White Book of the European Exhibition Industry - Vol. 1

Annex 3:
Description of the methodology involved in the quantitative survey

Here we describe the method used for the quantitative section of the study undertaken by BIPE Conseil and London Economics for MEDIA Salles.

The information which came from a survey undertaken in the first half of 1993, of a sample of exhibitors from different countries in the EU, has been included in the White Book. The aim of this quantitative work was to gather information complementary to the dossier of statistics already published by MEDIA Salles (the European Cinema Yearbook 1992 and 1993 editions) by providing original data on the location of cinemas, their size, commercial policy, facilities, programming and performance.

A questionnaire, created by BIPE, then submitted to the representatives of MEDIA Salles and subsequently vetted by the different national exhibitors' associations, was translated into eight national languages, and sent out, in the post, to a sample of 3,000 cinemas situated in the different countries within the Community.

This sample was amassed at random from a database of lists of cinemas sent by the national associations or by official organisations, always taking care, according to the stratified sample approach, to represent in each country the structure of the industry according to the size of cinemas (ie the number of screens). In certain countries, stratified sampling was also used to represent the distribution of the industry according to different regions.

When it was necessary, telephone follow-ups were used by the study's national correspondents, in order to maximise the response rate to the survey. 1,170 questionnaires were completed and returned, which is a response rate of nearly 40%, which is more than respectable for a survey of this type; however, there is no precedent for a quantitative survey concerning the industry throughout all the countries of Community. This response rate certainly attests to the interest manifested in such a study as much by the exhibitors' associations as by the management of the cinemas themselves.

The data capture and the statistical treatment of the questionnaires was entrusted to a specialist firm, under the direction of BIPE Conseil.

All the precautions taken and the existence of a very favourable response rate however does not necessarily avoid all risk of bias.

As for all surveys which require a voluntary collaboration by the people or companies questioned, it is possible that the pattern of non-responses alters somewhat the representativeness of the data obtained. When looked at country by country, the usable data showed, for example, an over-representation of single-screen establishments - this was particularly the case in Germany - and, in some cases, an under-representation of these cinemas and an over-representation of multi-screen complexes.

Whenever it was possible, these biases were corrected by using the classical technique of weighted coefficients, in order to adjust the sample in each country within the structure of the national universe.

The data capture and statistical processing of the questionnaires, extremely complex as numerous variables were involved, were entrusted to a specialist survey data processing company, under BIPE Conseil’s supervision.

Two series of results were provided: unweighted and weighted, in order to adjust the national samples to reflect the distribution of screens within the national universe (using the variable: number of screens per cinema).

We must state that the distribution of responses required from the correspondents were calculated on a the basis of distribution of screens by category of cinema; the rate of weighting used was calculated on the basis of the number of screens by category of screen, and not by the number of cinemas.

The weighted results, using the distribution of screens by category of cinema (Single-screen, 2 screens, 3 screens etc.) were the ones we mostly used in the analysis.

This type of bias was actually easy to detect, since precise country-by-country data on the variations in cinema distribution according to size of theatres was already available. It is not, however, certain that other biases were introduced for variables for which exhaustive information was not available. It seems, for example, that the response rate has generally been better from the large urban areas than from the smaller sized population areas. In the absence of reliable general data, however, it is impossible to redress this bias in a satisfactory manner.

Weighting Coefficients
No. of Screens	Belgium			Denmark			France
No. of Screens	Pop %	Sample %	Coef	Pop %	Sample %	Coef	Pop %	Sample %	Coef
1	22.0	22.2	0.991	34.5	27.9	1.236	32.0	22.9	1.397
2	10.7	6.3	1.698	21.5	11.7	1.837	13.5	12.4	1.088
3-5	36.4	34.9	10.42	36.27	37.8	0.959	36.0	42.5	0.847
6-7	8.8	11.1	0.793	4.0	5.4	0.741	13.5	13.7	0.985
8+	21.6	25.4	0.850	2.5	17.1	0.146	5.0	8.5	0.588
No. of Questionnaires	63			111			153
No. of Screens	Germany			Greece			Ireland
No. of Screens	Pop %	Sample%	Coef	Pop %	Sample %	Coef	Pop %	Sample %	Coef
1	34.5	55.0	0.627	99.0	99.0	-	17.2	10.8	1.592
2	19.5	16.3	1.196	1.0	1.0	-	25.8	15.4	1.675
3-5	35.0	24.0	1.458				36.5	49.2	0.742
6-7	7.0	3.1	2.25				3.2	9.2	0.347
8+	4.0	1.6	2.5				17.2	15.4	1.117
No. of Questionnaires	129			94			65
No. of Screens	Italy			Netherlands			Portugal
No. of Screens	Pop %	Sample %	Coef	Pop %	Sample %	Coef	Pop %	Sample %	Coef
1	96.0	90.1	1.065	14.9	11.3	1.318	76.0	55.3	1.374
2	2.0	7.2	0.277	21.0	23.9	0.878	9.5	14.5	0.655
3-5	1.5	1.8	0.833	59.3	60.6	0.978	8.5	17.1	0.497
6-7	0.0	0.0	0.0	4.6	4.2	1.095	3.0	-	-
8+	0.5	0.9	0.555	0.0	0.0	0.0	3.0	13.2	0.454
No. of Questionnaires	111			71			76
No. of Screens	Spain			UK
No. of Screens	Pop %	Sample %	Coef	Pop %	Sample %	Coef
1	62.0	56.5	1.097	18.0	28.8	0.625
2	7.0	8.2	0.853	12.3	11.4	1.080
3-5	22.9	22.4	1.022	32.7	30.9	1.058
6-7	4.0	8.2	0.487	11.3	18.1	0.621
8+	3.5	4.8	0.729	24.7	14.8	1.669
No of Questionnaires	147			149

For various reasons (the cooperation of associations, the commitment of exhibitors, the attitude of the major circuits, experience of this type of survey, the "culture" of enterprises etc.) the response rate and even the number of questionnaires returned varies considerably from country to country.

Numbers of observations per country
	Number of Questionnaires	Percentage of Questionnaires	No. of screens	%
Belgium	63	5.4%	431	14.6
Denmark	111	9.5%	315	35.2
France	153	13.1%	4402	3.4
Germany	129	11.0%	3630	3.5
Greece	94	8.0%	405	3.4
Ireland	65	5.6%	189	34.3
Italy	111	9.5%	3020	3.6
Netherlands	71	6.1%	416	17.0
Portugal	76	6.5%	232	32.7
Spain	147	12.6%	1807	8.1
UK	149	12.7%	1757	8.4
Total	1,020	100.0%	16,604	6.1

At the European level, the overall structure of the sample does not therefore reflect either the relative importance of each country, or the number of screens, or admissions. For this reason, we recognised it was appropriate to work with unweighted averages. The principal European results were therefore calculated on the basis of arithmetical averages, which gave the same weight to each of the European countries.

This method has the advantage of providing a representation of the corresponding reality in the "average" situation in Europe. But it also proves inconvenient: it is very difficult to apply to cross tabulate (between different variables). These latter results are therefore shown as weighted averages in the section of the synthesis report (weighted by the number of inhabitants in each country considered). The reader must not therefore be surprised by non-negligible differences which can exist between one and the other.

Despite the favourable response rate, the number of responses received and effectively usable was less than we had hoped for. All the exhibitors who had agreed to take part in the survey were not in fact in a position to answer all the questions in the document which we sent them. In addition, some exhibitors wished to keep certain information confidential.

Moreover, the Spanish association wished to simplify all the documents sent to exhibitors in that country; the questions appearing in the data which were more "sensitive" (turnover, distribution of turnover etc.) were therefore removed.

For certain questions, therefore, the replies received were relatively few in number. The results must therefore be interpreted with caution. And above all, it is very difficult in this situation to try and show clear relationships between different variables, because these split the sample into too many sub-groups which are too small in size. The variable "country", which, at the European level, appears with a dozen items, cannot be crossed with any other variable, at the risk of weakening considerably the representativeness of the results; this explains why the tables and figures which follow do not present the results country by country, but in terms of the European average.

Finally, and undoubtedly most importantly, a fourth risk of bias comes from the way in which exhibitors responded to the different questions which they were asked. One cannot exclude the possibility that some of them saw good reason to provide responses which only imperfectly matched reality: this applies particularly, for example, to the issues relating to comfort or performance of theatres.

Amongst other things, the mention in the questionnaire and in the accompanying letter from MEDIA Salles, will obviously have given the exhibitors questioned an incentive to minimise the proportion of their programming and their total revenues relating to US films. The regulatory intervention control has also played a part in much the same way, in all cases in countries where regulation exists relating to programming.

Lastly, the risk of errors must not be underestimated, due to insufficient control by the exhibitors themselves of the data provided for their own cinema.

Despite the many precautions taken, the risks of bias are far from minimised. It would therefore be erroneous, or in some cases hasty, to extrapolate from these results to the whole of the industry. Whatever its deficiencies, the information gathered does provide orders of magnitude in different areas, where, until now, information has altogether been lacking, and allows us to position the variables which were studied in relationship to each other.