Credit Scoring for the Supermarket and Retailing Industry: Analysis and Application Proposal
Author
Raffaele Manini
Oriol Amat
Abstract
This paper develops and tests a credit scoring model focused on the supermarket and retailing industry which can help financial institutions in assessing credit requests coming from customers belonging to these industries category. The empirical study has the objective of answering two questions:
(1) Which ratios better discriminate the companies based on their being solvent or insolvent?
(2) What is the relative importance of these ratios?
To do this, several statistical techniques with a multifactorial focus have been applied. The overall approach is the same as the one in Altman (1968), but the application of the design as well as the purpose of it are different. Through the application of several statistical techniques, the credit scoring model has been proved to be effective in assessing credit scoring applications within the supermarket and retailing industry under certain conditions.
Keywords
- Credit scoring
- Supermarket
- Retailing industry
1. Introduction
Credit scoring, is the process of collecting, analyzing and classifying different variables related to credit in order to asses credit decisions, Hand and Jacka, (1998) and Anderson (2007). The importance of this field has been increasing over time reaching its peak during the last financial crisis in 2008 when credit ratings and financial institutions in general showed severe limitations in assessing credit requests and companies’ solvency in general. Academic research in the field of credit scoring has the goal of identifying the variables that have significant influence in companies’ probability of default, so to allow credit institutions to discriminate between solvent customers and insolvent customers before actually granting credit. In addition, a slightly improvement in the credit scoring assessment leads to significant positive results for the lender (Schreiner, 2002; Schreiner, 2004; De Young, Glennon and Nigro, 2008). However, credit scoring techniques, might be influenced by industry specific characteristics as well as the institution applying it. This paper builds on the work of Amat, Antón and Manini (2016) which addresses the limitation of industry specificity when applying a Z score model for credit scoring purposes. In fact, their model once tested on real companies’ data, had extremely high overall results in assessing the solvency of specific companies including the television industry, the airplane industry and much more (Amat, Antón and Manini, 2016). However, this model could not be properly applied to both financial institutions and supermarkets/retailing industries because of their industry specificity. This paper, tries to partially bridge this gap by providing a Z score applicable to the supermarket and retailing industry. The rest of the paper is organized as follow. The second part will briefly revise part of the existing literature in credit scoring. The third part will be dedicated to describing the research methodology and design. Finally, conclusions and limitations of the paper will be discussed.
2. Literature Review
The main idea behind credit scoring is trying to develop a model which can forecast with as high predictive power as possible, if a company will have solvency issues. The most traditional credit scoring models tend to apply a simple basic concept which is comparing customers’ profiles. If the profile of a credit applicant is closer to a solvent one then it would be considered solvent, so it would be granted credit otherwise it would be considered not solvent. Hence, the financial institution would deny the credit request (Abdou and Pointon, 2011). To make this kind of comparison, financial institutions can choose between two main techniques. The loans’ officer subjective assessment that is the process through which the financial officer in charge after having performed several checks about the credit applicant decides if granting him credit or not. The advantage of this approach is the possibility of taking into account most of the aspects being part of a credit request including qualitative ones (Amat, Antòn, Manini, 2016). The second methodology financial institutions can apply to decide about granting credit or not is the application of a credit scoring model. A statistical model taking into account several factors of the credit application, but mainly relying on quantitative data including historical performance of the credit applicant and financial stability of the latter. Nowadays, more complex statistical models have been developed, but at the same time behavioral science are playing a heavier role within the business science. Therefore, it is still not possible to neutrally and properly decide between the two methodologies without missing some part of the big picture. However, what it is clear is the importance of credit scoring in today’s society which is based on borrowing and lending, so on credit and the relevance of credit scoring both in research and applications in the private sector. For instance, the Scoring can be applied to both companies and individuals and it is applied in different phases (Amat, Pujadas and Lloret, 2012):
- Customer identification phase. In this phase, financial entities can identify those customers having an appropriate profile to receive the loan (Arimany and Viladecans, 2015).
- Phase of initial study of the operation to decide whether to accept or not (acceptance scoring).
- Once the credit has been granted, there is a phase of post monitoring (Behavioral scoring). During this phase, the scoring is applied to the customers who obtained the loans and it is useful to assess if it is worth to keep the customer or not, if it is better to increase or reduce the limits allowed, to identify too risky customers before it is too late and establish interests and commissions for the renewals.
- Phase of default. In case the customer defaults, the scoring helps to evaluate the level of possible losses and the most appropriate actions to take in order to recover the defaulted payment.
Nowadays, many financial institutions have their own credit scoring and there are even some companies such as Experian and Equifax which fill in reports about their customers using credit scoring results. The history of credit scoring started with the FICO score designed in the USA for the FICO company in 1958. The first credit scoring for credit cards instead started in 1960 with Montgomery Ward. Still today, there is significant dispute among scholars and practitioners about what is the best statistical technique to apply to obtain the best credit scoring model in terms of predicting power. One of the most famous and respected credit scoring model is the Z score of Altman (1968) whose model is still one of the most used by credit institutions. Argenti (1983) proposed a model which aims at determining the probability of insolvency of a company using variables related to management and control. Along this line many others credit scoring models have been developed trying to add value to the previous ones. For instance, Hu and Ansell (2007) analyze for the retailing sector the usefulness of the models for credit risk evaluation. In this way, they compare four classical methodologies (Naïve Bayes, logistic regression, recursive partitioning and artificial neuronal networks) with the Sequential Minimal Optimization (SMO). They used a sample of 195 healthy companies and 51 that went bankrupt between 1994 to 2002. The five methodologies behaved well in predicting bankruptcy, in particular, one year before the event took place; moreover, it contrasts to how it was possible to predict up to five years in advance the bankruptcy with a level of accuracy superior to 78% and how none of these methodologies resulted to be superior in this classification. This agrees with the previous results where it was posed as a sample how bankruptcy prediction models have a predictive capacity of up to five years before a company goes bankrupt and how it was expected that the closer we get to the bankruptcy event the higher the predictive ability, so the values of the ratios deteriorate at a higher intensity (Marín, Antón and Mondragón, 2011), (Amat, Antón and Manini, 2016). The rest of this section will briefly describe some of the more interesting academic works related to credits scoring. For instance, in Spain among the first appearances of credit scoring techniques of automated valuation we can mention Bonilla, Olmeda and Puertas (2003). As briefly mentioned before, many of the most relevant academic contributions in credit scoring research attempt to develop, compare and comment on the statistical approaches to be used in order to obtain models yielding every time higher predictive power. Bardos (1998) develops a credit scoring model based on linear discriminant analysis after describing the tool used by the Central Bank of France to assess credit concessions. Zhou, Lai and Yen (2009) compare the SVM (support vector machines) technique with six traditional methods, concluding that the SVM approach yields better results than the traditional techniques. On the other hand, Shu-Ting, Cheng and Hsieh (2009) confirm the excellent results obtained with SVM, but they affirm that through the CLC (Clustering-launched classification) better results can be achieved. Paleologo, Elisseeff and Antonini (2010) compare the Subagging technique with other traditional techniques concluding that the latter approach yields higher predictive power. A very interesting contribution to the field of credit scoring, is Gutierrez (2008) who discusses the application of non-parametric techniques in credit scoring. He develops a model based on the Probit model and also comments on the dominance of parametric techniques in the field. His opinion is that parametric techniques are the most used because they are easier to apply and interpret. Antón (2007) shows how influential are the different decisions regarding coefficients and variables for the final outcome of a credit scoring model. Ochoa, Galeano and Agudelo (2010) adopt a discriminant analysis approach to develop a credit scoring model which includes not only the statistical analysis of quantitative variables, but also qualitative ones. The work mentioned so far, focuses mostly on traditional techniques including LDA and Probit model. However, throughout the development of the field, some researchers tried to develop models using more complex techniques such as neuronal networks and decision trees. For example, the study of Blanco, Pino-Mejías, Lara and Rayo (2013) uses neuronal networks. They develop several credit scoring models for the microfinance sector using different techniques including the linear discriminant analysis and the logistic regression. They base their study on a sample of 5.500 loan applicants for a Peruvian microfinance institution, concluding that for the microfinance sector, the results coming from neuronal networks yield better results than those obtained applying traditional techniques. Given that the main purpose of this paper is not to carefully review the existing literature in credit scoring, some of the main contributions in the literature have just been briefly described in order to give the reader a sense of the progresses and changes in general the field have been going through. However, in case the reader is interested in deepening his knowledge of the field analyzing the existing credit scoring literature, a more detailed revision of such works can be found in Allen, De Long and Sanders (2008) and more recently in Abdou and Pointon (2011).
3. Methodology
As Amat, Antón and Manini (2016) this study has the aim of identifying a function which discriminates the companies based on their ability of being solvent. Companies with higher probability to meet their debt obligations will be considered solvent whereas companies yielding lower a probability of meeting their debt obligations will be considered insolvent. To do so, this paper will try to answer the following questions:
- Which ratios better discriminate the companies based on their degree of solvency?
- What is the relative importance of these ratios?
To answer these two questions, the authors decided to follow as faithfully as possible the Altman (1968) approach.
4. Data
To perform the analysis, we use data coming from the SABI database focusing on companies operating in the following industries:
- Retail sale in non-specialized store (industry classification 471)
- Retail sale of food, beverages and tobacco in specialized store (industry classification 472)
Information are gathered on balance sheet information needed for building ratios from 1 to 40* and related to the period 2005-2006. The choice of using these 40 ratios (see Table 1) reflects the aim of the authors to follow the Altman approach and to continue the work started by Amat, Antón and Manini (2016).
Table 1. Ratios tested
Source: Amat, Antón and Manini (2016)
The time interval has been chosen taking into account the proximity to the last financial crises, so the effects it should have had on the Z score of the companies, but avoiding the big impact the crisis itself might have generated. Therefore, we consider it an appropriate time period to check for the robustness of our model and at the same time avoiding abnormal distortions which the financial crisis might generate to our analysis. Balance sheet data are deflated to keep track of inflation and cleaned according to standard procedure described in (Kalemli-Ozcan, 2015). The initial dataset comprises 33,397 companies. Data are winsorized each year at the 1st and 99th percentile to avoid extreme observations. Due to lack of information on receivables, COGS, suppliers, purchases and retained earnings, we start with no observation for ratios: 2, 18, 19, 20, 21, 22, 23, 39 and 40. For identifying bankrupt companies, we build a dummy (failed) taking the value one if the company is registered with one of the following status at any point in time in our dataset:
- Bankruptcy
- The company is going through a judicial dissolution process
- The company through liquidation of insolvency proceedings
- The company is immerse in an insolvency proceeding
- The company is going through an ipso jure dissolution process
- Insolvency proceedings
- Suspension of payments
Overall, we end up with 182 bankrupt companies. However, in the analysis will appear only 100 companies because not every firm year entry in the dataset has information in all remained ratios. Therefore, we select a subsample of companies with information for most of the ratios and we will end up with 18 ratios for our final analysis. Basically, the 18 final ratios are those for which we have the most information available.
5. Analysis and Results
As mentioned previously, the sample of bankruptcy companies which will be used in the analysis is of 100 observations. Also, this paper tries to be as faithful as possible to Altman (1968). Hence, data about the bankruptcy companies refer to one year before the bankruptcy. In order to identify the year of bankruptcy, we fix it as the last year for which there are information on total assets, as data on timing of status are sparse and unreliable. Due to the higher number of observation on active companies compared to bankruptcy ones, we can choose those observations that best match the distribution of total assets such as size in the spirit of Altman (1986). We do this by selecting for each bankrupt company an active one with closest size in absolute terms. Therefore, our analysis will be based on 200 different companies with nearly identical size distribution (see Table 2)
Table 2. Summary statistics of log (total assets)
Given these premises, we perform linear discriminant analysis (LDA) to discriminate between active and bankrupt companies. The null hypothesis that they are not jointly statistically significant is rejected at any level of confidence (result not reported)- the related statistic is reported as λ in (Altman, 1968). In Table 3 we report the standardized coefficients from LDA and the p-value from the F test with null hypothesis stating that the average ratio is equal across bankrupt companies and active.
Table 3. Standardied coefficients from LDA and p-values from bilateral t-test on mean
At this stage, we aim at reducing the dimensionality of the problem, so we select only those ratios with p-value close or smaller than 5%. When we have different ratios with the same numerator, we retain the one with highest absolute standardized coefficient. This strategy leaves us with 6 ratios for the final analysis.
- Ratio 1: Current Assets/Current Liabilities
- Ratio 3: Cash/Current liabilities
- Ratio 5: Working Capital/Sales
- Ratio 6: Net Worth/Assets
- Ratio 14: EBITDA/Current liabilities
- Ratio 24: Employment Costs/ Sales
We now perform the LDA using only these 6 ratios. The group centroids are the following (Table 4).
Table 4. Group centroids
Again, the λ statistic implies that the ratios are jointly significant at any level of significance. Notice that the negative centroid for active firms means that an increase in a given ratio implies a reduction of the probability of bankruptcy if the related standardized coefficient is negative and an increase if positive. The final Z function considering standardized coefficients as in Table 3 is the following:
Z= 0.103 * Ratio1 – 0.671 * Ratio3 – 0.079 * Ratio5 – 0.460 * Ratio6 -0.692 * Ratio14
+ 0.273 * Ratio24
A high level of EBITDA/Current Liabilities (Ratio 14) and Cash/Current Liabilities (Ratio 3) are the best predictor of firms that are not bankrupt against those that are. The classification is the following (Table 5).
Hence our algorithm correctly identifies 62% of bankruptcy companies and 71% of active companies. As a further check, we apply the algorithm to bankrupt firms in the sample 2 and 3 years before the bankruptcy. The results are described as follows (Table 6 and Table 7).
Table 5. Classification between active and failed companies using the algorithm
Table 6. Classification of ex-post bankrupt companies 2 years before the bankruptcy
Table 7. Classification of ex-post bankrupt companies 3 years before the bankruptcy
As expected, moving from 3 to 2 years prior to bankruptcy, the percentage of companies correctly identified as bankrupt increases from 45,8% to 66,7%.
The percentage of predicted bankruptcies increases rapidly one year before the bankruptcy itself. We observe a similar trend for the median Z-score, whereas the average Z-score depicts an inconsistent trend, but that is influenced by extreme realizations on the left tail (Table 8).
Table 8. Classification and average and median Z-score of bankrupt companies t years before bankruptcy
Note that sample size reduces when we move backward before bankruptcy, as we lose track of some firms balance sheet data.
6. Conclusions
This paper tries to complete the work started by Amat, Antón and Manini (2016) building a Z score which could help financial institutions assessing the solvency of companies within the supermarket and retailing industry. This is definitely a step further towards the cooperation between research and private sector and the results achieved show that the model is fairly reliable considering some limitations such as industry specificity. In fact this particular model is meant only for the supermarket and retailing industry. The geographical location of the companies analyzed, in fact the sample analyzed belong mostly to the Spanish market. Lastly, the methodology applied is a traditional one which is still the guideline within the credit scoring applications, but it is clear that academic research in credit scoring is trying to go other ways exploiting the growing knowledge on applied statistics and big data. Finally, some limitations can be considered. For example, this study focuses on companies within the Spanish market operating in a very specific industry field. Moreover, the statistical approach of the study follows the structure of both Altman (1968) and Amat, Antón and Manini (2016)
References
REFERENCES
Abdou, H.A., and Pointon, J. (2011). Credit Scoring, statistical techniques and evaluation criteria: A review of the literature. Intelligent System in Accounting, Finance and Management, 18, 59-88. https://doi.org/10.1002/isaf.325
Allen, L., De Long, G., and Saunders, A. (2004). Issues in credit risk modeling of retail markets. Journal of Banking and Finance, 28, 727-752. https://doi.org/10.1016/S0378-4266(03)00197-3
Altman, E. (1968). Financial ratios, discriminant analysis and the prediction of corporate Bankruptcy. Journal of Finance, 23(4)., 589-609. https://doi.org/10.1111/j.1540-6261.1968.tb00843.x
Amat, O., Antón, M., Manini, R. (2016). Credit Concession Through Credit Scoring. Analysis and Application proposal. Intangible Capital, 12(6), 1967-9818. http://dx.doi.org/10.3926/ic.903
Amat, O., Pujadas, P., and Lloret, P. (2012). Análisis de Operaciones de Crédito. Barcelona: Profit Editorial.
Anderson, R. (2007). The Credit Scoring Toolkit: Theory and practice for retail credit risk management and decision automation. New York: Oxford University Press.
Antón, M. (2007). Una propuesta alternativa en la valoración del riesgo de fracaso empresarial mediante la elaboración y aplicación a priori de modelos de predicción de alerta de crisis. Revista de Contabilidad y Tributación CEF, (288), 111-162.
Argenti, J. (1983). Predicting Corporate Failure, Institute of Chartered Accountants in English and Wales. Accountants Digest, 138, 1-25.
Arimany, N., and Viladecans, C. (2015). Analysis of the cash flow statement’s usefulness: An empirical study. European Accounting and Management Review, 1, 75-100. https://doi.org/10.2139/ssrn.2737292
Bardos, M. (1998). Detecting the risk of company failure at Banque de France. Journal of Banking and Finance, 10-11, 1405-1419. https://doi.org/10.1016/S0378-4266(98)00062-4
Blanco, A., Pino-Mejías, R., Lara, J., and Rayo, S. (2013). Credit scoring models for the microfinance industry using neural networks: Evidence from Peru. Expert Systems with Applications, 40, 356-364. https://doi.org/10.1016/j.eswa.2012.07.051
Bonilla, M, Olmeda, I., and Puertas, R. (2003). Modelos paramétricos y no paramétricos en problemas de Credit Scoring. Revista Española de Financiación y Contabilidad, 118(32), 833-869. https://doi.org/10.1080/02102412.2003.10779502
De Young, R., Glennon, D., and Nigro, P. (2008). Borrower-lender distance, credit scoring, and loan performance: Evidence from International-opaque small business borrowers. Journal of Financial Intermediation, 17, 113-143. https://doi.org/10.1016/j.jfi.2007.07.002
Gutiérrez, M.A. (2008). Anatomía de los modelos de credit scoring. Ensayos Económicos BCRA, 50, 61-96.
Hand, D.J., and Jacka, S.D. (1998). Statistics in Finance. London: Arnold.
Hu, Y., and Ansell, J. (2007). Measuring retail company performance using credit scoring techniques. European Journal of Operational Research, 183, 1595-1606. https://doi.org/10.1016/j.ejor.2006.09.101
Kalemli-Ozcan, S., Sorensen, B., Villegas-Sanchez, C., Volosovych, V., Sevecan, Y. (2015). How to Construct Nationally Representative Firm Level data from the ORBIS Global Database NBER Working Papers 21558, National Bureau of Economic Research, Inc.
Luo, S., Cheng, B., and Hsieh, C. (2009). Prediction model building with clustering-launched classification and support vector machines in credit scoring. Expert Systems with Applications, 36, 7562-7566. https://doi.org/10.1016/j.eswa.2008.09.028
Marín, S., Antón, M., and Mondragón, Z. (2011). Crisis bancarias, información financiera y modelos de predicción: Estudio de un caso. GCG: Revista de Globalización, Competitividad y Gobernabilidad, 5, 32-41.
Ochoa, J.C., Galeano, W., and Agudelo, L.G. (2010). Construcción de un modelo de scoring para el otorgamiento de crédito en una entidad financiera. Perfil de Coyuntura Económica, 16, 191-222.
Paleologo, G., Elisseeff, A., and Antonini, G. (2010). Subagging for credit scoring models. European Journal of Operational Research, 201, 490-499. https://doi.org/10.1016/j.ejor.2009.03.008
Schreiner, M. (2002). Ventajas y desventajas del scoring estadístico para las microfinanzas. Microfinance Risk Management, Washington University in St. Louis, 1-40.
Schreiner, M. (2004). Scoring arrears at a Microlender in Bolivia. Journal of Microfinance, 6(2), 65-88.
Zhou, L., Lai, K., and Yen, J. (2009). Credit scoring models with AUC maximization based on weighted SVM. International Journal of Information and Decision Making, 8(4), 677-696. https://doi.org/10.1142/S0219622009003582
APPENDIX
Olympic facilities and its current use
-Football Stadium Camp Nou, which was remodeled for the Olympics, in use by the FC Barcelona, its owner.
-Olympic Field Archery (Barcelona), temporary installation suitable for later use, in use.
-Olympic Shooting Range Mollet, in use by the administration.
-Catalan Olympic Canal, construction expresses, in use by the administration.
-Municipal Tennis Center Vall d’Hebron; in use.
-The Equestrian Club Montanyà, inactive riding club.
-Olympic Stadium in Terrassa, in use.
-National Institute of Physical Education of Catalonia (Barcelona), in use by the administration.
-Lake Banyoles, conditioning.
-Lluís Companys Olympic Stadium, no regular activity, managed by a municipal company.
-Football stadium Nova Creu Alta, conditioning, in use.
-Olympic Arena, in use by the administration.
-Pavilion of the Catalan Countries (Badalona), in use.
-Palau Blaugrana, in use by the FC Barcelona, its owner.
-Sports Palace of Barcelona, in use by the administration but regular activity.
-Granollers Sports Palace, built expressly for use by management.
-Palace of Metallurg, in private use
-Palau Sant Jordi construction expresses, no regular activity.
-Segre Olympic Park, used by management.
-Spain Sports Municipal Industrial, in use by the administration.
-Mar Bella Municipal Sports Complex, in use by the administration.
-Pavilion construction expresses Vall d’Hebron, in use by the administration.
-Montjuic Municipal Pool, in use by the administration.
-Bernat Picornell Pools, in use by the administration.
-North Station Sports, in use by the administration.
-Olympic Port of Barcelona, in use by the administration.
-Horta Velodrome, expansion and remodeling, not in use and in very poor condition.
In total 35 facilities were available to house sports competitions. The city of Barcelona had 19 facilities, divided in four Olympic areas: Montjuïc, Sea Park, Diagonal and Vall d’Hebron. In addition to 15 other cities were chosen as subsites of these Olympics.
Barcelona city
Montjuïc Olympic Area
The so-called Olympic Ring:
- Montjuic Olympic Stadium – Athletics, opening and closing.
Two. Palau Sant Jordi – finals Artistic Gymnastics and volleyball and handball.
Three. Bernat Picornell Pool – swimming, synchronized swimming and water polo final.
April. INEFC – Fight.
The facilities of the Barcelona Fair competitions were:
May. Palacio de los Deportes Barcelona – Rhythmic gymnastics and volleyball.
June. Metallurgy Palace – Fencing and fencing events of modern pentathlon.
And in surrounding areas were found two other facilities:
July. Montjuic pool – Diving and water polo.
August. Spain Pavilion Industrial – Weightlifting.
In addition to a circuit in the Free Zone for athletic gear tests of 20 and 50 km and 10 km feminine male.
Olympic Area Sea Park
Located in the district of San Martin, hosted the Olympic Village, the Olympic port and three installations:
- Olympic Port – Vela.
- Pavilion Mar Bella – Badminton.
- North Station Sports – Table tennis.
Olympic Area Diagonal
Located in the southwestern part of the city, the highest, along Avinguda Diagonal four facilities had already built:
- Camp Nou – Football.
- Palau Blaugrana – judo, taekwondo (demonstration sport) and hockey final (demonstration sport).
- Sarria Stadium – Football.
- Royal Polo Club – Equestrian and modern pentathlon (show jumping competition).
Olympic Area Vall d’Hebron
Located north of the city, in the district of Horta-Guinardó hosted four Olympic sports:
- Horta Velodrome – Track Cycling.
- Olympic Field Archery – Archery.
- Hebron Valley Pavilion – Volleyball Pelota (demonstration sport).
- Municipal Tennis Center Valle Hebron – Tennis.
Other Locations
- Badalona (Palacio Municipal de Deportes de Badalona) – Basketball.
Two. Badalona (Catalan Countries Pavilion) – Boxing.
Three. Granollers (Granollers Sports Palace) – Handball.
April. Banyoles (Lake Banyoles) – Remo.
May. Mollet del Vallés (Olympic Shooting Range Mollet) – Shooting Men.
August. San Sadurní de Noya (circuit rider) – Road Cycling – Hockey skates.
- Hospitalet de Llobregat (Llobregat Baseball Stadium) – Baseball.
- Vilamoura (Vilamoura Baseball Stadium – Baseball.
- Castelldefels (Castelldefels Olympic Channel) – Canoeing in calm water.
- Seo de Urgel (Segre Olympic Park) – Canoeing (slalom).
- Seva (stables The Montanyà) – Equestrian (dressage and eventing).
- Montmelo (Circuit de Catalunya) – Cycling (100 km team time trial).
- Zaragoza (Estadio La Romareda) – Football.
- Valencia (Luis Casanova Stadium) – Football.
- Sabadell (Stadium Nova Creu Alta) – Football.
- Terrassa (Olympic Stadium) – Field Hockey.
