Choosing a School

Choosing a School

Stephen Gorard - Extract From Phd Thesis

Introduction

Having established, in Chapters 8 and 9, the characteristics of the respondents, and the kind of schools that they use, this chapter moves on to describe further results from the questionnaire survey, and the analysis necessarily becomes more complex. The first section, reviews how families go about making a choice of a new school in terms of who is involved in the choice, where they obtain information from, and how they go about it. The second section looks at the respondents' ratings of the importance of the 73 reasons for choosing a school, dividing these reasons into four groups of decreasing relevance from very important to not important overall. The third section looks at the correlations between these ratings to establish which of them are significantly covariant. In the fourth section, the ratings are reduced, by Principal Components Analysis, to seven underlying factors, which are used to explain the patterned variance in the majority of the reasons for choice. Differences in the relative importance of these to different families, and links between the factors and other reasons for choice are discussed, as well the relationship of the factors, the patterns in the process of choice and the users of different schools.

In general, this chapter follows the same order as Chapter 5, in which reference is made to previous work in this field, and to which the present chapter frequently refers. As with the family background variables in Chapter 9, some values are converted to percentages for presentation. In the interests of clarity, not all of the results mentioned are included as tables within this chapter. All differences and relationships mentioned are significant at the 5% level, using Chi squared, Pearson's r, t-test, or Analysis of variance, as appropriate to the data.

The Process of Choice

Who Chooses The School?

Over 50% of respondents claim that the choice of a school is a joint one, where neither the parents nor the child can be said to have the larger role, while fewer than 10% of respondents claim that the main role is taken by the child (Figure 1). This last is an unexpectedly low figure, given the findings of much previous research as described in Chapter 5, which tend to envisage the choice as being either the prerogative of the child, or at least, a joint process with the parents (e.g. Thomas and Dennison 1991). It should be noted that the question asked 'who has the major say in choosing', and that therefore both parties could have been involved in all decisions to some extent. However, the scale of the difference between this finding and previous studies requires further explanation. The figure is more unexpected, as this survey is unusual in including the views of children, who are less likely to neglect their own role in comparison to their parents. If anything, the inclusion of children would be expected to lead to a greater emphasis on a child-centred approach. It could be that the difference is due to the occupational class, and parents' educational background, since there are some indications that it is especially working-class families who empower their children (Ball et al. 1992). It could be that, in previous work, parents have exaggerated the role that they gave to their children and that what has been discovered here is that children are more realistic, and know how little control they have had. It is also possible that asking families at different stages within the choice process would lead to different answers. Each of these possibilities is briefly discussed in this section, and then more fully in Chapter 14, which describes the process of choice as a three step procedure, in which the role of parent and child varies over time, as well as between families. However, some studies agree with this one, in assigning a much larger role to parents. West et al. (1994) reported that the choice was made by one or more parents 73% of the time, with the child alone only making the choice 7% of the time.

Figure 1: The Main Role in Choosing a School

It is clear from Table 1, in which the frequencies of Figure 1 are broken down by the generation of the respondent, that it is only children who are claiming a major choice role for themselves. It is not clear at this stage, which of the two generations, if any, is more reliable in this matter. However it is clear that the role of children has not been exaggerated by the parents interviewed in previous studies, in fact rather the reverse. Each generation feels that it is more involved in choosing a school than the other one reports, with parents claiming to be significantly involved in 98% of families.

Table 1: The Main Role in Choosing by Generation

	Parent's choice	Child's choice	Joint choice
Parents response	45	2	53
Child's response	36	14	49
Chi-Square	DF	Significance
56.47193	2	.00000

It could be argued that as the study focuses on fee-paying schools, with a greater proportion of service-class families in which parents may have a greater say, possibly because of the payment of fees, the finding with respect to choice roles is as expected, and as found by other researchers, such as Edwards et al. (1989). However, there is no hint of a difference in the allocation of roles to generations when the responses are cross-tabulated by occupational class (Table 2). The uniformity of choice roles across the three main occupational classes is impressive [the lower figures for the pupils role in Table 2 is because the table only represents the views of adults, since children were not asked about their parents' occupation]. This uniformity could be regional, and specific to South Wales, or due to the exact phrasing of this difficult question, but there is no clear reason to accept either of these explanations. The lack of variation by class, and the differences in the reports of each generation indirectly support the third possibility given above for the difference between the findings of this study and others, in that the answer to the question of who really chooses may be sensitive to variations in the time of asking. This finding does not deny that there is a difference by class, since the number of reportedly working-class families in the sample was small, but it suggests that the difference may not be as great as originally supposed.

Table 2: The Main Role in choosing by Occupational Class

	Parent chose	Pupil chose	Joint choice
Service	39	1	60
Intermediate	40	1	59
Working	43	7	50
Other	67	0	33
Chi-Square	DF	Significance
8.19210	6	.22436

The role of the child increases almost directly with age, while the role of the parent remains fairly steady in comparison (Table 3). It is the concept of a shared choice that decreases with age. Older children are more likely to claim that they made a choice alone, rather than with the help of a parent. Of course, these snapshot figures cannot reveal how the individual child's role varies over time, as discussed in Chapter 14, however there is a clear indication of at least one time-sensitive change. The majority of respondents in this survey relate to Years 6 and 7. In most of the schools in which Year 6 were surveyed, the majority of the families were intending to use one of the schools in which the Year 7 were also surveyed. If the survey had been carried out a year earlier, most of the Year 7 respondents would have been involved, and if a year later, most of the Year 6 would still have been involved. Thus, apart from the calendar difference of one year, there is no reason to suppose that the two year groups are any different, as they effectively represent different cohorts from the same schools. The differences between them can therefore be used to shed light on how the reasons of families change during the process. Since there is no reason a priori why the Years 6 and 7 should differ in this respect, it is interesting to note that the majority of respondents, 60%, feel that the choice is a joint one before it has been made, while the majority, 54%, see it as made by one party or the other, after the event.

Table 3: The Main Role in Choosing by Year Group

	Parent's choice	Child's choice	Joint choice
Year 1	44	0	56
Year 6	33	7	60
Year 7	45	9	45
Year 8	24	30	46
Year 11	37	20	42
Chi-Square	DF	Significance
65.16957	8	.00000

An alternative pattern appears in a breakdown of the figures for the choice roles in terms of the types of schools (Table 4). Each of the state primary schools involved fed directly into one of the two secondary schools, so that a large proportion of the users of these schools would be common from year to year, but these are the schools in which the largest variation is apparent. In fact, the pattern for the state secondary schools is similar to those of the fee-paying schools, who are all relatively uniform in terms of who, in the family, plays the main role in choosing. It is the state primaries that display a clear difference to the rest. There is no reason for the roles to vary so much from Year 6 to Year 7 in state schools, since the process of choice that the two groups are describing is, of course, the same one, with the difference being merely one of perspective. The parents are seen to have a much greater role in choice before it occurs, while, in retrospect, the choice is more commonly seen as a joint one, in these schools alone. The overall picture in Table 3, of changing over time from joint choice to parents or children choosing alone, and the finding in state schools, of changing from parents alone to joint choice, have this in common - the role of the child apparently increases over time in some families.

Table 4: The Main Role in Choosing by School Type

	Parent	Child	Both
State primary	71	12	18
Stand-alone proprietary	45	10	45
Catholic	43	8	50
State secondary	42	5	53
Grass-roots	38	8	54
Feeder proprietary	38	9	53
Traditional foundation	38	13	48
Chi-Square	DF	Significance
24.23257	12	.01891

It is well-known locally, by the schools, and is apparent from the interviews with parents, that one of the state secondary schools is highly desirable, and heavily over-subscribed. One of the focus primary schools is a feeder for it, and is assured that there will be a place for all of its leavers. Parents plan ahead, and use the appropriate primary school accordingly (Chapter 13). This may involve moving to the area, or sending children to the primary from outside the catchment of the secondary school. Such moves and applications take place while the child is of primary age, and cannot be expected to play a large role in the process. In summary, many of the choices represented in rows one and four of Table 4, are long-prepared, and planned by parents alone. The exaggerated role of the child in Year 7 is therefore anomalous. This confirms two ideas which are central to the models of choice used in this thesis. Firstly, the apparent role of the child increases as the choice process takes place, and secondly, if there is a dispute between the parents and children, there is some evidence here that the parents are more reliable on this question. The issue of who is really making the choice of schools is a complex one which can be elucidated by the stories from interviews, and is therefore taken up again in Chapter 14.

The differences between religions have already been linked to class, parent's education, and the use of particular types of schools. It is clear that both Catholics, and families with no religion, are more parent-centred in their choosing, while Protestants, and families with a minority religion more often make a joint selection of school. The users of Catholic schools overall are, however, not noticeably different in this respect from any other fee-paying schools.

Sources of Information about Schools

For most respondents, the most significant source of information used in making a choice about schools is a personal acquaintance, whether another parent, or a child at the school (Figure 2). This kind of second-hand information is even more popular than visits to the school. Prospectuses and examination results are used, and acknowledged by many parents, but few feel that they are really useful, despite the publication of league tables and the pressure to use them as performance indicators. These findings are generally in line with those of other studies outlined in Chapter 5 (e.g. GPDSA 1995), while they have implications for the likely outcome of any attempt to improve the quality of schooling via increased choice based upon formal performance indicators. The sources of information listed as 'other' in Figure 2 include a few not coded as one of the other four, such as visits to exhibitions, and the Holy Spirit appearing in a vision, but this category mainly represents families who had relied on more than one main source of information.

Figure 2: The Main Source of Information

The overall figures mask certain differences between the information sources used by various families when considering diverse schools. In addition, each family may use different approaches at different times for different children. It is clear from Table 5 that visits and prospectuses are far more important to the process of choosing a boarding school, while acquaintances and examination results are more relevant for day schools. Day pupils represent 94.3% of the sample, meaning either that they were day pupils, if in Year 1 and 7 at the time of the survey, or they intended to be day pupils next year. The very few boarders arises partly from the lower response rate from boarding schools, but it is also a reflection of the current trends in private education, away from boarding. Boarding is more common for children in Year 8, and for boys, since respondents in Year 8 represent the users of one traditional preparatory school.

Table 5: Sources of Information by Day or Boarding

Source of information	Day pupils	Boarders
Acquaintances	31	11
Visit	23	44
Results	10	0
Prospectus	8	28
Other	27	17
Chi-Square	DF	Significance
14.70948	4	.00534

Visits may be more important for boarders, because their choice involves the evaluation of accommodation, as well as the selection of an educational establishment. It is likely that boarders rate prospectuses as more useful because there is a tendency for their schools to be further from their homes. In this case, literature can be used to limit the number of long journeys to schools unlikely to be selected. This idea is supported by the fact that families from rural mid-Wales, Dyfed and Powys, rate the relevance of prospectuses much more highly than those in cities such as Cardiff, perhaps because remoteness makes visits difficult, and perhaps because a restricted choice in rural areas leads to more boarding and so to consideration of schools from even further afield (Table 6). Families from the valleys north of Cardiff also differ from the norm of rating acquaintances as the most important source. For them, visits are more important, which is perhaps due to the nature of the grass-roots fee-paying schools available in Mid Glamorgan. Regardless of the area, users of grass-roots schools do not rate examination results as highly as others, but as the child gets older, examination results become a more important source of information for all schools and users. There are no significant differences in the use of the various types of information by class or by parent's education.

Table 6: Sources of Information by Area

	Acquaint.	Visit	Exam	Pros	Other
Cardiff	30	19	14	3	33
Valleys	22	36	18	14	9
Vale	30	23	5	13	29
Gwent	32	27	0	14	27
Swansea	18	24	0	12	47
Mid-Wales	9	18	0	55	18
Chi-Square	DF	Significance
69.28032	20	.00000

The Choice Set

The "choice set" refers to the set of schools considered by each family at the start of the choice process. On the survey form, parents were asked to name the schools in this set and, in a clear majority of cases, all of the schools in each set were of the same type, as defined in Chapter 8. This finding further confirms the typology of schools, and suggests that in many families, the choice of a type of school is the first, perhaps unconscious, step in the process. Families report seriously considering around two schools on average (Table 7). They tend to read more prospectuses, but visit fewer schools than the number of schools they seriously consider. This suggests a coherent whittling-down process of reading about schools; rejecting some, and then visiting a smaller number.

Table 7: Size of the Choice Set

	Mean	Standard Deviation
Number of schools considered	1.70	0.95
Number of schools visited	1.47	1.46
Number of prospectuses read	2.02	2.12

Parents also report that they considered a larger number of schools than their children do. This ties in with the finding above that the participation of the child increases as the process progresses. In this case, one would expect children to report only the number of schools in the choice set from the point at which their involvement becomes significant. They may therefore be unaware of some of the schools initially considered by the parents, before whittling down. This suggestion is confirmed by Table 8, which shows the magnitude of the difference between the views of the two generations in this respect, which showed up despite the unfortunate restriction on the number of schools that parents could mention, as described in Chapter 7.

Table 8: Size of the Choice set by Generation

	Number considered	Standard Deviation
Parents	1.88	.950
Children	1.57	.923
t-value	DF	Probability
5.18	883.16	.000

Despite the differences between the three measures of the choice set - schools, visits, and prospectuses - all three are also highly correlated with each other, and after allowing for the inevitable measurement error, there is a suggestion that the three variables may, in fact, be measuring nearly the same thing - perhaps a level of engagement with the choice process. The correlation between the number of schools considered, and the number visited, is 0.57, and between the number of prospectuses and the number visited is 0.77, for example. Thus, people who consider more schools, in general also visit more, and read more literature. Confirmation of the reliability of these results comes from the finding that those families rating prospectuses or visits as more important sources of information, also tend to read more literature, and to visit more schools, for example (i.e. the respondents are being consistent).

It is also true that among all respondents, families with more children in private schools tend to consider, and visit, more schools, and read more prospectuses, perhaps because of the link between private schools and boarding, as outlined above.

Ratings Of The Reasons For Choice

Tables 9 to 12 show, in descending order, the mean ratings given by all respondents to each of the 73 variables representing choice criteria. On a scale from zero: not important to two: very important, the mean ratings ranged from 0.19 to 1.84, and this range is notionally divided into four sections. Very important is from 1.5 to 2.0, rounding up to 2. Important is from 1.0 to 1.5, rounding down to 1. Less than important is from 0.5 to 1.0, rounding up to 1, and the not important section is from 0 to 0.5. The number of responses used to calculate the mean score are also shown in each table, from the total of 1,267, which gives an indication of the difficulty of the question, or its appropriateness to some respondents. For example, 'lenient or child-centred approach to discipline' is a difficult statement for children to comprehend, and was only rated intelligibly by 962 respondents, while 'boarding facilities', for example, were probably felt to be irrelevant by many, who may therefore have left this question out despite the instructions to answer all questions, rating irrelevant ones as not important where necessary.

Also shown alongside the mean ratings in each figure are the standard deviations, and it is clear that these rise in the middle of the list, and fall way at both ends, in a kind of flattened normal distribution. This suggests two mutually supporting explanations. First, that there is firmer agreement between all respondents on what is very important and what is of no importance, while those ratings in the middle are there not because most respondents see them as moderately important, but because some see them as more important and some do not. As an example, 'dissatisfaction with state schools' is rated as less than important overall, but has the highest standard deviation, indicating that respondents are split on this issue. Secondly, the lower standard deviations of responses near the top and bottom of the list may indicate the effects of truncation by the scale used for measurement.

Respondents may have felt constrained by the scale, and unable to express particularly strong or negative support for an item (this is a disadvantage of using the three point scale). This explanation is confirmed by the behaviour of some respondents when they wished to show hostility, or occasionally very strong support for an item. In addition to writing many blunt and thought-provoking comments, some people also 'redesigned' the question before answering. The reasons used in the survey all appeared in the same form (Figure 3). It was felt unnecessary to provide a negative scale in the design, since all of the questions related to seemingly positive characteristics of a school, but some respondents added extra boxes to the left of the diagram for questions, such as that about the Welsh language, and then ticked these. Such responses appear in the results as a zero, equivalent to not important, but to treat them as such may be to ignore an interesting, if unanticipated, finding. The implications of some of these editions are discussed in Chapter 12.

Figure 3: The Appearance of the Questions

The analysis turns first to the items which most frequently appeared as very important reasons for school choice (Table 9). These very important reasons for choosing a school, come from only two of the categories for choice criteria defined in Chapter 5 - safety and academic outcomes - in almost equal measure. The variable pupil happiness heads the list, as predicted by previous work. One surprise is what this list does not include. None of the variables rated as very important can be classed as situational, organisational, or selective criteria. In not finding an important role for situational criteria, such as travel, this study is very much in a minority (e.g. West et al. 1994). However, many families represented by these findings are using, or intend to use, a fee-paying school. Their occupational class backgrounds show that they are not, in the main, working-class, and this study has also found parents to be more instrumental in making the choice than previously suggested. Since, other work has also suggested that middle-class families are less concerned with situational variables, while children are more concerned with them, it is perhaps not surprising that users of fee-paying schools are using the more adult, middle-class selection criteria described in Chapter 5 (e.g. Smedley 1995). It is also very likely, that by presenting respondents with only a limited range of reasons, previous researchers have ended up over-emphasising certain variables, as noted in Chapter 2.

Table 9: Very Important Reasons for Choosing a School

Criterion	Mean	SD	No. of cases
Happiness	1.84	.40	1029
Good teaching	1.77	.51	1248
No bullying	1.76	.53	1207
Pupil safety	1.75	.49	1070
Teachers' qualification	1.73	.51	1026
Work atmosphere	1.72	.51	1206
Teaches respect	1.71	.51	1076
Caring staff	1.71	.53	1201
Well-managed	1.71	.52	1073
Range of subjects	1.69	.56	1204
Later advantage	1.64	.58	1065
Well-equipped	1.63	.59	1243
Good facilities	1.63	.57	1051
Career prospects	1.61	.60	1224
Exam results	1.61	.60	1189
Balanced education	1.60	.58	1247
Care and welfare	1.59	.60	1253
University entrance	1.58	.64	1239
Well-behaved pupils	1.52	.63	1208

More surprising is the absence of direct evidence of a strong desire for any kind of pupil selection. Single-sex, religion, ability, and social class are all missing as significant influences on school choice. It could be argued that variables such as well-behaved pupils actually represent a type of unconscious code for class or social background, or that a desire for good examination results from a school is actually a desire for high ability pupils in an era of raw score performance indicators, but there are two arguments against this. First, such variables are linked by later analysis with others, that have no such underlying selective message. The school's departmental facilities, for example, are strongly linked to examination results, but do not appear to imply any kind of pupil selection at all. Secondly, several other variables, not rated as nearly as important overall, explicitly measure views on selection, including clever pupils; nice pupils; single-sex; middle-class pupils; religion; ethnic mix, and the proportion of British pupils. It could also be countered that families want selection, but are not willing to state it openly, as was suggested by the findings of Bagley (1995), but this view is both unparsimonious, and unsupported by the frank and unembarassed comments of many respondents on their forms, and later in the interviews. Less surprising, except perhaps to those researching school effectiveness and improvement, is the lack of the majority of the variables to do with school organisation and ethos. This finding has important implications for public choice theorists, since unless a market in schools operates in such a way that popular schools improve over time, the market can have no beneficial effect on standards of education. Families are only making a decision for the present cohort in school, and they want their child to go a school that is safe, and currently obtaining good examination results. Whether the school improves or fails at a later stage is not so relevant to them (Chapter 15).

The next category of variables are those ranked as important in school choice, as shown in Table 10. To a large extent these confirm the conclusions drawn above. Many of these reasons still relate to academic outcomes and the child's happiness, but a number are also concerned with the ethos of the school. The first few selection variables appear, although most of these are in fact to do with anti-selection, such as a good ethnic mix, and tolerance of all religions. While it must be considered that this finding is partly a consequence of what Bagley (1995) called a "category misinterpretation", it does confirm that selection is not reported to be as popular among these families as other studies may have suggested (e.g. Fox 1989). The first situational variable to be mentioned, ease of travel, just makes it into this category.

Table 10: Important Reasons for Choosing a School

Criterion	Mean	SD	No. of cases
Learning difficulty	1.47	.73	1006
Nice pupils	1.43	.65	1240
Visitors welcome	1.40	.66	1250
Responsive to parents	1.40	.64	1086
Exam emphasis	1.38	.69	1166
Equal opportunities	1.38	.71	1233
Child's wishes	1.37	.67	1080
High expectations	1.36	.68	1192
Competition	1.35	.69	990
Firm discipline	1.33	.70	1193
Traditional morality	1.32	.73	1033
Religious tolerance	1.29	.77	997
Sports range	1.26	.73	1082
Small class	1.24	.81	1200
Promotes confidence	1.24	.77	810
Sixth form	1.23	.80	1059
National Curriculum	1.21	.76	1068
Clubs	1.16	.72	1200
Character building	1.14	.77	1123
Progressive style	1.12	.78	1228
Traditional style	1.12	.78	1234
Proportion at 16+	1.11	.77	1044
Sport reputation	1.10	.77	1252
A particular sport	1.07	.84	1015
Ease of travel	1.03	.74	1238
Ethnic mix	1.02	.77	1040

Perhaps the most surprising results are shown in Table 11, where several situational reasons found to be popular or significant in other studies (e.g. Woods 1992), such as going to the same school as friends or relatives, are seen as less than important . These low ratings are even more interesting, since the results of this study include the views of children who are thought to be more concerned with such practical issues. Although, it is possible that this finding is a result of the relatively elevated occupational class of the sample, it also confirms the suspicion that previous studies may have exaggerated the importance of some reasons for choice, by ignoring others. The lists of 'less than important' and 'not important' reasons for choice are reproduced in full here, chiefly to make this point. The 73 reasons suggested in this study probably comprise the longest list used in any comparable study, and therefore currently give the best estimate of the effects of bias through omission in other studies. Another possible explanation of the difference is that the situational variables are only important very early on in the choice process, during the creation of the 'choice set', and that at the time of the survey this moment had passed. Such an interpretation is discussed further in Chapter 14. The list of less than important criteria for school choice, in Table 11, is mostly made up of situational, and organisational variables. The relatively high standard deviations suggest that some of these reasons are more important to some groups of respondents than others, and this possibility is examined once the reasons have been reduced by Principal Components Analysis.

Table 11: Less than Important Reasons for Choosing a School

Criterion	Mean	SD	No. of cases
Friends at school	0.96	.80	1259
Lenient discipline	0.95	.72	962
Head style	0.94	.77	1251
State dissatisfaction	0.94	.84	803
A particular subject	0.94	.74	1072
Useful social contacts	0.93	.76	1021
Mixed	0.93	.82	1061
Strict uniform	0.93	.77	1252
Lower fees	0.88	.79	1090
Higher status	0.88	.81	976
Clever pupils	0.84	.75	1236
Fees help	0.82	.82	1022
Music reputation	0.82	.74	1250
Small school	0.76	.80	1194
Sibling at school	0.63	.75	1238
Boarding facilities	0.61	.82	897
Attractive buildings	0.59	.66	1244
School bus	0.54	.74	1077

Finally, the least important criteria for school choice are shown in Table 12. The clear majority of the reasons rated as not important relate to pupil selection in some form. As well as containing all three of the variables relating to Welsh issues, this list shows that very little support exists for single-sex education, or for selection by social class, religion, or ethnicity. This is an important finding, especially as it comes from a study focusing on fee-paying schools, and one which needs to be examined seriously in the light of previous findings. Of course, despite the indications above, respondents might genuinely want selection and not be prepared to report it, but the absence of any such reports cannot be used to argue that selection is wanted. Also of interest is the finding that despite the preponderance of private school users in the sample, family tradition of using a particular school is of no importance. This result can act as a corrective to those whose research concentrated on the elite private schools, but whose findings have been quoted as indicative of the whole sector (e.g. Fox 1989). This issue is discussed more fully in Chapter 11.

Table 12: Unimportant Reasons for Choosing a School

Criterion	Mean	SD	No. of cases
Mostly middle-class	0.49	.64	1059
Religion	0.47	.70	1232
No Welsh ethos	0.46	.68	1047
Professionals	0.46	.65	857
Welsh language	0.38	.61	1080
Single-sex	0.34	.67	1046
Mostly British	0.34	.64	1022
Family tradition	0.32	.60	1241
Welsh ethos	0.27	.52	1069
Head gender	0.19	.51	1063

In summary, although various criteria may be found to be more significant to some groups of respondents than others, families are chiefly concerned with security, school resources, and academic outcomes, preferring tolerance and a pleasant ethos to selection, and convenience. These respondents express no great concern over management or discipline. The next section looks at the links between the 73 reasons listed here. It does so not by merely collapsing them into smaller groupings based upon theory and common sense, but by measuring the size of the correlations between them, and so uncovering the links in the minds of the respondents, rather than unwittingly confirming the prejudices of the researcher (Chapter 2).

Relationships between the Reasons for Choice

This section looks at the relationships between the choice variables listed above. These relationships are defined in terms of their correlations - the extent to which respondents tend to answer any two of them in the same way. As assessed by Pearson's r, all of the 73 choice variables show a correlation with several others, each of which would be significant at the 5% level if only one coefficient had been calculated. This suggests that there is common variance between groups of these variables, but the large number of respondents means that coefficients as low as 0.06 are tagged as significant. In order to reduce the number of associations, this section only considers coefficients of absolute value 0.3162 and above, so that 10% of the variance is common to the two variables concerned. This common variance is of a large enough size to suggest the existence of an underlying and unifying explanation. Although explanations of the common variance to be found in smaller associations might also be valid, their explanatory power would be weak, since the common variance they would explain could be as little as one third of one percent. All responses are included in the calculations, fee-paying and state-funded, parents and pupils, pilot and main study.

For 20 of the 73 variables, all 72 of their correlation coefficients are below |0.3162|, and since these variables show so little variance in common with any others, they are omitted from the Principal Components Analysis. In addition, 13 of these 20 variables were rated overall as being not important or less than important, and so they will generally be excluded from any further analysis, since they have been shown to be neither important to selection of a school by themselves, nor to be part of a larger underlying concept of any real explanatory value. The most important variable of all; pupil happiness also does not share much common variance with any other reason for school choice, which is in itself a finding of great interest, since it has been assumed in previous work that this reason is really a conglomeration of other contributory reasons (Coldron and Boulton 1991). Pupil happiness is, in fact, unrelated to any other reason for school choice, and one possible explanation of this is that there is so little variance in the responses to this question. This variable has the lowest standard deviation, as nearly all respondents gave it the highest rating. It is probable that happiness is an imprecise and non-technical concept which, although seen as important by all respondents, they may have differing interpretations of how to achieve. It is therefore of little use in explaining school choice until it can be examined in more detail in the light of interviews. This variable, and six other important, but uncorrelated, variables are excluded from the Principal Components Analysis, but are later fitted into the resulting model as far as possible, as reflects their importance to the process of school choice.

Ten further variables correlate strongly only with one of the other ten. The provision of a sixth form is linked to the proportion of pupils at 16+, a good ethnic mix is linked to tolerance of all religions, help with school fees is linked to low school fees, having mostly middle class pupils is linked to having clever pupils, and provision for Welsh language teaching is linked to a Welsh ethos. These variables are excluded from the Principal Components Analysis, but are then fitted into the resulting model as far as possible, in the form of five concepts: sixth form, tolerance, fees, selection, and Welsh. A further six variables correlate strongly with only one other variable, which in turn is related to several others. These six are also omitted, and are subsequently correlated with the Principal Components which emerge, with interesting results. The remaining 37 variables all show high correlations to at least two other variables. Each was linked to every other one of the 37, either directly, or through linkage to a common variable. These are the variables that are used in the initial Principal Components analysis. However it should be noted that the PCA has also been run with all 73 variables, then deleting the complex variables and minor factors and the outcome is broadly similar.

Seven Factors Underlying the Reasons for Choice

The reduced set of 37 variables was used in a Principal Components analysis, with orthogonal rotation to a Varimax solution, which yielded seven factors, accounting for 52% of the total variance in the responses, which is a reasonable figure since there is obviously some measurement error, and a recent review of school choice studies in the UK suggests that choice is frequently idiosyncratic anyway (Smedley 1995). The first figures in each of the next seven sections, such as Table 13, show the loadings of the variables for each of the seven factors, including all those with a value of |0.3162| or greater. The seven factors appear in descending order of the total amount of variance that they explained. However, this amount is not necessarily linked to their importance in the process of choosing a school. The factors extracted later are generally linked to fewer variables than those extracted earlier, and explain less of the variance, regardless of the importance of those variables to the respondents. For this reason, the mean importance of each variable is also shown again for convenience, coded as Very important, Important, Less than important, and Not important. Assuming that the measurement error in the variables is around 30% - a miserly estimate as the split-half reliability was over 0.9 - loadings of 0.84 and above can be seen as representing "pure" factor variables (Chapter 7). The abbreviated names, such as the "Sports factor", given to each factor are provisional, and intended to be mnemonic rather than fully descriptive at this stage. A factor score is calculated for each respondent, based upon their ratings of the reasons associated with each factor, and this new score is used in the later analysis instead of the "simple" elementary reasons. In this way, 37 variables are replaced by seven factors. As explained above, a further 36 variables were omitted from the initial Principal Components Analysis as they showed too little common variance with other variables, however, it is interesting to note that nine of these show correlations of greater than |0.3162| with at least one of the seven factors, and these are also discussed below.

The Sports Factor

The first of the seven factors extracted is clearly related to sports activities at the choice school. The high correlations between the three sport variables - range of sports, sporting reputation of the school, and a particular sport - and their large loadings suggest that they, and the factor itself, are in fact measuring the same thing, after allowing for the inevitable measurement error, which is why the factor is provisionally named "sports". This factor is important in choosing a school, but none of the variables related to it are in the very important category. One reason for this, suggested by the relatively high standard deviations for some of the variables, is that this factor is of more relevance to some respondents only. It concerns the provision of non-academic, or extra-curricular, activities, chiefly sports, which, along with a progressive style of education, are more relevant to children than adults in this study. The inclusion of boarding facilities does not imply that this factor is only of concern to those considering boarding, since this variable has a low loading, but because boarding is more commonly associated with fee-paying schools, and all state-funded schools in the survey are day schools, this factor is surely of more relevance to choice within the private sector.

Table 13: Sports Factor (20.1% of the variance)

Variable	Loading	Rating
Range of sports	.76	I
A particular sport	.76	I
Reputation for sport	.75	I
Useful contacts	.55	L
Boarding facilities	.52	L
Progressive style	.46	I
Social status	.44	L
Range of clubs	.43	I
Visitors welcomed	.34	I
Nice pupils	.33	I

As stated above, this sports factor is more relevant to children, and Table 14 shows the surprisingly large difference between the mean ratings from each generation in this respect. Sports are almost irrelevant to parents. The factor is also significantly more important in those families where the child has the major role in choosing, but less important in families from Year 1, in which only the parents took part in the survey, presumably for the same reasons. This finding is the first indication that the factors may not represent different compound reasons used to varying extents by all families, but that some factors may stand for a particular group, or sub-set, of choosers.

Table 14: Rating of Sports by Generation

	Rating of sports	Standard deviation
Parent	-.72	0.950
Child	+.57	.931
t-value	DF	Prob.
-16.81	584.44	.000

The sports factor is more relevant for boarders (Table 15). This finding provides the first indication that different factors may apply to different types of schools, as well as to different types of families. Since the sports factor is related to boarding, it is not surprising that it is also rated as more important by respondents with all of the characteristics so far associated with boarding, including families with boys; in Year 8; where the father attended a private school; and who make more visits, or read more prospectuses. To some extent, these findings are consistent with those of David et al. (1995).

Table 15: Rating of Sports, by Day or Boarding

	Rating of sports	Standard deviation
Day	-.91	1.136
Boarding	+.67	.960
t-value	DF	Prob.
-5.00	53.38	.000

Table 16 lists some of the uncorrelated variables, left out of the original Principal Components Analysis, but which now show a correlation of at least |0.3162| with the Sports factor which emerged. The majority are also rated as more important to children than adults, and the first three situational variables have also been found to be more important to children in previous studies (e.g. Echols et al. 1990). This finding confirms that the sports factor is one used as a criterion for choice, or justification, more by children than adults.

Table 16: Correlation of Sports and Omitted Variables

	Sports
Having friends at the same school	+.38
School runs a bus to your area	+.38
Ease of travel	+.33
Good ethnic mix	+.34
Lenient discipline	+.36
Reputation for music	+.38
Lower fees	+.32
Help with the fees	+.37

The inclusion of music in the list above confirms that the factor is to do with extra-curricular activities in general, of which sports may be merely the most significant. The link to fees also confirms the initial feeling that this factor is likely to be of more relevance to the larger, more expensive traditional foundation schools, which tend to be better equipped, to take boarders, and to be part of the Assisted Places Scheme.

This last point is backed up by analysis of the importance of the sports factor for the users of different school types. It is much more relevant to the choice of traditional fee-paying and state-funded schools, which can offer the facilities, staff, and grounds for many activities, than to any other school types (Table 17). In fact, its importance is in almost direct proportion to the standard of facilities available at the various schools. The Sports factor is particularly irrelevant for users of grass-roots schools, which is probably just as well since their facilities for sports, and their resources for the other traditional extra-curricular activities are almost non-existent (Chapter 8).

Table 17: Rating of Sports by Type of School

School	Sports	Grass	Feeder	Stand	Catholic	State	Traditional
Grass	-.96
Feeder	-.22
Stand	-.16
Catholic	-.05
State	-.03	*
Traditional	+.18	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	27.0296	5.4059	4.2663	.0008
Within	590	747.6047	1.2671
Total	595	774.6343

The Welfare Factor

The second factor, welfare, is clearly very important in school choice, possibly because it is seen as equally important by parents and children, and so it displays less variation by sub-groups of respondents than the Sports factor (Table 18). Those variables with particularly high loadings are concerned with the welfare and safety of the pupil. This factor represents a choice of school in which the parents feel their child will be safe, and allowed to get on with an education, since they will be taken care of. It is also a school in which parents are welcomed, and in which the other pupils are not much of a threat. As such, it is more important for families with children in Year 1, who may feel that the child requires more protection at that age, although it must also be remembered that the Year 1 respondents represent only parents from fee-paying schools.

Table 18: Welfare Factor (10.4% of the variance)

Variable	Loading	Rating
No bullying	.68	V
Caring staff	.61	V
Visitors welcome	.55	I
Nice pupils	.54	I
Well-behaved pupils	.54	V
Pastoral care	.50	V
Good teaching	.50	V
Work atmosphere	.47	V
Range of subjects	.44	V
Well-equipped	.34	V

The welfare factor is of more relevance to families in which the father has no degree, and who use visits as the main source of information, but consider fewer schools. These characteristics are linked to families in the working class in Chapter 9, and are also those particularly associated with the users of grass-roots schools, so it is not surprising that the welfare factor is more important to them than to users of the larger schools (Table 19). To some extent, this is a contrasting factor to that of sports, which is less significant for families from Year 1, and more favoured by users of large schools.

Table 19: Welfare by School Type

School	Welfare	Stand	State	Traditional	Feeder	Catholic	Grass
Stand	.10
State	.14
Traditional	.16
Feeder	.26
Catholic	.27
Grass	.88	*	*	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	9.4039	1.8808	3.1422	.0083
Within	590	353.1500	.5896
Total	595	362.5539

The Outcomes Factor

The third factor, outcomes, is also very important in school choice. The variables with high loadings in Table 20 make it clear that parents and pupils want high quality provision of services from their schools, and that the desired outcome of such provision is an advantage in terms of certification, later career trajectory, or social standing. This is the factor upon which performance indicators, in the form of league tables of examination results, are based in the UK. When policy makers and researchers speak of "school improvement", it is these kind of outcomes that they generally seek to improve.

Table 20: Outcomes Factor (5.0% of the variance)

Variable	Loading	Rating
Exam emphasis	.64	I
Exam results	.62	V
University entrance	.60	V
Career prospects	.60	V
Later advantage	.50	V
Social status	.44	L
High expectations	.44	I
Teachers' qualifications	.41	V
Competitive environment	.33	I

The outcomes factor is more relevant to families with a local minority religion, such as Islam, who also find league tables more useful than visits in selecting a school, and who, in fact, make fewer visits than others (Table 21).

Table 21: Outcomes by Family Religion

	Outcomes	Protestant	None	Catholic	Minority
Protestant	-.30
None	-.08
Catholic	+.20
Minority	+.55	*	*
Source	DF	SS	MS	F ratio	Prob.
Between	3	25.0993	8.3664	7.7642	.0001
Within	272	293.0963	1.0776
Total	275	318.1956

This factor is forward-looking, being less concerned with the school itself than with what happens after school, and so it is not surprising that interest in the examination and career outcomes of school rises almost directly with the age of the child in question.

The variable measuring the assertion that private schools produce more self-confident pupils, was left out of the Principal Components analysis, but it shows a correlation of +0.32 with the outcomes factor which emerged. It is an outcome in itself, but since it can only apply to private schools, it suggests that it is private school users who may be slightly more concerned with outcomes. Most fee-payers probably believe that they are buying an academic advantage, and so making an investment for the future (Chapter 11). In fact, although it is true that users of proprietary schools, among whom there are a higher proportion of families with minority religions, rate outcomes more highly, those using grass-roots schools stand out in their neglect of this factor, with state-funded schools somewhere in the middle, which again shows the danger of making simple generalisations about all fee-paying schools (Table 22).

Table 22: Outcomes by School Type

School	Outcomes	Grass	Feeder	State	Traditional	Catholic	Stand
Grass	-1.3
Feeder	-.20	*
State	-.16	*
Traditional	-.14	*
Catholic	-.08	*
Stand	+.35	*	*	*	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	52.7303	10.5461	10.6447	.0000
Within	590	584.5347	.9907
Total	595	637.2650

The Tradition Factor

The fourth factor, tradition, is only of moderate importance overall (Table 23), perhaps because it really only appeals to adults. Two of the variables with high loadings contained the word "traditional" on the survey form; a traditional style of education, and teaches traditional morality. These, along with firm discipline and strict uniform code, probably appeal to the restorationists, reminding them of their own school days, or at least an image of schools which they retain from that era (Chapter 13).

Table 23: Tradition factor (4.2% of the variance)

Variable	Loading	Rating
Traditional style	.70	I
Strict uniform	.67	L
Traditional morality	.57	I
Firm discipline	.54	I
Competitive environment	.50	I
Well-behaved pupils	.42	V
High expectations	.41	I

Children show significantly less interest in uniform, morality, and competition, and more interest in progressive education, tolerance, and lenient discipline (Table 24). As the tradition factor is more relevant to parents than children, it not surprising that it is rated as least important in families in which pupils have the major role in making the choice. Tradition is also rated more highly by families with a child in Year 1, whose responses only represent those of parents, and less highly by the "young adults" in Year 11, presumably for the same reason.

Table 24: Tradition by Generation

	Mean	SD
Parent	+.44	.826
Child	-.44	1.114
t-value	DF	Prob.
11.08	573.89	.000

Tradition is rated more highly by users of the smaller fee-paying schools, including Catholic schools, where Catholic parents have already been seen to have a larger role in the choice process, as well as grass-roots and feeder proprietary schools (Table 25). Of the first four factors, each has been rated more highly by users of different types of schools, and every type of school has at least one factor with which it closely linked. This is both an interesting research finding in its own right, and a further confirmation of the fruitfulness of the typology of schools derived in Chapter 8.

Table 25: Tradition by School Type

School	Tradition	Stand	Traditional	State	Feeder	Grass	Catholic
Stand	-.18
Traditional	-.15
State	-.10
Feeder	+.30	*	*
Grass	+.30
Catholic	+.52	*	*	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	32.7580	6.5516	5.8231	.0000
Within	590	663.8077	1.1251
Total	595	696.5656

The Safety Factor

This is the third of the very important factors. The unifying concept underlying the linked variables is not immediately apparent, nor is it entirely clear how it differs from the welfare factor, to which it should bear no statistical relationship, by definition (Table 26). It could be statistical artifact. The loadings for both the pupil safety, and the well-managed variables are high enough to suggest that they are almost pure factor variables. Perhaps the difference is that whereas the welfare factor is a characteristic of the people in the school, such a "nice" pupils and "caring" staff, safety is more to do with school organisation.

Table 26: Safety Factor (3.6% of the variance)

Variable	Loading	Rating
Pupil safety	.65	V
Well-managed	.65	V
Teaches respect	.60	V
Later advantage	.40	V
Good facilities	.38	V
Teachers' qualifications	.37	V
Traditional morality	.33	I

Safety is less important to pupils, especially those in Year 11, while it is more important to parents, particularly those using state schools. It is likely that this factor chiefly concerns those families with a child still in primary school, whether state-funded or feeder proprietary. Whether any great meaning can be attached to this factor will emerge as a result of the interviews in the following chapters.

The Size Factor

The sixth factor, size, or more strictly, avoidance of size, shows the link in people's minds between small schools, small classes, and dissatisfaction with state-funded provision. State-funded schools, and their classes, are frequently seen as too large (Table 27). The strong association between these variables suggests that this factor is a reason for choosing a private education in general, rather than a specific fee-paying school. This would partly explain the relatively low rating, and high standard deviations, of these variables, since not all families in the survey were considering a private school. The variable measuring the notion that private schools produce self-confident pupils, was left out of the Principal Components analysis, but shows a correlation of +0.32 with the size factor which emerged. This reinforces the suggestion that this factor is used to distinguish between the state and the private sector, rather than to decide upon a particular school.

Table 27: Size Factor (3.2% of the variance)

Variable	Loading	Rating
Small school	.76	L
Small class	.73	I
Dissatisfaction with state	.55	L

A small size of school is more important for families with children in Year 1, probably because it is more important to parents than children. Families looking for a small school also tend to consider more schools, as do users of private schools, confirming the relationship between the two groups. This factor is of much less relevance to families using the larger state funded and traditional foundation schools, but it is significantly more important for users of all private schools than users of state schools (Table 28). This factor can therefore be seen as a sector determinant.

Table 28: Size by School Type

School	Size	State	Traditional	Feeder	Stand	Catholic	Grass
State	-.55
Traditional	-.07	*
Feeder	+.25	*
Stand	+.37	*	*
Catholic	+.42	*	*
Grass	+.53	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	67.2555	13.4511	12.4843	.0000
Within	590	635.6879	1.0774
Total	595	702.9434

The Resources Factor

This is the fourth, and final, very important factor - the material provision and resources of the school - and it is a surprising one, given the poor facilities of most of the schools in the survey (Table 29). It would therefore be a good factor to use to discriminate between school types, since such descriptions as 'well-equipped' can really only be applied to state-funded and traditional foundation schools. The other four types of school are relatively poorly equipped, with restricted facilities and curricula. Unlike size, it would be a school, not a sector, determinant.

Table 29: Resources Factor (3.0% of the variance)

Variable	Loading	Rating
Well-equipped	.58	V
Good facilities	.46	V
Range of subjects	.44	V
Range of clubs	.44	V
Good teaching	.30	V

The resources factor is more relevant to adults than children, which is also surprising, since equipment and facilities are just the sort of practical 'holiday camp' reasons one might expect children to rate highly. It is also more important to families in which the parents have a degree; who use more than one source of information about schools, and who consider more schools. As suggested above, it is rated more highly by those using state-funded and traditional fee-paying schools, and less by proprietary and Catholic school users, and much less by the grass-roots families (Table 30). In fact, as with the sports factor, the rank order of the factor scores by school type, from grass-roots to traditional foundation, almost perfectly matches the relative resources of the various schools.

Table 30: Resources by School Type

School	Resources	Grass	Catholic	Stand	Feeder	State	Traditional
Grass	-1.62
Catholic	-.28	*
Stand	.00	*
Feeder	+.11	*
State	+.21	*	*
Traditional	+.25	*	*
Source	DF	SS	MS	F ratio	Prob.
Between	5	59.6411	11.9282	12.1031	.0000
Within	590	581.4749	.9856
Total	595	641.1160

Conclusion

Even with the heuristics used here, the data are too large and rich for all associations to be considered in the space of one thesis, and even this neglects the fascinating subject of those associations which were not found, but which might have been expected. For example, there is no difference in the rating of any factors by occupational class or area of residence, which is, in itself, a surprising result. The individual differences between schools could be a thesis in their own right. With 97 measured variables in the study, there are over 9000 possible simple comparisons between any two of them. When looking at the means or frequencies of different sub-groups, many possible relationships are present, and a decision has to be made which to follow up. Those avenues generally ignored are: where the link has an obvious and uninteresting explanation, such as the association between school type and year group, or where the link concerns a very specialised group. Some of the findings of specific relevance to fee-paying schools are discussed in Chapter 11, and issues specifically relating to Wales and the Welsh language are discussed in Chapter 12.

In summary, although, the process of choice is commonly seen as a joint one, both generations emphasise their own role in it. Older children have a greater lone input to the process, but there is also a tendency for families about to make a choice to see it as a joint process, while those who have made the choice see it more clearly as having been made by one party or the other. On the other hand, the role of the child generally increases during the process of choice. A personal acquaintance is usually the most important source of information about a prospective school, but families considering a boarding school use more prospectuses, and make more actual visits. The differences could be that boarding families live further away, and are not plugged in to a local grapevine, so they use literature to help decide which schools to visit. The visit is perhaps more important in respect to the school as a place of residence than as a place of learning. Those living in rural areas make more use of prospectuses, perhaps to cut down on unnecessary travel, while examination results are more important to older children, and to girls, who are more likely to be day pupils. Families, in general, appear to seriously consider two schools, although they may previously have read about more than this, but to visit only one of these, on average. This suggests a two-step whittling down process, of first creating a choice set, and then reducing it by research.

Over half of the reasons for choosing a school, originally suggested to respondents, can be reasonably well summarised by seven factors. These are:

Sports: other ECA, convenience criteria, and cost, all combine to create a factor which is reasonably important for children, particularly boys, considering state schools, and for parents considering traditional boarding schools, and relying on visits and prospectuses;
Welfare: a very important factor, used by less highly educated families, with younger children, considering small schools, such as the grass-roots establishments, and relying on visits;
Outcomes: a very important factor, used by families with minority religions, and older children, particularly those considering stand-alone proprietary schools, and relying on League tables of results;
Tradition: an important factor, used by parent-dominated families, especially Catholic ones, with younger children, considering small primary and preparatory schools;
Safety: and organisation, a very important factor, used by parents, considering state schools;
Size: a less than important factor overall, used by parents, of younger children, considering smaller schools in the private sector, who have rejected the state sector, and who consider more schools than most, and
Resources, a very important factor, used by highly educated parents, considering large schools, whether state or private, and who consider more schools than most.

There is little place for pupil selection, school organisation, or convenience in this seven factor model. Some of the factors, such as size, are more clearly sector determinants, more relevant to choice of a type of school, or to a decision as to whether to use fee-paying education. Others, such as outcomes, are more relevance to individual school choice within either sector. This suggests that there may be three steps in the process of choice, with the choice of a school type predating the creation of the choice set, and the two steps described above. Families might choose the fee-paying sector, for the traditional education that they had or would have liked, perhaps represented by grammar schools, and not for modern curricula, progressive teaching methods, or child-centred discipline. Such a traditional education is thought to produce adults with valuable personal and social advantages. Families, in general, also seek smaller schools and classes. Individual schools are chosen for their academic provision, the safety of the pupil, career prospects, and their extra-curricular activities, particularly in the case of boarding schools.

The attitude of parents and their children to five of the seven choice factors are quite clearly different from each other. The fact that the differences are so clear in terms of the generation of the respondents, and that some factors are so highly rated by one group while being almost irrelevant to the other, suggests that the factors extracted may be more than mere items on a check-list for potential consumers. Some may actually represent different categories of consumers. Parents are more concerned with a desire for a traditional type of education, the safety and welfare of their child, the size of the school, and its material provision. Since the size factor also represents dissatisfaction with state-funded education, this provides evidence that it is the parents who are making the choice of sector, and that some factors are more relevant to choice of sector, while others are more relevant to choice of a specific school within that sector. Children seem more concerned with the quality of extra-curricular activities, and the social value of a fee-paying school. The social value factor is only really of relevance to fee-paying schools, and may be the child's way of justifying the choice of this type of school.

The process of choice then, involves parents looking back to the schooling that they had, as well as looking forward to the child's life after school. As they relinquish their child to a school, they are concerned for the child's well-being in an institution which has a large proportion of older children. In response they prefer, and where possible actively seek, a well-run small school with a family atmosphere, where their child will be noticed, and mixing with other pupils that are seen as desirable. Children are the ones who will actually be going to the school day to day, and they also want this experience to be socially pleasant, and fun, which for them is more practical and immediate. They do not want to be bullied, and they do want the school to be easy to get to, but above all, it must offer fun activities and sports, which for them usually means a much larger school than their parents envisage. Although children are also concerned with later life, this is expressed in terms of social networking, as much as academic results. The sports factor, with its association to progressive style and good boarding facilities has a practical everyday theme, and may be seen as a portmanteau of reasons for a child to choose a traditional fee-paying school.