823 0.753 0.758 0.679 0.931

0.926 1.027 0.969 0.971 0.969 0.878 0.986 0.926 0.935 0.974 1.052 0.952 0.958 0.946 0.967 0.940 0.967 0.913 0.916 0.963 0.930 0.909 0.966 0.957 0.898 0.931 0.927 0.954 0.940 0.889 0.953 0.916 0.949 0.937 0.884 0.932 0.921 0.844 0.940 0.876 0.915 0.926 0.986 0.886 0.941 0.877 0.906 0.936

1.045 1.018 1.059 0.989 1.068 0.945 1.000 1.119 1.038 1.000 1.000 1.031 1.000 1.036 0.988 1.083 0.978 1.113 1.000 0.946 0.979 1.000 0.997 0.985 1.064 1.002 0.997 1.037 0.989 1.000 1.040 0.997 0.930 1.029 0.981 1.031 1.000 1.000 0.934 0.955 1.019 0.945 1.000 0.979 0.911 1.000 0.793 1.000

1.096 1.001 1.000 1.030 0.955 0.829 1.000 0.943 1.000 1.000 0.920 0.984 1.000 0.974 0.996 0.908 0.976 0.903 1.000 1.000 0.998 1.000 0.936 0.950 0.935 0.950 0.952 0.886 0.932 0.970 0.860 0.934 0.956 0.871 0.966 0.866 0.893 0.952 0.903 0.943 0.829 0.866 0.758 0.840 0.826 0.758 0.856 0.931

1.060 1.048 1.026 0.989 0.988 0.987 0.986 0.977 0.976 0.974 0.968 0.965 0.958 0.954 0.951 0.925 0.924 0.918 0.916 0.911 0.909 0.909 0.901 0.896 0.894 0.886 0.881 0.876 0.866 0.862 0.853 0.853 0.843 0.839 0.837 0.832 0.822 0.804 0.801 0.789 0.772 0.758 0.747 0.729 0.708 0.665 0.615 0.872

signifying that the there was growth in technical efficiency. The mastering or diffusion of best-practice technology improved in the period. Moreover, the breaking down of the technical efficiency into pure technical efficiency and scale efficiency shows mixed results with stores which obtained simultaneous gains in both areas and others which obtained gains in one, but losses in the other. The improvement in pure technical efficiency denotes that there was investment in organisational factors associated with the store management,

such as better balance between inputs and outputs, marketing initiatives, more accurate reports, improvement in quality, and so on. The scale efficiency, which is the consequence of dimension, improved in the period for many stores. It is noteworthy to verify that scale efficiency change is the only variable for which the mean value is one. Technological change is the consequence of innovation, that is, the adoption of new technologies, by bestpractice stores. We verify that, with the exclusion of a

ARTICLE IN PRESS 276

C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

small group of stores, this index is smaller than one, which indicates that innovation deteriorated in the period, meaning that there was not investment in new technologies (methodologies, procedures and techniques) and the commensurate skills upgrades related to it. Overall, we observe four combinations of technical efficiency and technological change: (i) A single store in which the improvements in technical efficiency co-exist with improvements in technological change, namely, Tapada das Merc#es. This is the best-performing store in the period, with improvement registered in technical efficiency, denoting up-graded organisational factors associated with the use of inputs (low number of employees, low costs), with the use of outputs (high number of sales and operational results) and the relation between inputs and outputs. Moreover, this store demonstrates economies of scale related to its dimension, as well as pure, technically efficient growth. (ii) Stores in which the improvements in technical efficiency co-exist with deteriorating technological change, namely, Portalegre, Montijo, Albufeira, Tomar, Beja, Fanzeres, Santarem, Caldas, Felgueiras, Torres Novas, Porto Alto, Marco de Canavezes, Loule! and Portim*ao. These are stores with up-graded organisational factors, but without the innovation inherent to investment in new technology which would leverage the organisational factors. These stores need to acquire new technology and the necessary, commensurate skill upgrades in order to improve their performance. (iii) Stores in which deteriorating technical efficiency co-exists with deteriorating technological change, namely, Olh*ao, Vila Real Maia, Covilh*a, S*ao Jo*ao da Madeira, Elvas, Santo Tirso, and all the stores hierarchically listed below Santo Tirso in the table. These are the inefficient stores that need both to improve their organisational factors related to the use of inputs and outputs and their interrelationship and to adopt new technology related to upgrading organisational skills. (iv) One store in which deteriorating technical efficiency co-exists with improvements in technological change, namely, Gaia. This is a store that invested in new technology, but not in the use of inputs and their relationship with outputs, which is a contradictory procedure. Hence, our findings encompass all possible combinations, signifying that there is room for adjustment in almost all of the above-mentioned stores in order to achieve best-practice procedures in retail store management.

7. Managerial implications of the study This paper has proposed a simple framework for the evaluation of retail chains and the rationalisation of their management activities. The analysis is based on a DEA model that allows for the incorporation of

multiple inputs and outputs in determining the relative efficiencies. Benchmarks are provided for improving the operations of poorly performing stores. We emphasise two managerial implications of our findings. First, the group management should change its managerial procedure in order to adopt an efficient, enhancedincentive policy, which would enable the inefficient stores to catch up with the efficient frontier. Second, the adjustment must be based on the improvement of technical efficiency, as well as technological change. Technical efficiency is, in a broad economic sense, about a unit allocating resources without waste and refers to a movement towards, or away from, the bestpractice frontier production function. A movement towards the latter is an improvement, while a movement away is a deterioration. In a dynamic way, it is characterised as efficiency change (diffusion), related to change between two successive technical-efficiency frontiers. Technical inefficiency is a consequence of one or more of the following factors: (i) sub-structural rigidities associated with the policy of rotation of senior store-managers on a 2-yearly basis adopted by the retail chain analysed, in order to prevent the principal–agent relationship (Jensen and Meckling, 1976); (ii) structural rigidities associated with the labour market; (iii) unequal access to information on operational activities between the various outlets in the chain (Tucker and Tucci, 1994); (iv) time lags in outlets’ acquisition of new technology and the necessary, commensurate skills upgrades among the staff; (v) differential incentive systems; (vi) organisational factors associated with Xefficiency (Leibenstein, 1966); (vii) organisational factors associated with human capital, such as lack of incentives for the improvement of efficiency (Mirrelees and Miller, 1996); (viii) dimensional factors associated with scale and scope economies (Johnston et al., 2000). Due to any, some or all of these factors, stores may produce at a level below their potential which is the maximum possible output, given the production environment which applies to retailing activity. The results of our research highlight the importance of economies of scale in the retail chain analysed. Moreover, less efficient stores are situated in middle-dimension cities, meaning that the location and agglomeration plays a role in demand attraction, as well as in diseconomies of scale, which contributes to the efficient scores. Finally, the tenure of the store’s senior manager, which is currently based on a 2-yearly rotation principle, is also a contributory factor to inefficiency. Technological change (innovation) is, in a broad economic sense, about investment that improves the total productivity of a productive unit. It arises due to capital accumulation, which drives the adoption of technology by best-practice decision units shifting the frontier of technology. In retailing, technological change means investing in new methods, procedures and

ARTICLE IN PRESS C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

techniques with the aim of improving the results. In this context, a good policy should be to copy foreign methodologies that have shown excellent results. The general conclusion is that there is room for improvement in the management of the stores. However, one note of caution should be take into account, namely, that the data set is short, and thus, the conclusions are limited. Furthermore, as DEA only determines relative efficiency, it cannot identify all of the inefficient stores. All of the units in the sample may, in fact, be inefficient (Bessent and Bessent, 1980).

8. Limitations and extensions of this study The DEA model does not impose any functional form on the data, nor make distributional assumptions for the inefficiency term. This efficiency measurement assumes that the production function of the fully efficient outlet is known. In practice, this is not the case and the efficient isoquant must be estimated from the sample data. In these conditions, the frontier is relative to the sample considered in the analysis. Moreover, without statistical distribution hypotheses, the DEA does not allow for random errors in the data, assuming away measurement error and chance as factors affecting outcomes. A variety of extensions to this paper can be undertaken. First, in this analysis, the DEA model allowed for complete weight flexibility. In situations in which some of the measures are likely to be more important than others, DEA allows for restricting factor weights through linear constraints. These linear constraints represent ranges for relative preferences among factors based on managerial input. Such analysis enables effective incorporation of managerial input into the DEA evaluations. Second, the input and output dimensions considered are context specific. More comprehensive input and output measurements, namely, allowing for no discretionary factors, such as environmental, socio-economic and quality inputs and outputs, need to be taken into consideration. The influence of no discretionary variables, excluded from the analysis, amounts to an assumption that these factors are constant across the sample. Third, non-parametric, or alternatively, parametric, free-disposal hull analysis can be used to assess the efficiency scores. However, previous research has shown that the DEA scores are inferior in value to econometric scores, but the ranking is preserved (Bauer et al., 1998). Third, the data set is short, thus the conclusions are limited. In order to be more conclusive, we would need to have a panel data set. Finally, The hypothesis of the homogeneity of the outlets under analysis is based on their nature, that they compete in the same market and that they all have the

277

same stockholders. However, non-discretionary factors can render them non-homogenous.

9. Conclusion This article has proposed a simple framework for the evaluation of retail outlets and the rationalisation of their operational activities. The analysis is based on a DEA model that allows for the incorporation of multiple inputs and outputs in determining the relative efficiencies. Benchmarks are provided for improving the operations of poorly performing retail outlets. Several interesting and useful managerial insights and implications from the study are raised. The general conclusion is that the majority of the outlets analysed are efficient, while a proportion of them is not efficient. For the latter group, we identified peer groups among the efficient outlets, as well as the slacks that they should adjust in order to achieve the efficient frontier. The results suggest that scale economies are determinant factors of efficiency in this sector. More investigation is needed to address the limitations mentioned.

References Athanassopoulos, A., 1995. Performance improvement decision aid systems (PIDAS) in retailing organisations using data envelopment analysis. The Journal of Productivity Analysis 6, 153–170. Balakrishnan, P., Desai, A., Storbeck, J., 1994. Efficiency evaluation of retail outlet networks. Environment and Planning B 21 (4), 477–488. Bauer, P.W., Berger, A.N., Ferrier, G.D., Humphrey, D.B., 1998. Consistency conditions for regulatory analysis of financial institutions: a comparison of frontier efficiency methods. Journal of Economics and Business 50 (2), 85–114. Bessent, A.M., Bessent, E.W., 1980. Determining the comparative efficiency of schools through data envelopment analysis. Educational Administration Quarterly 16, 57–75. Caves, D.W., Christensen, L.R., Diewert, W.E., 1982. The economic theory of index numbers and the measurement of input, output and productivity. Econometrica 50, 1393–1414. Coelli, T.J., 1996. A guide to DEAP version 2.1: a data envelopment analysis (Computer) program. Working Paper No. 8/96, Center for Efficiency and Productivity analysis, University of New England, Armidale, Australia. Coelli, T.J., Prasada, R., Battese, G.E., 1998. An Introduction to Efficiency and Productivity Analysis. Kluwer Academic Publishers, Boston. Donthu, N., Yoo, B., 1998. Retail productivity assessment using data envelopment analysis. Journal of Retailing 74 (1), 89–105. Doutt, J., 1984. Comparative productivity performance in fast food retail distributions. Journal of Retailing 60, 98–106. Dunne, P., Lusch, R., Gable, M., Gebhardt, R., 1992. Retailing. South-Western Publishing, Cincinnati, OH. Fare, R.S., Grosskopf, S., Lovel, C.A.K., 1994. Production Frontiers. Cambridge University Press, Cambridge. Fare, R., Grosskopf, S., Yaisar Warng, S., Li, S., Wang, Z., 1990. Productivity growth in Illinois Electric Utilities. Resources and Energy 12, 383–398.

ARTICLE IN PRESS 278

C.P. Barros, C. Alves / Journal of Retailing and Consumer Services 11 (2004) 269–278

Farhangmehr, M., Marques, S., Silva, J., 2000. Consumer and retailer perceptions of hypermarkets and traditional retail stores in Portugal. Journal of Retailing and Consumer Services 7, 197–206. Farrel, M.J., 1957. The measurement of productive efficiency. Journal of the Royal Statistical Society, Series A 120 (3), 253–290. Hjalmarsson, L., Veiderpass, A., 1992. Productivity in Swedish electricity retail distribution. Scandinavian Journal of Economics 94 (Suppl.), 193–205. Jensen, M.C., Meckling, W., 1976. Theory of the firm: managerial behaviour, agency costs and capital structure. Journal of Financial Economics 3, 305–360. Johnston, A., Porter, D., Cobbold, T., Dolamore, R., 2000. Productivity in Australia Wholesale and Retail Trade. Productivity Commission, Canberra. Kamakura, W., Lenartowicz, T., Ratchford, B., 1996. Productivity assessment of multiple retail outlets. Journal of Retailing 72 (4), 325–330. Leibenstein, H., 1966. Allocative efficiency vs. ‘X-efficiency’. American Economic Review 56 (3), 392–414. Lusch, R., Moon, S., 1984. An explanatory analysis of the correlates of labour productivity in retailing. Journal of Retailing 60, 37–60. Mahajan, J., 1991. A data envelopment analytical model for assessing the relative efficiency of the selling function. Journal of Operational Research 53, 189–205. Mahajan, V., Sharma, S., Srinivas, D., 1985. An application of portfolio analysis for identifying attractive retail locations. Journal of Retailing 61 (4), 19–34. Mahajan, V., Sharma, S., Kerin, R., 1988. Assessing market opportunities and saturation potential for multi-store, multimarket retailers. Journal of Retailing 64, 315–332.

Malmquist, S., 1953. Index numbers and indifference surfaces. Trabajos de Estatistica 4, 209–242. Mirrelees, B., Miller, D., 1996. Retailing Management: a Best Practice Approach. RMIT Press, Victoria, Australia. Price, C.W., Weyman-Jones, T., 1996. Malmquist indices of productivity change in the UK gas industry before and after privatisation. Applied Economics 28, 29–39. Ratchford, B., Brown, J., 1985. A study of productivity changes in food retailing. Management Science 4, 292–311. Ray, S.C., 1991. Resource-based efficiency in public schools: a study of Connecticut data. Management Science 37, 1620–1628. Thanassoulis, E., 2001. Introduction to the Theory and Application of Data Envelopment Analysis: a Foundation Text with Integrated Software. Kluwer Academic Publishers, Dordrecht, Netherlands. Thomas, R.R., Barr, R.S., Cron, W.L., Slocum Jr., J.W., 1998. A process for evaluating retail store efficiency: a restricted DEA approach. International Journal of Research in Marketing 15, 487–503. Tucker, J., Tucci, L., 1994. Why traditional measures of earnings performance may lead to failed strategic marketing decisions. Journal of Consumer Marketing 11 (3), 4–17. Walters, D., Laffy, D., 1996. Managing Retail Productivity and Profitability. Macmillan Press, London. Weitzel, W., Schwarzkopf, A., Peach, E., 1989. The influence of employee perceptions of customer service on retail store sales. Journal of Retailing 69, 27–39. Worthington, A.C., Dollery, B.E., 2002. Incorporating contextual information in public sector efficiency analyses: a comparative study of NSW local government. Applied Economics 34, 453–464.