- 19% of the firms have been engaged in manufacturing for the whole period in question or started doing that (mostly in fashion sector).

The set of explanatory variables was determined by either already common or only gaining popularity strategic practices that retailers can potentially employ in order to increase their efficiency. In compliance with the previous chapter these are: organic and inorganic growth, franchising techniques, multichannel selling, and private labels.



Fig. 2. Data by retail sector

The aggregated statistics for independent variables is as follows:

- 92% of the firms organically expanded their market presence throughout the period in question. However, the average growth in the number of outlets has been constantly declining since 2013.

- 18% of the firms acted as bidders in M&A during the period in question.

- 57% of the firms have been strategically employing multichannel selling for the whole period in question or started doing that.

- 52% of the firms have been strategically developing private labels for the whole period in question or started doing that.

- 30% of the firms have been strategically employing franchising schemes for the whole period in question or started doing that.

Thus, the majority of large-sized retailers in Russia prefer organic growth as the safest strategy towards higher returns, while M&A are still not in great request. Meanwhile, one can easily notice the growing interest in additional strategic tools such as multichannel selling, private labels and franchising that help increase sales with minimum costs. However, almost half of market leaders still never tried these tempting opportunities, which leaves room for investigation.

To avoid multicollinearity within further regression analysis, it is important to examine pairwise correlations between independent variables (Appendix 2). The highest correlation (0.42) is between production and prlabel_share, since these are usually producer-retailers with almost 100% of private labels in sales. The results indicated no suspiciously high correlation coefficients (neither exceeds 0.8), so any combination of these variables can be included into the model without any concern about possible multicollinearity.

3.2 Model Specification

Appendix 3 provides the results of a series of panel estimates with EBITDA margin as a dependent variable. According to Breusch-Pagan test, the random-effects regression models always appeared to be better than the pooled ones. Hence, the results on the latter were suppressed to save space. The aim of conducting these four models was to determine the best model specification, i.e. by trying different variables and their modifications.

We started with constructing the model with only control variables included (M1). All the controls, except for the age and leverage variables, take expected coefficients that are both statistically and economically significant. Not only leverage is almost insignificant (10%), but it also turned out to have no actual impact on margin (=0). In this regard, the decision was made to exclude the leverage variable from the model. As for the age variable, its unexpectedly negative coefficient is insignificant at any reasonable significance level. Deduction of the `Detsky Mir' retailer as an outlier according to age (around 60 years) did not help. Having plotted the graph indicating the relation between age and margin, the slight U-shape relationship has been noticed. This means one can try adding the polynomial values of the age variable into the regression model.

The next model (M2) is to mostly check whether the polynomial values of age make it any more significant. As a result, the age variable is now significant at 1% level, which testifies for the U-shape relationship between the firm's age and its efficiency. Practically it means that the firm is most efficient either in its earliest years or in the latest, with lower efficiency in the middle. Now all the controls are finally settled.

In the third model (M3), we expand the specification by additionally including industry controls and the whole set of independent variables: growth of outlets, the fact of participating in M&A, private labels in sales, online sales in total sales, and the fact of using franchising schemes. This did not bring any significant changes to the significance levels and signs of the controls. Out of all the independent variables, only online_share appeared significant at 1% level, while others are either significant at 10% or insignificant. Hence, we needed to perform some modifications in order to achieve higher explanatory power.

The final model (M4) includes specifications for some of the independent variables. The gr_outlets was most significant when paring it with the region variable, i.e. including the interaction term between these two variables. The coefficients indicate that opening new outlets is most reasonable when doing it in a new geographical market area. In addition, lagged values of M&A appeared in the model to see how the effect changes years after the merge. Though, the variable itself still remains insignificant, its lagged values are significant at 1% level. However, the test for the three of them being jointly equal to 0 indicated these should stay in the model. In this model prlabel_share remained a significant determinant. We did not manage to come up with a worthy specification for franchising, so it was removed from the future models.

Thus, from all the specifications we consider the forth model as the best. The overall significance of the model can be proven by relatively high value of Wald statistics (180.64). In the next paragraph the same model specification but with fixed-effects will be analysed. Then, the Hausman test is to help us conclude on which model better fits the data.

3.3 Random-effects VS Fixed-effects Model

Appendix 4 indicates the random-effects model chosen in the previous paragraph and two fixed-effect models: with (M6) and without (M5) time effects. The core distinction between to types of models is that unlike fixed-effects model, the random one goes under the assumption of no correlation between unique errors and regressors. However, this is not always the case, so for some data fixed effects may provide more adequate results.

In our particular case, the results obtained through the fixed-effects approach turned out to be less coherent in terms of both signs and significance of the coefficients. While the M&A variable as well as its derivatives became less significant, the variables gr_outlets, region, and their interaction term are all insignificant now. The coefficients for both production and online_share variables changed their sign to negative, which contradicts common sense. Though the coefficients produced by fixed-effect models are consistent for regressions with no endogeneity, part of them can still be ineffective. This can make the coefficients on the major regressors insignificant. Thus, we consider the model as not a good fit for our data. As for the fixed-effects model with time dummies, the test for all time dummies being jointly equal to 0 indicated that these are jointly insignificant (p-value=0.345). This means that the coefficients are time invariant.

In order to finally decide between random- and fixed-effects models, we are to conduct a Hausman test where the null hypothesis is that unique errors do not correlate with regressors. The results of the test (p-value=0.603) thus indicated that the random-fixed model is more preferable for our dataset.

3.4 Interpretation of the Regression Coefficients

The Model 4, our random-effects regression model, was considered the most preferable according to the previous paragraphs. High value of Wald statistics (p-value=0.000) speaks for the overall significance of the model. Altogether, controls and independent variables explain roughly 48% of the variance in margins. The franchising variable turned out to be insignificant in any of model specifications. Probably, the reason is its rare application by Russian retailers due to the potential problems arising from limited managerial control and imperfect legislation.

The control variables are all significant at 1% level, have expected signs and account for approximately 15% of R-squared. As we predicted, the company's size and own production facilities positively affect retailer's efficiency, while the debt load - negatively. Interestingly enough, modification of the age variable testifies its U-shape relationship with our dependent variable. This means that having been functioning for a number of years, companies accumulate stock of experience in the particular industry, create solid customer and supplier bases, and acquire new technologies that result in increasing overall efficiency.

The organic growth as a strategy is represented by annual growth in the number of outlets. The corresponding gr_outlets variable only appeared significant when being included with the region variable and their interaction term. It means that opening new outlets becomes more efficient when entering a new geographic market area: each additional 1% growth in the number of outlets opened in a new region brings two times bigger margin than if there was no regional expansion. In 2015, despite the recession, market leaders showed the record 30% growth in floor space since 2009. The phenomenon is due to the large-sized retailers trying to take advantage of the weakened market with its decreasing competition for floor space and low rental rates. Another trend is geographic expansion to remote market areas of Russia, characterized by low penetration of modern retail formats. Both of these tendencies are reflected in our coefficients on the group of `organic' variables.

The extensive growth is represented by a dummy variable indicating whether the firm participated as a bidder in M&A in the year of observation or not. Though the MA variable is insignificant, its both lagged values are significant at 1%. The interpretation of the coefficients is as follows: initiating M&A is to pay off only two years after the merge (2.2% margin). The margin will be characterized by 6% decline in efficiency in the next year and 6.6% increase - in the second. The results seem to be realistic and do not contradict to common sense. Given economic crisis, Russian retailers are reluctant to participate in M&A. However, we consider it as having potential due to low retail market concentration in Russia (10 largest retail players account for only 24% as of 2015), even though real benefits can only be achieved at least two years after the merge.

As for the private labels, these have a positive effect on retailer's efficiency: each additional 1% of private labels in sales increases margin by 3%. It is therefore no surprise that in recent years retailers started to introduce their private label lines. The positive impact can be explained from two perspectives: reduction in costs and sales stimulation. The first results from closer working relationships with suppliers, thus providing better contractual terms and lower input prices. This in turn creates better value for money that is so appreciated by consumers during recession times.

Private labels can be thus considered as an additional dimension on which retailers can compete for customers by innovating and changing their own-brand offerings.

The multichannel selling strategy is the last to be addressed. The coefficient on the corresponding online_share variable can be interpreted as follows: each 1% increase in the share of online sales will bring 1.3% margin. The positive relation can be explained by multichannel selling being a low-cost way of building trustful relationships with the customers, increasing market coverage and providing around-the clock hours of operation. According to Russian consumers, the key strengths of online shopping include time saving (47%) and the opportunity to choose the best price (45%). The recession is thus considered to be the perfect time for implementing the strategy.

Finally, the Shapley Value Decomposition technique helped us to determine the independent variable of highest positive impact on our dependent variable. The results indicated that the M&A variable is of highest contribution (12%), followed by organic growth with its 11%. Then, private labels and multichannel selling taken separately explain 9% and 5% of the margin variance, respectively. Overall, one can conclude that mergers and acquisitions conducted in retail during the last 6 years all had a positive impact on the efficiency of the bidder. Given Russian retail is characterised by low level of concentration, we would recommend taking action even though the first synergy is to be obtained no earlier than two years later.

CONCLUSION

retail market business strategy

The current research examines the Russian retail market and is to provide its leaders with the information on their drivers of efficiency. The topic is relevant due to the recession of 2015 caused by the combination of sharp drop in oil prices and continuing Western sanctions imposed on Russia due to its policy in Ukraine. With inflation high, the real consumer spending power has fallen, thus having an adverse effect on the consumer market. In addition, the business investment has also fallen sharply due to high capital cost and limited profitable opportunities.

The recession thus becomes a real test for the market players on how they can better adapt to the changes in the marketplace. Though Russian large-sized retailers felt financially steady to the macroeconomic challenges, the major problem they face now is fierce competition for customers and high pressure on prices. The paper thus proposes 5 perspective dimensions of future development and chooses the most marginal one.

The literature review indicated a substantial gap in the field from two perspectives. First, to the best of our knowledge, such a research has never been conducted for the Russian retail market. Second, most of the foreign case studies on the topic estimate the efficiency of a single retail chain rather than the whole market. Moreover, the existing studies usually consider the effect of only one driver of efficiency, which is in most cases not retail specific. In order to fill those gaps our research is to consider 5 business strategies towards higher margins, best reflecting the specifics of retail and most relevant to the current economic situation in Russia.

The practical part of the proposed research is based on the panel regression analysis with EBITDA margin as a proxy for retailers' efficiency and dependent variable. The dataset contains financial and operational information on 90 large-sized Russian retailers over a 6-year period, from 2010 to 2015. As independent variables business strategies of highest potential are used: organic and extensive growth, multichannel selling, franchising, and private label, each represented by a group of corresponding variables. After running several panel regression models, the one with random effects appeared as the best fit for our data. Having finally conducted the Shapley Value Decomposition technique, we managed to reveal the contribution of each variable to the explanation of Russian retailers' efficiency.

The major findings indicated all positive relationships between the business strategies and EBITDA margin. It was revealed that the strategy of organic growth represented as a growth in the number of outlets becomes two times more effective when entering new geographic market areas. The result is especially relevant given the vast size of Russia and hence enough room for geographic expansion. With regard to the extensive growth, it was revealed that M&A is to pay off at least two years after the merge, following a sharp drop in efficiency in the first year. Though retailers seem to be reluctant to participate in M&A now, the strategy still has the potential due to low concentration of the Russian retail market. The results for the private labels and multichannel strategies appeared to significantly affect retailers' margin, while franchising was removed from the model for being insignificant. The reasons may be lack of managerial control, non-compliance between the parties, and imperfect franchising legislation in Russia.

According to the Shapley Value Decomposition results, the business strategies in ascending order according to their contribution to the explanation of retailer's efficiency are as follows: M&A, organic growth, private labels, and multichannel selling. As organic growth appeared to be of almost the same effect on margin as M&A deals, we did not manage to provide statistical evidence to the heated discussion of which of the two growth types is more efficient. The effects appear comparable, though, only when organic strategy and regional expansion are pursued together. Hence, for those retailers with already high territorial presence we would recommend to better focus on finding worthy M&A opportunities, while smaller retailers may consider opening new outlets in other regions of Russia to accelerate margin growth. It would also be of benefit for Russian retailers to continue finding new ways to improve their own brand ranges as the current needs of the retail environment and consumers will continue to evolve. For those retailers with no financial or operational opportunity to launch private label lines we would recommend to pay closer attention to the strategy of multichannel selling, though it not as efficient as other business strategies coved in the paper. The results obtained might have important managerial implications for the Russian retailers aspiring to enhance their margins after the 2014-2015 recession.

REFERENCES

1. Ahmad, Z., Abdullah, N. M. H., Roslan, S. (2012). Capital structure effect on firms' performance: Focusing on consumers and industrials sectors on Malaysian firms. International Review of Business Research Papers, 8(5), 137-155.

2. Aigner, D., Lovell, C. K., Schmidt, P. (1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics, 6(1), 21-37.

3. Akinboade, O.A., Mokwena, M.P., Kinfack, E.C. (2008). Estimating profit effciency among South African retail firms using stochastic frontier approach. The international review of retail, distribution and consumer research,18(5), 511-520.

4. Bahri, M., St-Pierre, J., Sakka, O. (2011). Economic value added: a useful tool for SME performance management. International Journal of Productivity and Performance Management, 60(6), 603-621.

5. Barros, C.P. (2005). Efficiency in hypermarket retailing: a stochastic frontier model. The international review of retail, distribution and consumer research, 15(2), 171-189.

6. Barros, C.P., Alves, C.A. (2003). Hypermarket retail store efficiency in Portugal. International journal of retail and distribution management, 31(11), 549-560.

7. Cameron, A., Trivedi, P. (2009). Microeconometrics using Stata. Texas: Stata Press.

8. Charnes, A., Cooper, W.W., Rhodes, E. (1978). Measuring the efficiency of decision making units. European journal of operational research, 2(6), 429-444.

9. De Jorge, J. (2008). Productivity growth of European retailers: A benchmarking approach. Journal of Economic Studies, 37(3), 288-313.

10. Dobbs, R., Koller T. (2005). Measuring long-term performance. The McKinsey Quarterly, 4, 16-27.

11. Dong-mei, W., Peng, W. (2012). A study on similar financial models in manufacturing sector. Journal of Modern Accounting and Auditing, 8(1), 77-86.

12. Gaur, V., Fisher, M., & Raman, A. (1999). What explains superior retail performance? (Working Paper). The Wharton School, University of Pennsylvania.

13. Gleason, K. C., Mathur, L. K., Mathur, I. (2000). The interrelationship between culture, capital structure, and performance: evidence from European retailers. Journal of Business Research, 50(2), 185-191.

14. Global Agricultural Information Network. Russian Federation retail foods. Retrieved from Global Agricultural Information Network official website: http://agriexchange.apeda.gov.in/marketreport/Reports/Retail%20Foods_Moscow%20ATO_Russian%20Federation_12-9-2015.pdf

15. Goldman, A. (1992). Evaluating the performance of the Japanese distribution system. Journal of Retailing, 68(1), 11-50.

16. Good, W. S. (1984). Productivity in the retail grocery trade. Journal of Retailing, 60(3), 81-97.

17. Gupta, A., Mittal, S. (2010). Measuring retail productivity of food and grocery retail outlets using the DEA technique. Journal of strategic marketing, 18(4), 277-289.

18. Kaya, H. (2007). Are retailers more sensitive to changes in business conditions compared to wholesalers? International Journal of Business and Social Science, 5(10), 1-8.

19. Lu, W., Hung, S. (2011). Exploring the efficiency and effectiveness in global e-retailing companies. Computers and operations research, 38, 1351-1360.

20. Lusch, R. F., Moon, S. Y. (1984). An exploratory analysis of the correlates of labor productivity in retailing. Journal of Retailing, 60(3), 37-61.

21. Matrosova, Y. (2016). Estimation of the drivers of the Russian retail market efficiency. Higher School of Economics in Saint Petersburg.

22. Meeusen, W., Van den Broeck, J. (1977). Technical efficiency and dimension of the firm: Some results on the use of frontier production functions. Empirical Economics, 2(2), 109-122.

23. Nooteboom, B. (1982). A new theory of retailing costs. European Economic Review, 17(2), 163-186.

24. Oxera Consulting. The economic benefits of retailer own-brands. Retrieved from Centromarca Marcas Originais official website: http://www.centromarca.pt/folder/conteudo/632_Oxera%20report%20on%20retailer%20own%20brands.pdf

25. Ratchford, B. T., Stoops, G. T. (1988). A model and measurement approach for studying retail productivity. Journal of Retailing, 64(3), 241-263.

26. Reynolds, J., Cuthbertson, C., Bell, R. (2004). Retail strategy: The view from the bridge. Burlington: Elsevier Butterworth-Heinemann.

27. Russian Information & Business Center, Global Investment Center. (2015). Russia: investment and business guide. Washington: International Business Publications, Inc.

28. Patel, G.N., Pande, S. (2013). Measuring the performance of pharmacy stores using discretionary and non-discretionary variables. OPSEARCH, 50(1), 25-41.

29. PwC. Total Retail 2015: Retailers and the age of disruption. Retrieved from PwC official website: http://www.pwc.ie/media-centre/assets/publications/ 2015-pwc-ireland-total-retail-february.pdf

30. Sarantopoulos, P., Kioses L., Doukidis G. (2007). Productivity and efficiency in grocery retail. IELKA's White Paper Series. The institute of Retail Consumer Goods.

31. Selden, A., Gipson R., Parker A. (2008). An Introduction to franchising. Minneapolis: Minnesota Department of Employment and Economic Development.

32. Sellers-Rubio, R., Mas-Ruiz, F. (2008). Technical efficiency in the retail food industry. European Journal of Marketing, 5/6(43), 652-669.

33. Sellers-Rubio, R., Mas-Ruiz, F. (2006). Economic efficiency in supermarkets: evidences in Spain. International Journal of Retail & Distribution Management, 34(2), 155-171.

34. Sellers-Rubio, R., Mas-Ruiz, F. (2007). Different approaches to the evaluation of performance in retailing. International Review of Retail, Distribution and Consumer Research, 17(5), 503-522.

35. Shockley, J., Turner, T. (2014). Linking productivity and efficiency to retail firm performance: empirical insights from US retailing segments. Production planning and control, 26(5), 393-406.

36. Turner, T., Shockley, J. (2014). Creating shopper value: Co-creation roles, in-store self-service technology use, and value differentiation. Journal of Promotion Management, 20(3), 311-327.

37. Waldorf, W. H. (1966). Labor productivity in food wholesaling and retailing, 1929-1958. The Review of Economics and Statistics, 48(1), 88-93.

38. Wet, J. H. (2005). EVA versus traditional accounting measures of performance as drivers of shareholder value - A comparative analysis. Meditari Accountancy Research, 13(2), 1-16.

39. Xavier, J.M., Moutinho, V.F., Moreira, A.C. (2015). An empirical examination of performance in the clothing retailing industry: A case study. Journal of retailing and consumer services, 25, 96-105.

40. Zhao, Y., Trejo-Pech, C. O., Weldon, R. N. (2013). The profitability of the US food supply chain: Financial indicators, cross-section and time-series effects. Journal of Applied Business Research, 29(1), 23-34.

APPENDIX

Definition of variables and descriptive statistics (N=90)

Group of variables	Variable	Definition	Type	Mean	SD
Dependent	ros	Return on sales (EBITDA to sales), %	Quant	1.69	7.34
Control	age	Time span between the year of observation and the year of establishment of the firm	Quant	14.7	7.2
	ln_assets	Logarithm of assets	Quant	14.59	1.83
	production	Whether the firm is a producer-retailer (1) or not (0)	Binary	0.18	0.38
	liquidity	Current ratio (current assets to current liabilities)	Quant	13.6	138.44
	debt	Debt ratio (total debt to total assets)	Quant	0.83	0.28
	listed	Whether the firm is listed(1) or not (0)	Binary	0.07	0.26
Organic growth	country	The number of countries the firm operates in	Quant	1.36	1.23
	region	Whether the firm entered a new region (1) or not (0)	Binary	0.21	0.41
	city	The number of cities the firm operates in	Quant	104.92	267.54
	gr_outlet	Growth in the number of outlets, %	Quant	23	0.32
	gr_square	Growth of selling area, %	Quant	34	0.36
Inorganic growth	MA	Whether a firm has acted as a bidder in M&A (1) or not (0) in this particular year	Binary	0.08	0.27
Franchising	franchising	Whether the firm uses franchising (1) or not (0)	Binary	0.23	0.42
Private labels	prlabel_share	Percentage share of private labels in sales	Quant	12.56	26.9
Multichannel selling	online_share	Percentage share of online sales in total sales (the variable is set equal to 100 for online retailers)	Quant	15.82	33.55
Sector effects	sector	8 sector dummies (FMCG, electronics and home appliances, DIY, household, drogery, fashion, children's goods, art)	Binary
Time effects	year	6 year dummies (2010-2015)	Binary

Matrix of pairwise correlations between independent variables

	age	ln_assets	production	debt	liquidity	gr_outlet	MA	franchising	online_share	prlable_share
age	1
ln_assets	0.293*	1
production	0.041*	-0.145*	1
debt	0.025*	0.161*	0.017*	1
liquidity	0.065*	0.076*	-0.040*	-0.042*	1
gr_outlet	-0.145*	-0.074*	0.109*	0.027*	-0.025*	1
MA	0.108*	0.199*	-0.066*	0.202*	-0.024*	0.057*	1
franchising	0.076*	0.162*	0.111*	-0.054*	0.150*	-0.065*	0.035*	1
online_share	-0.193*	-0.244*	-0.154*	-0.085*	-0.022*	0.046*	0.062*	-0.038*	1
prlabel_share	-0.002*	0.060*	0.422*	0.030*	-0.041*	0.048*	-0.058*	0.347*	-0.172*	1

Deciding between different model specifications

___________________________________________________________

Dependent variable EBITDA margin

Method: random-effects panel

_____________________________________________________________

Explanatory variables Model1 Model2 Model3 Model4

_____________________________________________________________

Age -0.035 -1.014*** -1.683*** -2.553***

(0.09) (0.37) (0.52) (0.62)

Age^2 0.033*** 0.061*** 0.089***

(0.01) (0.02) (0.02)

Ln(firm size) 0.759** 0.937*** 1.126** 0.986**

(0.34) (0.35) (0.47) (0.49)

Debt to assets -6.307*** -6.338*** -6.172*** -5.814***

(0.42) (0.42) (0.46) (0.57)

Liquidity 0.000*

(0.00)

Production 4.204*** 4.491*** 5.113*** 4.217*

(1.58) (1.59) (1.88) (2.19)

Growth of outlets 1.940* 1.662**

(1.17) (0.67)

New region 1.298***

(0.48)

Outlets*Region 0.597***

(0.15)

M&A -1.969 1.596

(1.61) (2.08)

L1.M&A -5.959***

(2.25)

L2.M&A 6.596***

(2.40)

Multichannel selling 1.109*** 1.302***

(0.29) (0.42)

Private labels 2.003** 2.908*

(0.87) (1.55)

Franchising -0.334

(1.49)

Constant -6.517 -2.940 -4.183 6.959

(4.69) (4.84) (7.11) (7.92)

Industry dummies No No Yes Yes

Observations 534 528 528 528

R-sqr (within) 0.387 0.396 0.512 0.494

___________________________________________________________

Notes: The table shows estimated coefficients. Robust standard errors are in parentheses. ***Statistically significant at the 1% level; **at the 5% level; *at the 10% level. Model1 and Model2 are regression models with only control variables included. Model3 and Model4 include independent variables, as well.

Deciding between random- and fixed-effects model

_________________________________________

Dependent variable EBITDA margin

Method: random- and fixed-effects panel

_____________________________________________________________

Explanatory variables Model4 Model5 Model6

(re) (fe) (fe)___________

Age -2.553*** -2.921*** -2.869***

(0.62) (1.01) (1.02)

Age^2 0.089*** 0.082*** 0.079***

(0.02) (0.03) (0.03)

Ln(firm size) 0.986** 4.255*** 4.434***

(0.49) (1.40) (1.41)

Debt to assets -5.814*** -6.643*** -6.735***

(0.57) (0.62) (0.62)

Production 4.217* -2.502 -2.167

(2.19) (6.55) (6.56)

Growth of outlets 1.662** -0.422 -0.142

(0.67) (0.40) (0.13)

New region 1.298*** 0.233 1.331

(0.48) (0.39) (1.72)

Outlets*Region 0.597*** -1.916 -1.033

(0.15) (1.18) (1.75)

M&A 1.596 0.375 0.491

(2.08) (2.21) (2.22)

L1.M&A -5.959*** -6.621*** -6.758***

(2.25) (2.22) (2.23)

L2.M&A 6.596*** 2.937 3.278

(2.40) (2.75) (2.75)

Multichannel selling 1.302*** -0.051 -0.050

(0.42) (0.15) (0.10)

Private labels 2.908* 0.148 0.149

(1.55) (0.16) (0.14)

Constant 6.959* -35.572* -38.264**

(7.92) (18.50) (18.71)

_________________________________________________

Industry dummies Yes No No

Time dummies No Yes Yes

Observations 528 528 528

R-sqr (within) 0.49 0.53 0.53___________

Hausman p-value 0.603

Testparm p-value 0.345__________

Notes: The table shows estimated coefficients. Robust standard errors are in parentheses. ***Statistically significant at the 1% level; **at the 5% level; *at the 10% level. Model3 is a random-effects regression model, Model4 - fixed-effects. Model5 is a fixed-effects regression model with time dummies.

Размещено на Allbest.ru

...