Shorenstein Center
Kennedy School of Government
Harvard University
Cambridge, MA 02138
Fax: 617 495 8696
Email: Pippa_Norris@Harvard.edu www.ksg.harvard.edu/people/pnorris
Synopsis:
Department of Government
University of Essex
Wivenhoe Park, Colchester
Essex, CO4 7SQ, UK
Fax: 011 44 1206 873894
Email: Sanders@Essex.ac.uk
The principle of political balance is one of the most basic standards commonly u sed to evaluate bias, fairness and impartiality in television coverage. Given its importance, this study examines the impact of three types of balance commonly employed in newsrooms - stop-watch balance, directional balance, and agenda balance. The first s ection outlines the conceptual and theoretical framework. The second section describes the research design based on experiments within the context of the
1997 British general election. The third section establishes the core findings about the effects of b alanced news coverage for party preferences. The conclusion considers the consequences for our understanding of media effects and for the principles guiding fair and impartial election coverage.
© Paper prepared for Panel 38 -12 'The ABC of Media Effects in British
Elections: Agenda, Balance and Change' at the Annual Meeting of the
American Political Science Association, Boston 3 -6 September 1998.

Does Balance Matter? Experiments in TV News
The concept of balance is commonly invoked as one of the main criteria used to evaluate the quality of news, along with related notions such as how far stories are 'fair', 'impartial', and 'objective'
(for a discussion see Westerstahl 1983; McQuail 1992:200 -231). The principle of political balance is often used to dete rmine legal rights to election broadcasts, as well as influencing the news agenda for television producers and editors, and functioning as one of the conventional yardsticks used for scholarly evaluations of bias in journalism. The issue of political balan ce is most sensitive in how broadcasters cover parties and candidates during election campaigns and there is perhaps no area of the news media which is more strictly scrutinized.
Given its importance, this study examines the impact of three types of balance commonly employed in newsrooms - stop-watch balance, directional balance, and agenda balance. The first section outlines the conceptual and theoretical framework. The second section describes the research design based on experiments within the contex t of the 1997
British general election. The third section establishes the core findings about the effects of balanced news coverage for party preferences. The conclusion considers the consequences for our understanding of media effects and for the princip les guiding fair and impartial election coverage.
The Principle and Practice of Balance
Although widely agreed in principle, the concept of balance raises complex and difficult issues in practice. One person's notion of
'balance' may be another's idea of 'bias'. At least three distinct meanings can be distinguished in conventional usage.
Stop-watch Balance
First, the concept is often understood in terms of ' stop-watch ' balance, meaning whether particular candidates, parties or groups have been given equal or proportional coverage in the media (McQuail 1992:
224-5). We commonly tot up the number of column inches for X and Y, and hunt for disproportionality, especially in controversial or disputed matters with two or more opposing sides. Which candidate go t more
2

coverage after a debate? Did we hear more from management or the unions in a dispute? Which party convention got more attention on TV? The simple stimulus-response model behind this notion is that greater exposure means more persuasive influence. Stop-watch balance focuses strictly upon the amount of coverage of each party or candidate, not its contents or effects.
Ever since pioneering work by Lazarsfeld et al. (1944), comparing press and radio coverage of Roosevelt and Wilkie, scholars have trie d to assess whether election reporting has been balanced by this standard.
Content analysis frequently compares the amount of news coverage given to each candidate or party to judge whether the media has been 'fair'
1
.
Following this convention, in the 19 87 British election Miller et al. added up the total hours of coverage of each party on television and concluded that there was a "massive imbalance" which was overwhelmingly right-wing: "The parties did not receive equal treatment...The
Conservative gover nment got two bites of the television cherry: once as a party, once as a government." (Miller et al. 1989; Miller 1991:77; for a critique of this approach see Harrison 1989). British broadcasters and politicians carefully and assiduously stop -watch campaign news to monitor impartiality. In the same way U.S. Presidential campaign debates use rules regarded as fair if equal time is allocated to each candidate. Similar standards are used to grant free political advertising to parties on an equal or proport ional basis, for example in
Denmark, the Netherlands and Israel (Kaid and Holtz -Bacha 1995).
Yet while it is relatively easy and mechanical to count the numerical balance of coverage, this may tell us little about the actual bias in coverage unless we go on to explore its impact. As McQuail notes: "The stop-watch measure of balance in output only takes one to a certain point. It reflects the concerns of interested 'senders' rather than any rational weighing of likely consequences of balance or imbalance. Unless audience reach (and, if possible, response) is also taken into account …[the] amount of media attention is limited as an indicator of media performance." (McQuail 1992: 226). If the conventional rules are changed, for example if one party receives fa r more coverage than another, or if all parties (large and small) receive equal airtime, would this matter?
Directional Balance
Directional balance takes the next step and measures the positive,
3

negative or neutral contents of reporting. In this familiar perspective, which draws on classical liberalism, news is defined as impartial if even-handed towards the pros and cons of any argument: Democrats v.
Republicans, Government v.
Opposition, Tweedledum v.
Tweedledee.
Coverage is regarded as biased if it is disproportionately critical, negative or hostile to one side or the other, or alternatively too hagiographic and propagandistic (see, for example, the Glasgow Media
Group 1976, 1980). Directional accounts emphasize the contents and tone of news more than the more mechanical length of exposure. This notion often features in professional codes of conduct and frequently crops up in informal discussions among journalists. Participant observation studies of British newsrooms during election campaigns have found that editors and producers commonly stress the need for equidistant coverage of the main political parties, as well as even -handedness in commentary, interviews with party leaders, and reports from the campaign trail
(Blumler and Gurevitch 1995; Semetko 1 996). The typical story in most
British election broadcasts tends to present one party's policy proposals or record, and then a rebuttal from opponents, in a familiar
'on the one hand and on the other' sort of format, rather like watching the ball at Wimbl edon. Similar news values have been found in other countries: a five -nation survey by Paterson (1998) found that balance, or 'expressing fairly the position on both sides of a dispute' was one of the commonest ways for journalists to understand objecti vity, especially in the United States and Britain.
Agenda Balance
Lastly, the notion of agenda balance is based on the idea common in theories of voting behaviour that parties often have 'issue ownership': hence, rightwing parties are usually regarded as stronger on issues like defense, crime and inflation, while leftwing parties are usually seen as stronger on education, welfare and unemployment (Budge and Farlie 1983). By focussing election coverage disproportionately on one set of issues, such as proble ms of inner city poverty or immigration, this plays on conventional party stereotypes and can therefore skew the reporting in a partisan direction. Miller et al, for example, argued that the media focus on international security issues in the 1987 British election campaign constituted a "massive and consistent rightwing bias in the issue agenda" which favored the Thatcher government(Miller et al. 1989: 650).
4

The Context of British Broadcasting
How are these different notions of balance applied in the con text of broadcasting British election campaigns? The BBC Producer Guidelines stress the strict need for political impartiality, written into the BBC
Charter, but are fairly vague about how this is to be achieved in practice: "There is an absolute obligati on for the BBC's journalism to remain impartial as the people of the United Kingdom exercise their right to vote.… Editors should ensure that, through the course of the campaign, their coverage has proved wide -ranging and fair."
2
Requirements of 'due impa rtiality' are also written into the Independent
Television Commission's Program Code
3
. Like much British electoral law, the regulations governing broadcasting embodied in Section 93 of the
Representation of the People's Act only controls coverage of parliamentary candidates standing in particular constituencies.
4
Television coverage of parties is determined less by law than by conventions which have evolved since the creation of the BBC in 1922
(Blackburn 1995: 258 -261).
During British election campaigns b roadcasting is dominated by the concept of 'stop-watch' balance, meaning proportional (not equal) coverage of political parties. The allocation of free party political broadcasts is determined by the Committee on Political Broadcasting, created in 1947, a body whose proceedings are never published and which has not actually met in person since 1983, but which is composed of senior representatives from the broadcasters and the parties. Through negotiations this body agrees the time allocation for regular par ty political broadcasts (PPBs) outside of elections, and also for party election broadcasts (PEBs) during the official campaign (for details see
Scammell and Semetko 1995). As shown in Table 1, during the 1997 general election, in line with many previous contests, the ratio was
5:5:4, meaning that the Conservatives and Labour were each given five 10 minute PEBs, while the Liberal Democrats were allowed four, and minor parties which mustered a minimum of fifty candidates, like the UK
Independence Party and the Greens, received at least one five -minute broadcast each, with additional arrangements for Scotland, Wales and
Northern Ireland (for details see Harrison 1997:149 -154)
5
.
Most importantly, the agreed ratio of time allocated to each party also operates in election newsrooms. During the official campaign coverage of parties in election news on all public and commercial
5

television channels reflects the agreed proportion of time allocated to parties for election broadcasts. The stop -watch principle does no t apply to each daily broadcast, but rather to party coverage for each news or current affairs program during the course of the campaign. Participant observation studies of newsrooms during campaigns have found that the
'stop-watch' principle is conscienti ously implemented and continually monitored by news organizations and by parties (Blumler and Gurevitch
1995; Semetko 1996). In 1992 ITN put slightly greater emphasis on news values driven by editorial judgements rather than the stop -watch rule, but in practice the overall time ITN allocated to parties differed little from the 5:5:4 formula which would otherwise have applied (Tait
1995:60). 'Stop-watch' balance is one of the prime ways that British broadcasters try to be impartial. The 1997 campaign follow ed this pattern: the proportion of coverage given to each of the parties in television news closely reflected the allocation of time for PEBs (see
Table 1). Nevertheless other criteria of impartiality are also applied by producers, for example even -handedness in the relative position of parties in the running order and equivalence of tone in reporting campaign events. Leader is matched against leader, issue against issue, and press conference against press conference. In this sense the notions of directional and agenda balance are also pervasive in election newsrooms.
[Table 1 about here]
The question raised by the formal conventions of balance is whether 'due impartiality' is achieved in practice. The situation is carefully monitored by broadcaster s and campaign managers alike, with vociferous complaints if the party balance is seen as unfair. The extent of directional balance in the 1997 election was measured in content analysis by Scammell and Semetko which analyzed the favourability of stories about each party on the main evening news bulletins on BBC1, ITN and Sky News during the April 1997 election camapign
6
. The tone of the story as a whole was coded from the perspective of that party on a 7 -point scale ranging from 'very negative' (1) to 've ry positive' (7). If we compare the mean scores the results in Table 2 show that the mean tone proved very similar for all major parties, although the Conservatives received slightly less favourable coverage on all three channels. If we then recode d the data into three categories to examine the distribution of stories, we can confirm that over 80 percent of all stories fell into the 'internally
6

balanced' or mixed category. Nevertheless the Conservatives did suffer from twice as many negative stories as Labour, while the Libe ral
Democrats had the strongest positive directional bias.
[Table 2 about here]
Nevertheless the effects of balance upon viewers remain largely unknown. The ITC regularly surveys the public's perceptions of bias in the media (see, for example, Gunter et al. 1994). When asked after the
1997 campaign how far the major channels provided news programs which were 'balanced', viewers tended to give high marks to Channel 4, and about half thought that BBC2 and ITV achieved this 'just about always' or 'most of the ti me', compared with one third who thought this of
BBC1's news (Norris 1998). Studies suggest that such perceptions of bias matter for voting behaviour (Mughan 1992). But perceptions may or may not be accurate. Moreover surveys of public opinion can tell us little about the effects of changing the existing news conventions. If the rules were radically altered, for example if third parties like the
Liberal Democrats were given equal airtime, or if there were no rules governing party balance and the contents o f news programs was determined purely by news values, would this matter? Despite the importance of the principle of stop -watch balance for regulating political broadcasting, we do not know. What we need to do is to systematically vary the contents of the m edia message and then monitor the effects.
Research Design, Hypotheses, and Methods
To explore these issues, as one part of a larger multi -method study (Norris et al. forthcoming), the approach adopted here follows the classic logic of experiments. During the 1990s experimental methods have gradually entered the standard repertoire of political research (Iyengar and Kinder 1987; Iyengar 1991; Ansolabehere and Iyengar 1997).
Nevertheless because this approach remains less familiar than survey analysis we will outline our research design in some detail. In order to examine the effects of stop -watch, directional and agenda balance on voters' perceptions, we carried out a series of 15 experiments. To summarize the process, respondents completed a pre -test questionnaire and then randomly assigned to separate groups. Each group was exposed to a distinctive 30 -minute selection of video news and a post -test questionnaire was then administered to each respondent. The purpose of the experiments was to establish the extent to which any changes between
7

pre- and post-test responses varied according to the type of video footage that had been seen.
Fieldwork and the Selection of Respondents
The experiments were conducted in a central London location
(Regent Street). W e included 1125 respondents in total, more than most experimental designs. Participants were drawn primarily from Greater
London and south-east England. Respondents were not selected explicitly as a random sample of the British electorate, but they did g enerally reflect the Greater London population in terms of their social background and party preferences. We chose a busy central London location during the day to provide a diverse group of Londoners including managers, office -workers and casual shoppers . The generalisability of the results rests not on the selection of a random sample of participants, as in a survey design, but on the way that subjects were assigned at random to different experimental groups. Any difference in the response of groups sho uld therefore reflect the stimuli they were given rather than their social backgrounds or prior political attitudes.
One potential problem of experiments is that participants may alter their own behaviour given the artificiality of the research setting and their perceptions of the aims of the study. In order to counter this, respondents were told that they would be participating in research to learn how people evaluate and understand television news.
Prior to the experiment, we informed respondents (fal sely) that we were primarily interested in "selective perception", that is, whether young people and older people, or men and women, are interested in different stories in the news. We did not mention that the news would be about the election, which might well have discouraged participation by the politically apathetic, and we found that many participants believed we were carrying out television market research. We used a single -shot rather than a repeated design so that respondents would not become unduly conditioned by the research process itself.
Participants completed a short (15 -minute) pre-test questionnaire about their media habits, political interests and opinions and personal background. They were then assigned at random to groups of 5 -15 to watch a 30-minute video compilation of television news. Respondents subsequently completed a short (15 -minute) post-test questionnaire. The
8

experiments were carried out in April 1997 during the middle of the official general election campaign. This timing w as deliberate: we wanted to examine the attitudes of participants who had been subjected to the intensive barrage of political coverage that characterizes television news during an election period.
The Construction of Video Stimuli
The video compilations o f news stories were chosen to represent a
"typical" evening news program during the campaign. We drew on stories recorded from all the main news program on the terrestrial channels
7 from mid February until early to April 1997. The videos all had the same format. They consisted of a "sandwich", with identical, standard footage at the top and bottom of each program and one of fifteen different experimental video stimuli in the middle "core" (see the list in Appendix B). Respondents were not told which video was being shown to which group or even that different videos were being watched by different groups of respondents.
To test for the effects of stop -watch balance on party preferences, we monitored the reactions of 261 participants who were divided at random into three treatment groups shown respectively a 20 minute core of Conservative coverage (CON20), 20 minutes of Labour coverage (LAB20), and 20 minutes of Liberal Democratic coverage (LIB20).
Most importantly, the stories selected for the comp osite news bulletins were judged to be 'internally -balanced' or neutral overall. Each story usually opened with the proponent's case and then presented a rebuttal by opponents. A claim that, for example, taxes had been cut under the
Conservatives would be followed by counterclaims by Labour and Liberal
Democrat spokespersons, with equally pro and con comments from vox pop and outside experts. We recognize the difficulty of objectively determining whether or not the content of a particular video selection is genuinely 'internally -balanced'. Our decision rule as to what constituted positive, negative and neutral news coverage was that an item had to be coded as such by two independent coders. We have every confidence, however, that similar codings would have been produced by other researchers. Full transcriptions of the content of the videos concerned are available from the authors.
The four directional experiments involved 258 respondents who were shown consistently positive or negative coverage for the L abour or
Conservative parties as the 10 minute core (see Appendix B), for example
9

one group was shown a series of critical stories about the Conservative government's record in office, divisions over Europe and internal party splits. Lastly the seven agend a experiments, with 484 participants, varied the subject matter of the video core in terms of policy issues, such as whether news about Europe, pensions or taxation. A full description of the directional and agenda experiments are provided elsewhere (Sanders and Norris 1997; Norris and Sanders 1997), and further brief details can be found in Appendices A and B.
In addition to the treatment groups, we also monitored the reactions of an explicit control group of 110 participants who were shown a core with th e standard 5:5:4 ratio of stop -watch party balance.
This control group was used for the descriptive means. Our simple expectation is that subjects in this control group did not significantly change their preferences for the major parties between the pre- and the post-test whereas subjects in the stop -watch groups did.
For the multivariate analysis to compare the relative effects of the stop-watch, directional and agenda balance experiments we recoded all the 15 groups into dummy variables, comparing the effects of those who received one treatment against all the other participants, excluding the control group.
Hypotheses
Based on common assumptions about the effects of news balance, and building upon the results of previous analysis of the dataset presented elsewhere (Norris and Sanders 1997; Sanders and Norris 1997), we developed a series of specific testable propositions. The first set of hypotheses concern the direct effects of television exposure upon party preferences, following the simple logic of direct stimulusresponse, without taking account of any mediating or intervening variables. The core assumption of the principle of stop -watch balance is that, ceteris paribus, (H#1) greater exposure to one party will have a direct positive impact on pr eferences for that party . The core assumption makes no distinction between different types of voters or messages, instead the effects of exposure are regarded as equally likely across all groups in the electorate. We can examine this hypothesis by analyzing changes in the reactions of respondents exposed to each of the different Conservative, Labour and Liberal Democrat 20 minute video stimuli compared with the control group.
10

Critics of the stop -watch principle often argue that to expect any effects simply from greater exposure to one side or another is naïve, at best, and misunderstands the complex nature of the interaction between the news media and viewers (Harrison 1989:655). The principle of directional balance presupposes (H#2) that 'positive' television news of a particular party will tend to create a more favourable preferences for that party while 'negative' coverage will tend to produce less favourable preferences. We have demonstrated elsewhere that direction does matter: positive news was foun d to have a significant impact upon party images, even controlling for a wide range of potentially confounding factors, whereas negative news failed to exert a significant influence (Sanders and Norris 1997). To build upon earlier work, in this study we can compare the relative influence of stop -watch and directional effects.
In contrast, the model of agenda balance suggests that focussing upon certain types of issues will advantage one party or another, depending upon the policy areas which are traditional ly regarded as their greatest strength or home ground. Given the position of British parties across the ideological spectrum we would therefore hypothesize
(H#3) that news programs focussed on taxation will tend to boost
Conservative party support, whereas television stories about social policy (such as jobs, health and pensions) would probably improve Labour party preferences.
Foreign policy issues are more complex to predict, but given the type of coverage during the campaign we would anticipate a priori that stories focussing upon Europe would probably be to the disadvantage of the Conservatives , given well-publicised splits within the back-benches over this issue, whereas news about overseas aid for developing countries would probably slightly benefit La bour . Accordingly we can compare the effects of each of these issues using the groups in the seven agenda experiments.
Mediating Variables
These simple hypotheses, however, say nothing about the different tendencies of various sorts of voters to shift th eir party preferences.
In line with evidence reported by Iyengar and Kinder (1987), we might well expect that the effect of media exposure on party preferences would not be uniform across the electorate but would vary in systematic and predictable ways ac cording to the prior political predispositions,
11

social characteristics and television habits. Many studies, starting with Lazarsfeld et al. (1944), have suggested that the groups most susceptible to influence by the media during the campaign are the waverers who have yet to make up their minds how to vote even late in the campaign. Specifically, we would therefore expect that the politically well-informed, interested and strong partisans would be less likely to change their pre -test to post-test responses than the less informed, uninterested and undecideds. By examining the relative influence of exposure to the videos broken down by the respondent's pre test voting intentions, pre -test interest, and level of political knowledge we can examine these propos itions.
Two other sets of potential effects on voters' changing perceptions need to be considered. Given the information that we obtained from our pre -test and post-test questionnaires, we are also in a position to apply controls both for (a) the stand ard set of socio-demographic variables (age, gender, education and so on) that are normally found to exert significant effects on voters' political preferences and (b) respondents' television -watching habits (such as frequency of watching and trust in TV n ews). As a result of our experimental design, however, our hypotheses in both of these contexts favour the null. The experimental approach is predicated on the random assignment of subjects to test and control groups. If the stimulus
(video exposure) ge nuinely affects the response (a shift in the respondents' party image), the group effect should be observable regardless of respondents' socio -demographic characteristics or television-habits because the test and control groups should contain roughly equal proportions drawn from all groups. At the individual level, we could, of course, simply assume that socio-demographic and television-habit effects do not confound any bivariate statistical relationships that we might observe between video exposure and ch anging party-images. We prefer, however, to conduct formal tests for any such possible effects. Specifically, we hypothesize that the observed relationship between video exposure and pre - to post-test changes in party preferences will not be confounded b y the application of statistical controls for the effects of (a) the standard battery of socio-demographic variables or (b) respondents' television -watching habits. In the subsequent models we therefore test for the relationship between media messages an d party preferences controlling for the mediating effects of the political attitudes, social background
12

and television habits.
Dependent Variable: Party Preferences
We designed the research to achieve conceptual replication of responses, that is, tests we re repeated with conceptually similar but empirically different measures of the variables under scrutiny. We included nine related, BES -based, measures of party support in both the pre-test and the post -test. For each of the three major parties, we asked respondents to assess, on 0 -10 scales, (a) how likely it was that they would vote for the party, (b) how much they liked the party, and
(c) how highly they rated the (named) party leader. For each party, and for both the pre-test and the post -test measures, the three scales were averaged to produce a single party support index where a 0 score meant a respondent had a very poor image of the party and 30 meant a very good image. Tests proved that these items were strongly inter -correlated and formed reliable scales (the Cronbach Alpha was, respectively, .90 for the Conservative index, .89 for the Labour index, and .82 for the
Liberal Democrat index). Accordingly composite party preference indexes were used summing scores on the three separate items. Calcula ting each respondent's "change in party support" score was simply a matter of subtracting the pre -test score from the post -test score: a positive
(negative) change indicated that a respondent had a more (less) favourable view of the party in question after the experimental manipulation than before it. We assume that our composite support scores, precisely because they are based on three different response items, more accurately measure each respondent's party preference than any single party-response item considered in isolation.
Analysis of Results
First in Table 3 we can compare the mean response on the pre -test and post-test average scores on party preferences and the average change over time. The overall pattern shows that respondents greatly preferred the Labour party across all three indicators, but the party enjoyed a particular advantage on the probability to vote score. This increased our confidence in the results, since this pattern was broadly in line with the substantial Labour victory on poll ing day two weeks later. In terms of the change over time from the pre - to the post-test scores, there was very modest change across all parties. Given the limitations of the research design, with only a single stimulae in a 30
13

minute compilation of news, we would expect only very modest short -term effects to be evident. Indeed, if we found any strong results from such a small stimuli we would become slightly concerned about preconditioning participants in our methodology. The lack of overall variance, comb ined with the limited numbers within each video group, reinforces the importance of any findings which do prove statistically significant.
Overall the Liberal Democrats most consistently improved their support while Labour tended to do less well, with the exception of a slight boost in the proportion liking Tony Blair. Exposure to television news also produced a slight increase in the propensity to vote, as found in other studies of the influence of television news (Norris 1997).
[Table 3 about here]
Table 4 goes on to test one of our primary hypotheses by analyzing the mean size and direction of changes in party preferences caused by exposure to the stop -watch video stimuli compared with the control group shown a 'balanced' (5:5:4) news broadcast. The res ults show a mixed pattern. Those who watched the Conservative news bulletin (CON20) experienced a small but significant increase in Conservative support, as predicted by Hypothesis 1, however there was also a significant and stronger increase in Liberal De mocrat support among these participants.
Viewers who watched the Labour video (LAB20) also modestly increased their support for Labour and the Liberal Democrats, although these changes did not achieve statistical significance. Exposure to the
Liberal Democrat bulletin (LIB20) also proved to have no significant effects.
[Table 4 about here]
To test our primary hypotheses about the effect of watching each of the video stimulae, with the range of control variables described earlier, we can turn to the OLS mult iple regression models in Table 5.
The results of the analysis shows that directional balance proved the most significant effect, supporting Hypothesis 2. Specifically the positive news bulletins (where the 30 minute video showed a 10 minute core of positive stories about each party) produced a significant impact in the expected direction for both the Conservative and Labour parties, even after including the full battery of control variables. This confirms and reinforces our previous conclusions about the effects of positive news on preferences for the major parties (Sanders and Norris
1997). In contrast negative news had no significant effect on levels of party support.
14

[Table 5 about here]
The effects of watching the stop -watch party balance videos prov ed slightly more complex to interpret. The simple hypothesis we are testing here is that greater exposure to a party in the news will strengthen preferences for that party. The effects of watching the 20 minute series of stories about the Labour party was straightforward since this had no impact on Labour party preferences. Viewing the
Conservative party bulletin (where the 30 minute video showed 20 minutes of neutral or 'internally balanced' footage about the Conservative party) also produced no significa nt improvement in Conservative party preferences, after controlling for other factors. Indeed, contrary to the hypothesis we were testing, watching the Conservative video unexpectedly served to boost Liberal Democratic preferences. In contrast to our hypothesis, viewing the 20 minute Liberal Democratic news bulletin was associated a slight fall in Liberal Democrat support(although only significant at the .08 level), not an increase.
These results give no support for the core hypothesis about stop -watch balance and the modest effects we did find remain a puzzle. It may be that although we evaluated the selected stories for each party as
'internally-balanced' or neutral, others perceived them as negative. We did not control the creation of these news stories , only their selection, since we were editing typical stories from the campaign.
Perhaps our selection of the Conservative and Liberal Democrat stories provided some unintentional directional cues against these parties.
Lastly the assumptions of agenda ba lance, as suggested by Miller et al. (1989), also find no support in these tests. Showing respondents a 30 minute bulletin which includes a 10 minute core of stories about one of six different issues produced no significant boost for any party. On these g rounds we have to discount Hypothesis 3.
Conclusions and Implications:
The BBC Producer's Guidelines stress: " Due impartiality lies at the heart of the BBC. It is a core value and no area of programming is exempt from it. It requires programme makers to show open-mindedness, fairness and a respect for truth.
"
8
Requirements of due impartiality are written into the BBC Charter and the Independent Television
Commission's Programme Code
9
. The principle of balance is the standard most commonly used by journal ists, broadcasting regulators and scholars
15

to judge the impartiality of news. The question explored by this study is how far stop-watch balance, directional balance and agenda balance actually exerted any influence on experimental subjects during the 1997
British election campaign.
The assumptions of stop -watch balance are that there needs to be proportional treatment of the major parties in the news because greater coverage of one of the major parties would automatically give them an electoral advantage. The principle assumes that media exposure, ceteris paribus, will lead to persuasion. But the results of this analysis strongly suggests that the principle of stop -watch balance, which is so pervasive in regulating coverage of election news in British broadcasting, is not supported by this evidence. The results suggest that the short-term effects of the amount of coverage of each party in the news, measured by the stop -watch principle, does not provide an automatic boost in that party's support, or a consi stent decrease in support for other parties. On this basis we find little evidence for
Hypothesis 1. The assumptions of directional balance are that what matters is less the amount than the positive or negative contents of television news. In this view im partiality is achieved by an evenhanded approach to criticism and praise of each party. Our evidence, confirming
Hypothesis 2, suggests that directional balance is important, in particular that positive news has the capacity to provide a short -term and modest boost to that party's fortunes. Lastly we found no support for Hypothesis 3 about the need to maintain agenda balance in newsrooms: the policy topic of stories which we examined had no differential effect upon levels of party preferences. A newsroo m which focussed heavily on coverage of health and jobs, for example, would not thereby be improving Labour support any more than one which carried many stories about tax would necessarily help the Conservatives. In a nut shell, we conclude that what matt ers is how parties are covered in television news, in particularly positive stories, rather than how much coverage they receive or on what topics.
These results may have important implications for public policy and the conventions which dominate British broadcasting during election campaigns. If we can generalize from the results of short -term experiments to the cumulative experience of the real world, and this is a real 'if', then it follows that all the paraphernalia of stop -watch balance which is so c arefully monitored by the party managers and by
16

news executives and producers may, at best, be irrelevant, and, as worst, may obscure any real biases in British television journalism. A more laissez-faire attitude to time balance, so that stories are driv en by news values more than by the need to give proportional coverage to each party, irrespective of what they are actually doing during the campaign, might produce more stimulating and effective coverage. But since it is relatively easy and mechanical to measure the time allocated to different parties, while it is highly problematic and controversial to measure directional bias, we can probably expect that even if irrelevant, the principle of stop -watch balance will continue to be observed in British broad casting. If nothing else, it provides a modest fig-leaf of impartiality to clothe naked and unprotected journalists hit by the chill winds of charges of party bias.
17

Appendix A:
Video News Compilations Used in the Party Balance Experiments
All video compilations had th e same beginning and end segments consisting of:
Beginning
1. Report on the opening day of the official campaign.
Commentary by Robin Oakley. Statements by Major, Blair and Ashdown.
Discussion by Oakley of the four main issues of the campaign: the economy, Europe, constitutional reform and leadership. [5 minutes].
2. Continuation of news on the opening day of the campaign.
Description of Major's, Blair's and Ashdown's activities during the day.
Description of the timetable for the election and the ti ming of the next
Queen's Speech. Discussion by John Pinnaar on the failure of the
"feelgood factor" to re -kindle Conservative fortunes thus far. [5 minutes].
End
1. Description of a runaway horse incident at a racecourse. [2 minutes].
2. Discussion of m otorway traffic congestion. [1.5 minutes].
3. Incident of two bombs left beneath flyover on the M6. [1.5 minutes].
4. Discussion of the introduction of self -assessment tax forms.
[1.5 minutes].
5. Discussion of freemasonry in public life. [1.5 minut es].
6. Results of the Oscar award ceremony in US. [2 minutes].
18

Appendix B
Timing of 30 minute Video Experimental Stimuli
Experiment
STOPWATCH BALANCE
Start Core
Control
Conservative Balance
10 neutral 4 Con: 4 Lab: 3
5 neutral
LibDem
20 minute Con neutral
Labour balance 5 neutral
Liberal Democrat 5 neutral
20 minute Lab neutral
20 minute LibDem neutral
DIRECTIONAL BALANCE
Con Positive 10 neutral 10 Con positive
Con Negative
Lab Positive
Lab Negative
10 neutral 10 Con negative
10 neutral 10 Lab positive
10 neutral 10 Lab negative
AGENDA BALANCE
Tax
Jobs
10 neutral 10 tax neutral
10 neutral 10 jobs neutral
Health
Pensions
End
10 minute neutral
5 minute neutral
5 minute neutral
5 minute neutral
10 neutral
10 neutral
10 neutral
10 neutral
10 neutral
10 neutral
10 neutral 10 health neutral 10 neutral
10 neutral 10 pensions neutral 10 neutral
10 neutral 10 Europe neutral 10 neutral Europe
Overseas Aid
Issue Control
10 neutral
10 neutral
10 overseas aid
10 mixed issues
10 neutral
10 neutral
19

References
Ansolabehere, Steven and Shanto Iyengar. 1997. Going Negative.
New
York: Free Press.
Blackburn, Robin. 1995. The Electoral System in Britain . New York: St
Martin's Press.
Blumler, Jay G. and Michae l Gurevitch. 1995. The Crisis of Public
Communication . London: Routledge.
Budge, Ian and David Farlie. 1983. Explaining and Predicting Elections .
London: Allen & Unwin.
Glasgow Media Group. 1976. Bad News . London: Routledge and Kegan Paul.
Glasgow Media Group. 1980. More Bad News . London: Routledge and Kegan
Paul.
Gunter, Barrie, Jane Sancho -Aldridge and Paul Winstone. 1994. Television and the Public's View , 1993 . London: John Libbey.
Harrison, Martin. 1989. 'Television Election News Analysis: Use and
Abuse - A Reply' Political Studies XXXVII(4): 652-658.
_____. 1997.'Politics on the Air'. In David Butler and Dennis Kavanagh.
The British General Election of 1997 . London: Macmillan.
Iyengar, Shanto, 1991. Is Anyone Responsible?
Chicago: Chicago
University Press.
Iyengar, Shanto and Donald R Kinder. 1987. News That Matters.
Chicago:
Chicago University Press.
Kaid, Lynda Lee and Christina Holtz -Bacha (1995). Political Advertising in Western Democracies.
Thousand Oaks, CA: Sage.
Lazarsfeld, Paul, Bernard Be relson and Hazel Gaudet. 1944. The People's
Choice . New York: Columbia University Press.
McQuail, Denis (1992). Media Performance . London: Sage.
20

Miller, William L., Neil Sonntag and David Broughton (1989). 'Television in the 1987 British Election Campaign: Its Content and Influence.'
Political Studies XXXVII (4): 626-651.
Miller, William L. 1991. Media and Voters . Oxford: Clarendon Press.
Mughan, Anthony. 1996. 'Television Can Matter: Bias in the 1992 General
Election'. In David M. Farrell et al.
(eds.) British Elections and
Parties Yearbook, 1996 , pp.128-142. London: Frank Cass.
Norris, Pippa and David Sanders. 1997. It Was the Media, Stupid:
Agenda-Setting Effects During the 1997 British Campaign . Paper prepared for delivery at the 1997 Annual Meeting of the American
Political Science Association, Sheraton Washington Hotel, August
28-31.
Norris, Pippa. 1997. Electoral Change since 1945 . Oxford: Blackwell.
Norris, Pippa (1998). 'The Battle for the Campaign Agenda'. In Anthony
King (ed) New Labour Triumphs: Britain at the Polls . Chatham. NJ:
Chatham House.
Norris, Pippa, John Curtice, David Sanders, Maggie Scammell and Holli
Semetko. (1999). Communicating the Campaign . London, Sage.
Sanders, David and Pippa Norris. 1998. 'Negative News and Political
Cynicism'. Paper presented at the PSA Annual Conference, Keele,
April.
Scammell, Margaret and Holli Semetko. 1995 'Political Advertising on
Television'. In Lynda Lee Kaid and Christina Holtz -Bacha (eds.)
Political Advertising in Western Democracies.
Thousand Oaks: Sage.
Semetko, Holli. 1996. 'Political Balance on Television'. The Harvard
International Journal of Press/Politics . 1(1):51-71.
Tait, Richard. 1995. The Parties and Television. In Ivor Crewe and Brian
Gosschalk. Political Communications: The General Election Campaign of 1992.
Cambridge: Cambridge University Press.
Westerstahl, J. 1983. 'Objective News Reporting'. Communications
Research . 10;403-424.
21

Table 1: Ratio of Election Broadcasts and Proportion of Election News,
Major Parties 1964-1997
Party Election Proportion of BBC TV Proportion of ITV
1964
1966
1970
1974F
1974O
1979
1983
1987
1992
1997
Broadcasts
Con
5
5
5
5
5
5
5
5
5
5
Lab
5
5
5
5
5
5
5
5
5
5
LDem
3
3
3
3
4
3
4
5
4
4
Election News
Con
42
39
46
39
35
35
34
35
32
35
Lab
41
42
42
40
35
35
36
31
32
31
LDem
17
16
10
18
26
22
26
25
31
24
Election News
Con
41
43
45
37
37
37
34
38
31
35
Lab
39
40
43
36
37
37
37
31
34
30
LDem
Sources : Information about PEBs from Margaret Scammell and Holli Semetko
1995 'Political Advertising on Television' Table 3.1. In Lynda Lee Kaid and Christina Holtz -Bacha (eds.) Political Advertising in Western
Democracies (Thousand Oaks, Sage). The proportion of election news per party from successive volumes of T he British Genera l Election of.. by
David Butler et al.
20
16
11
22
21
21
27
26
32
28
22

Table 2: Directional Balance of Television News, 1997 Campaign
Conservative
Party
Labour Party Liberal
Democratic
Negative
Mixed
Positive
12.2
82.5
5.4
6.3
84.9
8.8
Party
0.9
92.0
7.1
Negative Minus Positive
Mean Directional Score
-6.8
3.92
+2.5
4.04
+6.2
4.08
N. of Stories 1267 1276 1269
Note: Coding of political stories on BBC1 9 O'Clock News, ITN News at
10, and Sky News evening bulletin. The Directional Score is calculated on a 7-point scale from 1 (Neg ative) to 7 (Positive).
Source: Seketko, Scammell and Goddard 1997 British General Election
Campaign Content Analysis.
23

Table 3: Average Rating of Party Preference Variables
PrePosttest test
Mean
Change
N. for mean
Probability of voting Conserv ative
Liking of Conservatives
Liking of John Major
Conservative party preference index
Probability of voting Labour
Liking of Labour
Liking of Tony Blair
Labour party preference index
2.82
3.19
3.86
5.67
5.30
4.94
16.2
3.01
3.32
0.21
0.14
3.84 -0.02
9.98 10.31 +0.34
5.50 -0.09
5.29 -0.03
5.08 +0.13
16.0 -0.02 change
945
1082
1084
916
1003
1085
1089
936
Probability of Voting Liberal Democrat
Liking of Liberal Democrats
Liking of Paddy Ashdown
Liberal Democrat party preference index
Probability of Voting
3.25
4.46
4.69
12.4
7.24
3.43
4.64
4.76
12.9
+0.20
+0.13
+0.04
+0.31
7.41 +0.19
889
1053
1068
859
1113
Note : All measured on a 10 -point scale except for the index which is summed to a 30 point scale. The index score is the summed aver age of the other three scores in each party grouping. It is only measured for those respondents who answered all six component questions.
Source: Television News Experiments, April 19 97.
24

Table 4: Mean Changes in Party Preferences by Stop-Watch Balance Video
Stimuli
Control video
Change in
Conservative
Change in
Labour preferences preferences
Mean P. N. Mean P. N.
-.28 92 -.26 98
Change in
Liberal Democrat preferences
Mean P. N.
+.22 90
Cons. video 65
Labour video
+.69 *
.01
70 +.14 78 +1.29 *
76 +.50 76 +.51 71
LibDem video +.36 56 -.84 * 61 -.34 50
Note : The significance of the mean difference between the control group and the exposed group is measured through ANOVA. P. *=05. **=.01.
Source: Television News Experiments, April 19 97.
25

Table 5: Changes in the Party Preference by Video Balance Experiments
Video Group Change in
Conservative
Preferences
Change in
Labour preferences
Change in
Liberal
Democrat preferences
Constant
B
-2.0
Sig B
-.59
Sig B Sig
.57
STOPWATCH BALANCE
CON20
LAB20
LIBDEM20
DIRECTIONAL BALANCE
Con Positive
Con Negative
.06
-.02
.02
.12 .04
.64 .04
.11 ** .01 -.04
.03 .35 .06
.29
.33
.26
.13
.09 * .02
.01 .88
.68 -.08 * .02 -.07 .08
-.01
.02
.91
.69
Lab Positive
Lab Negative
.06
.04
.09 .11 ** .01
.31 .03 .50
AGENDA BALANCE
Tax Issue
Jobs Issue
.02
.01
.62
.86
.02
.01
Health Issue
Pensions
Europe
.02
.05
-.03
.67
.21
.49
.04
-.01
-.03
Overseas Aid
CONTROL VARIABLES
Gender
.06
.04
.12
.26
.04
.01
Graduate
Age
Ethnicity
.09 **
-.02
-.01
.01
.61
.69
-.01
.04
.03
TV Trust
TV Use
-.01
.06
.04
.87 -.01
.09 -.06
.30 -.01 Political Knowledge
Political Interest
Own vote
.10 **
.05
.01
.20
.05
.01
Undecided vote
R 2
N.
.04
.04
916
.32 .05
.04
969
S.e. of estimate 2.64 3.18
Note : OLS regression models. P. *=05. **=.01.
Source: Television News Experiments, April 1 997.
.58
.89
.01
.02
-.08
3.36
.80
.54
.88 -.03 .49
.30
.93
.43
.36
.83
.75
.24
.27
.71
.07 -.03 .42
.20
.85
.22
.01 .80
-.01
-.01
-.06
-.01
.01
.01
-.01
.01
.08 * .02
-.02
-.09 **
-.01
.03
859
.65
.84
.14
.69
.70
.85
.76
.98
.07
.68
.02
.75
26

1
See, for example, the chapter on broadcasting in D.E. Butler and
Anthony King The British General Election of 1964 (1965, London:
Macmillan) and successive editions.
2
See the BBC Produ cer Guidelines 1996 Chapter 19 (5.1). http://www.bbc.co.uk/info/editorial/prodgl
3
See ITC Programme Code: http://www.itc.org.uk/regulating .
4
Section 93 of the Representation of the People's Act is designed to ensure strict impartiality by requiring the prior consent of every candidate in a seat to any broadcast about the constituency during the campaign. But it only regulates coverage in of part icular parliamentary candidates running in particular constituencies, not the national campaign.
5
It should be noted that the parties determine the editorial contents of the programs although as the publisher the broadcasters must ensure that the programs follow the law on libel and contempt, and also follow accepted standards of taste and decency. The precise timing of election broadcasts is also within the hands of the television companies.
6
We would like to thank Holli Semetko and Maggie Scammell for access to this data, which will be forthcoming in Pippa Norris et al.
Communicating the Campaign (Sage, forthcoming)
7
The programs sampled were Nine O'Clock News (BBC1), News at Ten (ITN),
Channel Four News and Newsnight (BBC2).
8
See the BBC Producer Gu idelines 1996 http://www.bbc.co.uk/info/editorial/prodgl .
9
See the ITC Programme Code http://www.itc.org.uk/regulating .
27