Opening Hours and Employment in the Retail Sector: Quasi

Home , Ladenschlussgesetz, Shopping hours

Opening Hours and Employment in the Retail Sector:

Quasi-Experimental Evidence from Germany

Bossler, Mario * Oberfichtner, Michael † Institute for Employment Research University of Erlangen-Nuremberg

Preliminary, August 2013.

Abstract We study the effect of a deregulation in shop opening hours on employment in food retailing. Our identification relies on changes in restrictions of business hours across German states that allow for a difference-in-differences approach. Using data on the universe of German establishments, we find that lifting restrictions on business hours increases total employment modestly. This effect is driven by an increase in part-time employment while full-time employment is not affected. The statistical significance of these effects hinges on the assumptions on error correlation and we hence report inference robust to clustering at different levels.

* Institute for Employment Research, Department Establishments and Employment, Regensburger Str. 104, 90478 Nuremberg, e-mail: [email protected], phone: +49-911-179-3043. † Friedrich-Alexander University Erlangen-Nuremberg, Department of Economics, Lange Gasse 20, 90403 Nuremberg, e-mail: [email protected], phone: +49-911-5302-771. For helpful comments and suggestions, we thank Sascha Becker, Thiess Büttner, Boris Hirsch, Robin Naylor, Thorsten Schank, Claus Schnabel, Till von Wachter, Fabian Waldinger, Matthias Wrede, Jeffrey Wooldridge, and Thomas Zwick as well as participants of the 16th BGPE workshop and seminars at the universities of Erlangen-Nuremberg and Warwick. Michael Oberfichtner visited University of Warwick during the work on this article, and he is thankful for their hospitality.

Introduction

Despite a general trend towards deregulation, restrictions of shop opening hours are still widespread. Restrictions often take the form of blue laws and restrict business activities on

Sundays. However, some regulations also prohibit trade for some hours during weekdays. For instance, in a number of European countries like Austria, Belgium and Finland such restrictions are in place as they also were in Italy until 2012 (Metro 2012, 64–65).

Changes in the regulation of shop opening hours typically cause heavy public debates. A recent example is Italy, where Monti’s reforms in 2012 intended to foster the economy included a deregulation of opening hours. Interestingly, the discussion of labour market effects is very vivid with proponents claiming that a deregulation will create jobs in retailing while opponents claim that it costs jobs and harms small retailers (see for instance Financial

Times 2012). Similarly, a major trade union in Germany claims that deregulating business hours would create part-time jobs at the expense of full-time jobs (verdi 2006).

In this paper, we provide quasi-experimental evidence on the effects of deregulating daily opening hours on total employment as well as part-time and full-time employment. Our analysis focus on the total effect on the food retail sector, taking into account changes in market entries and exits. Our identification strategy relies on differences in the regulation of opening hours across German states (“Bundesländer”) after a major reform of federalism in

2006. While regulations were identical prior to the reform, states could implement their own law afterwards, which all states but Bavaria did. This creates exogenous variation in the restrictions faced by retailers.

Comparing shops in Bavaria to shops in other states, we implement a difference-in- differences approach. Unlike in many other settings, where one group is treated and one has to come up with a suitable control group, we need to construct a treatment group that matches

2 our control group. As a first step, we restrict our attention to West Germany since the parallel trends are more plausible within this group. Furthermore, the sample is limited to states with similar employment trends before the reform. To ensure that our control group is not affected by the deregulation, e.g. via cross-border shopping, we restrict our attention to food shops and perform additional robustness checks. Given the construction of our sample, the estimates should only be interpreted as an average treatment effect on the treated for the respective states and sector.

Using the 100% sample of the IAB Establishment History Panel, which comprises information on all establishments in Germany with at least on employee, we find a small positive effect on employment in food retailing. This effect is driven by an increase in part- time employment, while full-time employment is not affected. Looking at the effects on shops already existing before the reform, we find that the deregulation of opening hours most likely had a negative impact on employment in small shops whereas employment in larger shops benefited. The statistical significance of our results pends on the assumptions one is willing to make regarding the error distribution and we therefore report results using different types of inference.

Previous economic research neither offers clear theoretical predictions nor empirical evidence regarding the employment consequences of deregulating opening hours during weekdays.

From a theoretical point of view, Gradus (1996) argues that the total employment effect of longer opening hours at the plant level is unclear, since it is a combination of several effects.

Longer opening hours could increase sales and hence the derived labour demand. Another positive effect on employment could stem from the minimum number of workers necessary to keep a shop open. In contrast, if there are congestion costs that decrease labour productivity, the smoothing of sales over more hours could decrease labour demand. However, all of these arguments apply to a single shop, though changes at the market level, for instance in terms of

3 shop closure and changes in the market structure as modelled by Wenzel (2010; 2011), could also affect labour markets. On the empirical side, there is some quasi-experimental evidence supporting a positive employment effect of lifting Sunday opening restrictions (see Burda /

Weil 2005; Goos 2005; Skuterud 2005). However, empirical research has little to offer regarding restrictions of opening hours during day time.

The paper proceeds in section 2 with a description of the institutional setting of the regulation of shop opening hours. Section 3 presents the dataset and sample construction for the analysis.

Section 4 starts with a descriptive analysis, while section 5 presents the sophisticated regression estimates.

Institutional setting

The 2006 reform of German federalism provides the framework for the analysis. In Germany, the federal states have legislative power on any topic unless the constitution ( Grundgesetz ) states otherwise. In these cases, there are different forms of federal competence, ranging from areas, where the federation has exclusive competence to fields where competences are shared in different ways. The reform of federalism aimed to ease legislation procedures by clarifying federal and state competencies as well as the modes of cooperation. At the same time, it had to keep a balance of power between the federation and the states. To achieve this, legislative power for some areas has been transferred from the federal states to the state and vice versa.

The draft of the reform bill was discussed for the first time in both chambers of parliament in

March 2006. This draft was based on the work of a multi-party commission headed by leading politicians of the two major parties, and it was already informally approved by major political actors: the parliamentary groups of the governing coalition, the conference of the prime ministers of the federal states and the federal cabinet. It was formally approved by both chambers in June and July 2006. It finally came into force on 1 September 2006. Given the

4 broad support for the first draft, it is not surprising that the final bill was very similar to the draft and the reform could hence be anticipated already in early 2006.

The reform moved legislation on shop opening hours from the exclusive federal domain to the state domain. Before the reform, a federal law (Ladenschlussgesetz ) prohibited shops to open on Sundays and holidays as well as on Monday to Saturday before 6am and after 8pm; with few exceptions such as for shops in train stations and airports. After the reform of federalism, this federal law stayed in place unless a state issued a law of its own.

Between 15 November 2006 and 1 April 2007, new regulations of shop opening hours came into force in all West German states but Bavaria. Table 1 gives an overview of the regulations applying Monday to Saturday after April 2007. In seven out of these nine states, all restrictions on opening hours from Monday to Saturday were lifted. One state allowed for longer opening hours in the evening, until 10 pm, and one state, the Saarland, adopted a law with only minor changes in comparison to the federal law. The restrictions of opening hours were hence changed for shops in eight states in a very similar way. Restrictions for Sunday openings were by and large unaffected and are therefore not further discussed.3

While there is no systematic evidence on changes of opening hours after deregulation, the previous restrictions appear to have been binding for many shops. Kaapke et al. (2007) report the results of a survey in several cities in early 2007. Shoppers were asked whether the shops where they typically buy some product are open after 8pm. Depending on the type of shop, between 9 per cent (stores for leather goods) and 47 per cent (food stores) said that this was the case, with 37 and 21 per cent not knowing. As a whole, these numbers show that the previous restrictions on opening hours were a binding constraint for a substantial portion of the retail industry and that the reform affected shops’ behaviour.

3 Regarding East Germany the timing and the content of the new laws were very similar. 5

Looking at the two states without major changes, Bavarian shops appear to be better suited as control group for two reasons. First, there were some minor changes in the Saarland, for instance shops are allowed to open for 24 hours once per year (the exact date can vary across municipality and years). Hence, including the Saarland in the control group would lead to a control group that includes slightly treated plants as well as non-treated plants. Furthermore,

Bavaria did not deregulate more or less accidentally. The parliamentary group of the governing party voted on lifting restrictions on shop opening hours and because of a tie in votes decided not to do so. What is more, one supporter of the deregulation could not vote due to another appointment. Since the Saarland purposely denied a deregulation and also considering the similarity to the treatment group, Bavaria is more suitable to define the control group of establishments.

Data

In our empirical analysis, we use the 100 per cent sample of the Establishment History Panel

(BHP) of the Institute for Employment Research (for an extended overview of the dataset see

Spengler 2008; a detailed description of the latest version can be found in Gruhl et al. 2012).

The dataset is based on information from the social insurance system, where every establishment is obligated to report information on all employees who are subject to the social security system. The individual information as of 30 June is then aggregated at the establishment level in each year. This procedure yields a panel dataset that comprises all establishments with at least one employee.

The plant level characteristics in the data set include the total number of employees, the main outcome variable for our analysis, as well as information on the numbers of part-time and full-time employees. Furthermore, the data set includes information on plant location such that each establishment can be assigned to a federal state, regional district and municipality.

This is of particular importance since the estimation and identification of the present analysis

6 relies on a variation in the deregulation of opening hours on the state level. The data set also includes a 5-digit industry identifier. Since the industry classification changes in 2003 and

2009, we restrict our analysis to the years 2003 to 2008.

As our analysis is on the establishment level, let us briefly describe the exact meaning of this term in our dataset. An establishment is defined as a worksite of a firm in a certain municipality, which can either be the whole firm or a branch. If a firm has more than one worksite, here stores, in the same municipality, it may use the same establishment number for all notifications from these plants or it can use a different number for each site (and it may change this decision). Hence, an establishment can comprise information on a firm, all branches of a chain in one municipality or a single branch. Each establishment has a unique identifier. This identifier is normally time constant and changes only in case of certain events, like a change of ownership, the legal form or in certain cases of store’s location. Thus, the identifier allows us to follow establishments, whatever the term describes exactly for a given identifier, over time.

Construction of the Analysis sample

To make the parallel trends assumption and the unaffectedness of the control group plausible, we confine our analysis in terms of industry as well as in terms of location. Based on the industry identifier, we restrict the sample to food retailing for three reasons. Firstly, food retailers reacted to the deregulation by extending opening hours. The survey mentioned above already points towards a relatively large impact of the deregulation and this is supported by some further evidence. 3,000 out of 5,700 stores of a major German food retailer, REWE, opened after 8pm by early 2012 (Welt 2012), which is fairly similar to regionally limited numbers for the same retail chain from spring 2007 (FAZ 2007). Hence, the restriction on opening hours was binding for a substantial portion of food retailers making plausible that lifting the constraint affected shop behaviour. Secondly, food retailers are least likely to be

7 affected by cross-border shopping caused by the change in opening hours such that these effects should be very limited. While shoppers might be willing to drive further to get longer opening hours for more event-like shopping, e.g. a new TV, this appears to be less relevant for day-to-day groceries. Finally, food retailers are least affected by the rise of online retailing. It is plausible that the effects of online retailing depend on regional characteristics, like income and population density, which would violate the common trends assumption.

Focusing on food retailers hence strengthens both the unaffectedness as well as the common trends assumption.

Regarding shop location, we restrict the sample to West Germany. The parallel trends assumption in the absence of treatment is more plausible for West German states since East

Germany is still affected by lower economic development, differing industrial structure, high unemployment rates, substantial migration to the West, and finally East German states are still heavily subsidies. All of these issues point towards large differences between East and West

German states making the parallel trends assumption less plausible than for the more homogeneous group of West German states.

Finally, the FIFA World Cup took place between 9 June and 9 July 2006 in cities all over

Germany making 2006 unsuitable as the last pre-treatment period. During the World Cup, some of the federal states relaxed the restriction on opening hours, while others allowed municipalities to decide on the opening hours for that period. On the political side, the willingness to relax the restriction temporarily is likely to correlate with the willingness to relax it permanently. Furthermore, the reform of federalism has already been discussed in parliament by that time and thus, the deregulation could already be anticipated in June 2006.

On the firms’ side, shops anticipating that a change will become permanent soon could be more likely to react to a temporary change. Thus, employment in treatment and control states

8 could be affected differentially in 2006 and we exclude this year from the analysis sample and use 2005 as the last pre-treatment period.4

To obtain the overall effect of the deregulation on employment in retail establishments, we use all plants which are active in at least one year between 2003 and 2008. If a plant is not active in at least one of these years, employment is defined to be zero. Doing this, we obtain a balanced panel. This treatment of missing observations allows following all establishments irrespectively of market entries and exits, and it also ensures that changing establishment identifiers do not affect our estimates. Using this artificially balanced panel, the estimated treatment effect therefore presents the effect accounting for the change in employment which is due to establishment entries and exits and can thus be interpreted as an overall average effect on employment of the treatments states’ plants, where the average is over all potentially active establishments.

Panel A of Figure 1 yields a graphical illustration of the analysis when using Bavaria and

Saarland as the control group and the eight other West German states as the treatment group.

The lines present the average employment per potentially active plant by treatment status over the period 2003 to 2008 centred at the averages in 2005. Before the treatment, employment rose faster in the treated states than in the control states, making a difference-in-differences analysis for these groups implausible. We therefore take a closer look at the single states to find comparable treatment and control groups.

Panel B shows the evolution of employment until 2005 by state. Two features stand out in this graph. In all states but Bremen and Saarland shops experienced increasing employment between 2003 and 2005. Furthermore, employment decreased between 2004 and 2005 in shops in all states but North Rhine-Westphalia and Schleswig-Holstein. Shops in the remaining six states have very similar employment trends, making a difference-in-difference

4 In fact, we observe a small placebo effect in 2006, which does not change the major regression results of the analysis. 9 analysis feasible. Hence, shops in these states appear to be well-suited as treatment and control group and we focus on these shops throughout our analysis.

Graphical and Descriptive Evidence

Panel A of Figure 2 redoes Panel A of Figure 1 using only shops in Bavaria, Baden-

Württemberg, Hamburg, Hesse, Lower Saxony, and Rhineland-Palatinate. Looking at the pre- treatment period, employment trends are indeed very similar in both groups making the parallel trends assumption in the absence of the treatment plausible. Looking at 2007, the first post-treatment period, we observe that employment increased more in deregulated shops than in Bavarian shops. This difference amounts to about 0.4 employees per plant, which remains stable in 2008. This first result indicates a modest positive effect of the deregulation on employment.

Since the period used so far is rather short, we have also extended our sample over both reclassifications of the industry identifiers. Panel B replicates the graphical analysis over for the years 1999 to 2010. As before, we see parallel movements in employment until 2005 and an increase in employment in treated shops directly after the deregulation that remains stable until the end of the observed period. Hence, using this extended period does not promise additional insights, but adds noise to the analysis. We thus only look at the years 2003 to 2008 in the remaining paper.

Figure 3 shows the development of part-time employment (Panel A) and full-time employment (Panel B) by treatment status. For both types of employment pre-treatment trends are very similar suggesting parallel trends in the absence of treatment. In the treatment period, we see an increase in part-time employment, but no differences in full-time employment.

These patterns suggest that the effect on total employment is driven by the change in part-time employment.

To give some information on our sample, Table 2 presents descriptive statistics by treatment status for the years 2005 and 2007 focusing on total employment. The sample consists of 10

11,487 Bavarian establishments, forming the control group, and 12,133 treated plants in the other states. By construction of our sample, the number of establishments does not change over time.

Turning to average employment before the deregulation, Bavarian establishments had 10.98 employees and were thus slightly smaller than establishments in the treatment group, which had 11.38 employees. Looking at the employment changes between 2005 and 2007 shows that average employment increased by 0.42 employees in Bavarian between 2005 and 2007. For the treated plants, we observe an increase by 0.82 employees. These numbers imply a first descriptive difference-in-difference estimate of the treatment effect of 0.4 employees per establishment.

Regression Analysis

In the regression analysis the treatment effect of interest is obtained from a standard

Differences-in-Differences specification:

= + + + (1)

The dependent variable in this baseline specification is the total number of employees in plant i at time t. Deregulated is the treatment dummy of interest, which is 1 for plants in treatment states during the treatment period. The baseline specification also includes year dummies as well as plant fixed effects.

Turning to statistical inference, we use two different approaches relying on different error assumptions. Firstly, we allow for arbitrary error correlation within districts, but not across districts. Since plants are located across 248 districts, we report Huber-White cluster robust standard errors. The number for districts is large enough to calculate meaningful asymptotic standard errors (Cameron, et al. 2008). While the assumption on the error distribution is rather restrictive, this approach should be reasonably powerful to detect moderate effects. Secondly,

11 we relax the assumption of independent errors across districts and instead cluster at the state level, and thus on the level of the identifying variation. Since this leaves us with only six clusters, we cannot rely on analytical clustered standard errors and bootstrap p-values using the wild cluster bootstrap procedure imposing the null hypothesis as proposed by Cameron et al. (2008). The second approach is robust to correlated state-level shocks and appears to provide valid p-values even with as few as six clusters. However, it is very conservative

(probably more conservative than necessary) and has little power to detect effects of moderate size (see the simulation results of Cameron et al. 2008). Given these strengths and weaknesses of both approaches, we report results from both and readers may decide which result to follow.

The first column of Table 3 reports the results from the baseline specification. We find an increase of 0.42 employees after deregulating opening hours. Compared to the average establishment size of about 11 workers this effects appears of reasonable size. Using standard errors clustered at the district level, we get a t-statistic of about 2.5 and would hence reject he null of no effect at the 5% level. Clustering at the state-level gives a p-value of 0.16 and we would thus not reject the null at conservative levels.

Columns 2 and 3 of Table 3 show results from robustness checks. Firstly, we add state specific linear and quadratic time trends to our regression. Secondly, add the logged disposable income and logged population at the district level as control variables. In both specifications, we get almost identical point estimates for the treatment effect of 0.38 and 0.41 supporting our identification strategy. Also, the results regarding statistical significance are very similar, though the p-value when clustering at the state level falls to 0.05 when including the control variables.

Finally, we do a standard placebo test and estimate a placebo effect. Using pre-treatment observations only and assigning the treatment status to plants in treatment states in the last 12 pre-treatment period. This estimation tests whether an anticipation of the treatment and therefore a reaction before the actual treatment period is observed. Furthermore, it is a regression based test of a deviation in the parallel trends assumption in the last pre-treatment period. The placebo effect is practically zero, supporting the parallel trends assumption.

In a second step we disentangle the average employment effect into an effect on part-time and full-time employment. Table 4 presents the treatment effect for part-time and full-time employment. As suggested by the graphical analysis we find an increase in part-time employment by 0.35 workers, but practically no change in full-time employment. The effect on part-time employment is statistically significant when clustering at the district level as well as when clustering at the state level with a p-value of 0.05. These results are confirmed when doing the same robustness checks as before and again we do not find a placebo effect.

Robustness checks

As discussed above the treatment group consists of plants in states which are selected from the population of German states. The selection is entirely driven by the trends in average employment per plant in the absence of treatment, i.e. before the deregulation took place. To check robustness against a different selection of treatment states and rule out that our results are driven by shops in a single state, we redo our baseline estimation dropping shops from each treatment state at a time. Table 5 presents the results showing a treatment effect close to

0.4 irrespective of the excluded state. Point estimate range from 0.36 to 0.51 and the results in terms of statistical significance are again mixed.

Next, we address cross-border shopping as a channel through which the treatment could also affect shops in the control group and the estimated effect could then be driven by redistributing employment across states. As argued above the extent of cross-border shopping should be rather limited in food retailing. To further reduce it, we drop shops from all districts

13 which flank the border between a treatment state and Bavaria. Column 1 of Table 6 shows that this does not alter our results.

Finally, we test whether the 2004 enlargement of the European Union (EU) violates the parallel trends assumption. The Czech Republic joined the EU in 2004 and Bavaria is the only state in our sample neighbouring the Czech Republic. Hence, shops close to border could be affected by the expansion while shops in the treatment group where not. We hence drop all shops in districts located at the border to the Czech Republic from the sample. As can be seen from Column 2 of Table 6 this does not alter our results, which also hold when combining both restrictions as shown in the final column.

Conclusion

We analyze the effect of the deregulation in opening hours on employment in the German food retail sector and find a moderate increase in average employment. It is the first comprehensive analysis of employment effects induced by a deregulation of opening hours on weekdays. A quasi-random deregulation of opening hours across German states in 2006 is studied which allows using a difference-in-differences method. We have selected treatment states on the basis of parallel trends before treatment and thus claim to uncover the causal effect of interest on the deregulated plants.

The effect of 0.4 employees per plant represents an increase in employment by 3.2% in the food retail sector of treated states or equivalently a raise in employment by about 8,000 employees. This estimate is robust to adding state specific time trends and controlling for covariates. In further analyses we show that the effect is mostly driven by an increase in part- time employment, whereas full-time employment does not react to the deregulation.

Additional robustness checks show that the treatment effect is not driven by cross border shopping or the 2004 EU expansion. Furthermore, the results are robust to excluding single

14 treatment states and we thus claim that the selection of treatment states does not explain the employment increase.

Table 1: Opening hours regulations by state

Federal state Regular potential opening hours State law came into force in in treatment period Baden-Württemberg 0-24 March 2007 Bavaria 6-20 No state law Bremen 0-24 April 2007 Hamburg 0-24 January 2007 Hesse 0-24 December 2006 Lower Saxony 0-24 April 2007 North Rhine-Westphalia * 0-24 November 2006 Rhineland-Palatinate 6-22 November 2006 Saarland ** 6-20 November 2006 Schleswig-Holstein 0-24 December 2006 *) A stricter regulation came into force in May 2013, several years after our observation period ends. **) The state law in Saarland implemented some rather minor changes compared to the old federal law like allowing for longer opening hours once per year at the municipality level. Source: Federal law and various state laws.

Table 2: Descriptive statistics

Establishments Number of Average number of employees Difference Difference-in- located in Establishments Differences 2005 2007 Bavaria 11,487 10.976 11.397 0.421 Treatment 12,133 11.376 12.193 0.817 0.396 states Treatment states are Baden-Württemberg, Hamburg, Hesse, Lower Saxony, and Rhineland- Palatinate. Source: Establishment History Panel.

Table 3: Deregulation effect on total employment Baseline State Specific Control variables Placebo trends Treatment effect 0.415 0.382 0.414 0.060 (0.165) (0.175) (0.175) (0.083) [0.16] [0.17] [0.05] [0.20] Observations 148,020 88,816 Note: Dependent variable is the number of employees at an establishment. Standard errors clustered at the district level in parentheses. P-Values using a wild cluster bootstrap (state level clusters) with 1,000 iterations in brackets. All regressions include plant as well as year fixed effects. Column (2) includes linear and quadratic state-specific trends, column (3) includes log(disposable income) and log(population) at the district level as controls. Analysis is based on artificially balanced panel, 2003 to 2008. In column (4), only data until 2005 is used and 2005 is coded as treatment year. Source: Establishment History Panel.

Table 4: Treatment effects on part-time and full-time employment Dependent Baseline State Specific Control variables Placebo variable trends Number of part- 0.346 0.282 0.344 0.030 time employees (0.089) (0.117) (0.092) (0.052) [0.05] [0.27] [0.03] [0.71] Number of full- 0.073 0.102 0.071 0.002 time employees (0.010) (0.089) (0.108) (0.042) [0.29] [0.12] [0.35] [0.81] Observations 148,020 88,816 Note: Dependent variables are the number of part-time employees at an establishment and the number of full-time employees. See further notes from Table 3.

Table 5: Regression results excluding each treatment state at a time Excluded state Hesse Baden- Lower Saxony Hamburg Rhineland- Württemberg Palatinate Treatment effect 0.363 0.422 0.505 0.396 0.405 (0.158) (0.187) (0.181) (0.166) (0.171) [0.11] [0.14] [0.07] [0.12] [0.19] Observations 126,554 115,321 116,223 143,383 134,580 Note: Dependent variable is the number of employees at an establishment. See further notes to baseline regression from Table 3.

Table 6: Regression results with exclusion of border regions Excluded border regions Bavarian - Treatment Bavarian - Czech Both Treatment effect 0.390 0.456 0.433 (0.252) (0.230) (0.266) [0.11] [0.07] [0.12] Observations 128,751 144,471 125,207 Note: Dependent variable is the number of employees at an establishment. See further notes to baseline regression from Table 3.

Figure 1: Average employment per establishment in West-Germany

Notes: The graphs show the average numbers of employees per plant centred at the values of

2005.

Figure 2: Average employment per establishment by treatment status, analysis sample only

Notes: The graphs show the average numbers of employees per plant centred at the values of

2005. Treatment states are Baden-Württemberg, Hamburg, Hesse, Lower Saxony, and

Rhineland-Palatinate.

Figure 3: Average full-time and part-time employment per establishment by treatment status, analysis sample only

Notes: The graphs show the average numbers of part-time and full-time employees per plant centred at the values of 2005. Treatment states are Baden-Württemberg, Hamburg, Hesse,

Lower Saxony, and Rhineland-Palatinate.

References

Burda, M. and Weil, P. (2005). Blue laws. Working paper.

Cameron, A. C., Gelbach, J. B., and Miller, D. L. (2008). Bootstrap-based improvements for inference with clustered errors. The Review of Economics and Statistics , 90(3):414-427.

FAZ (2007). Vorbild Rewe. Frankfurter Allgemeine Zeitung, March 1, 2007.

Financial Times (2012). Italian lobbies apply brakes to Monti’s reforms. January 2, 2012. Online available at http://www.ft.com/intl/cms/s/0/fc36edea-3554-11e1-84b9- 00144feabdc0.html#axzz2dKhcu2UX.

Goos, M. (2005). Sinking the blues: The impact of shop closing hours on labor and product markets. Working paper.

Gradus, R. (1996). The Economic Effects of Extending Shop Opening Hours. Journal of Economics , 64(3): 247-263.

Gruhl, A., Schmucker, A. and Seth, S. (2012). The Establishment History Panel 1975-2010. FDZ-Datenreport 04/2012. Nürnberg.

Kaapke, A., Lindstädt, B., Preißner, M. (2007). Nutzung und Auswirkungen liberalisierter Ladenöffnungszeiten im Einzelhandel, Handel im Fokus , 59(3): 156-171.

Metro (2012). Metro Retail Compendium 2012/13. Düsseldorf.

Skuterud, M. (2005). The impact of Sunday shopping on employment and hours of work in the retail industry: Evidence from Canada. European Economic Review , 49(8):1953-1978.

Spengler, A. (2008). The establishment history panel. Journal of Applied Social Science Studies , 128(3), 501-509. verdi (2006). Erfahrungsbericht zu den verlängerten Öffnungszeiten für Verkaufsstellen des Einzelhandels während der Fußballweltmeisterschaft 2006. Berlin.

Wenzel, T. (2010). Liberalization of Opening Hours with Free Entry. German Economic Review , 11: 511-526.

Wenzel, T. (2011). Deregulation of Shopping Hours: The Impact on Independent Retailers and Chain Stores. Scandinavian Journal of Economics , 113(1): 145-166.

Welt (2012). Rewe-Chef wehrt sich gegen früheren Ladenschluss. January 23, 2012. http://www.welt.de/wirtschaft/article13829123/Rewe-Chef-wehrt-sich-gegen-frueheren- Ladenschluss.html.