There is a widespread view that low self-esteem is a risk factor for a broad range of psychological and behavioural problems. However, neither public discussion nor decisions to invest in prevention and treatment have been strongly informed or guided by hard evidence, either about the effects or the causes of low self-esteem. This review of the available research evidence, by Nicholas Emler (London School of Economics) aimed to fill this gap. The review found that:

• There is not perfect agreement among researchers about the nature of self-esteem. The most significant division is between the view that self-esteem is a generalised feeling about the self, and the view that it is the sum of a set of judgements about one's value, worthiness, and competence in various domains.
• Despite imperfect agreement about its nature, levels of self-esteem can be reliably and easily measured.
• The design of much, perhaps most, published research means it cannot show whether self-esteem has a causal influence on behaviour patterns. The most informative evidence comes from longitudinal studies, following the same individuals over time. This shows that:
   
  - relatively low self-esteem is not a risk factor for delinquency, violence towards others (including child and partner abuse), drug use, alcohol abuse, educational under-attainment or racism;
  - relatively low self-esteem is a risk factor for suicide, suicide attempts and depression, for teenage pregnancy, and for victimisation by others. In each case, however, this risk factor is one of several and probably interacts with others;
  - there are indications that childhood self-esteem is associated with adolescent eating disorders and with economic outcomes - earnings, continuity of employment - in early adulthood, but the causal mechanisms involved remain unclear.
   
• Low self-esteem in an absolute sense is rare. Most of the comparative research contrasts the consequences of very high self esteem with more moderate levels.
• The strongest influences upon self-esteem are the individual's parents. Parenting style, physical and particularly sexual abuse play a significant role, as do genetic factors.
• Planned interventions can raise self-esteem but knowledge of why particular interventions work, or whether their effects are more than short term, is very limited.

The emerging climate has also had its effect on public policy. For example, the converging notions that high self-esteem is both an asset to society and an individual right has had distinct effects in education. Observers have noted a growing disinclination among teachers to criticise or to set high performance standards, through fear that more objective feedback will damage the self-esteem of pupils. At the same time, there has been pressure on educational authorities to bring in programmes of personal development specifically aimed at enhancing and strengthening self-esteem. In the US, the state of California has invested significant public funds in projects to raise the self-esteem of its citizens.

However, despite an extensive body of research which can inform policy and practice, such initiatives - and the public perceptions that lie behind them - have not been securely rooted in evidence. The aim of this study was therefore to determine what is known from research about three key questions:

• what are the consequences of low self-esteem?
• what factors and conditions determine a person's level of self-esteem?
• can self-esteem be raised through planned interventions?

There is not perfect unanimity within the scientific community as to exactly what self-esteem is, but the major options currently boil down to two. Either self-esteem is primarily an emotional response: it is a generalised feeling about the self that is more or less positive. Or self-esteem is primarily the cumulative result of a set of judgements. These are judgements about one's adequacy across a range of dimensions - intellectual competence, social skills, appearance, physical co-ordination, and so on. According to the first view these judgements are substantially shaped - indeed biased - by the generalised feeling people have towards themselves. According to the second, the generalised feeling is the net result or effect of these more specific judgements.

These alternatives have practical consequences for how self-esteem is measured and there are well-established and widely used procedures based on each of them. There are, moreover, advantages to each kind of procedure. However, the simpler procedures derived from the more emotion-based definition have generally proved more useful and more informative.

The review focused upon the following outcomes: crime and delinquency (including violent crime), racial prejudice, abuse of illegal drugs, illegal (under-age) tobacco use, alcohol abuse, risky sexual behaviour and teenage pregnancy, child maltreatment, educational underachievement, economic circumstances, eating disorders, suicide and suicide attempts.

It is more difficult to prove, unequivocally, the absence than the presence of a link. Nonetheless, in several cases the evidence was about as clear as it could be in ruling out a causal influence of low self-esteem. These cases are crime/delinquency (including violent crime), racial prejudice, teenage smoking, and child maltreatment. What make some of these cases particularly clear is that high, not low self-esteem, is the more plausible risk factor.

A second category covers cases in which the influence of low self-esteem is not proven (these may merit further attention) or very slight. These include educational under-achievement, alcohol abuse and drug abuse.

One commonly voiced assumption is that low self-esteem increases the risk of behaviour damaging to health among young people - notably drug and alcohol abuse and smoking - because it increases vulnerability to negative peer group pressure. In fact, very low self-esteem if anything reduces sensitivity to conformity pressures. It also appears that engaging in physically risky pursuits, such as driving too fast or under the influence of alcohol is associated with high, not low self-esteem.

Finally, with respect to four cases, low self-esteem does appear to be a risk factor: teenage pregnancy, eating disorders, suicide attempts and suicidal thoughts, and (for males only) lower earnings and more extended periods of unemployment in early adulthood.

In each case, however, it is unclear precisely why low self-esteem increases the risk. Moreover, the increased risk is typically small and low self-esteem is only one of a number of risk factors. Nonetheless, these are the strongest cases for further inquiry.

Belonging to a minority ethnic community, particularly one that has experienced a history of persecution and discrimination by the majority population, would seem to pose a threat to any individual's sense of their own worth. If anything, however, quite the contrary is the case. Notably, black Americans enjoy higher self-esteem than their white counterparts.

It is just possible that black Americans are a special case but research points to the same conclusion about all manner of stigmatised groups: being an object of prejudice does not damage self-esteem.

Social class position in adulthood is modestly related to self-esteem, but adolescent and childhood levels of self-esteem are unrelated to the socio-economic status of parents. There is indication, however, of some 'hidden injuries', albeit minor, of social class including father's level of education and whether the father is unemployed.

Gender is also only very modestly related to self-esteem. Females on average have slightly lower self-esteem than males, the gap being widest in the late teens. The many explanations that have been advanced for females' lower self-esteem therefore seem to over-explain the difference.

Among factors that appear to have a clearer effect on self-esteem are successes and failures. Most attention has been given to the effect of academic achievement. The effect is undeniable but it is not large. The same goes for successes and failures of other kinds - in finding work or holding on to employment, for example, and for such public attacks on a person's worth as being diagnosed an alcoholic, referred to a drug abuse rehabilitation programme or convicted of child abuse.

Much attention has been given to the impact of appearance upon self-esteem and strong claims have been made about its effects among adolescents. The evidence, however, does not unequivocally support these claims. It does show clearly that self-esteem is related to beliefs about appearance. It does not rule out the plausible conclusion that these beliefs are themselves substantially determined by self-esteem.

As to what are the most important influences on self-esteem, the simple answer is: parents. Part of this influence is attributable to parenting style. The key qualities contributing to positive self-esteem appear to be approval and acceptance. Among the most damaging things parents can do is to abuse their children, physically or sexually. Family conflict and breakdown are likewise sources of damage.

Biological parents also exercise a genetic influence; a part of the difference between the self-esteem of one individual and the next is inherited. This source of influence is significant and substantial - it is the single most important source of variations in self-esteem so far identified. But it still leaves most of the differences between people to be produced by events after they are born.

Finally, close and loving relationships with others later in life do contribute positively to self-esteem. But the likelihood of forming and sustaining successful relationships of these kinds is itself higher when self-esteem is higher in the first place.

Most of these questions remain for the present unanswered. Undeniably some interventions do work. Moreover, those work best that are grounded in relevant theory and evidence, that are intended specifically to raise self-esteem rather than to produce some other change, and that are targeted at those identified with a relevant problem. But very few evaluations of effects have so far considered long-term outcomes. And next to nothing is known about either cost-effectiveness or why interventions work when they do.

As for the large social problems, those with significant costs for the community and the public purse such as juvenile crime, alcohol abuse, and racism, this research suggests that there is no warrant for the view that low self-esteem plays a significant part. Its clearest effects relate to more private troubles - depression, suicide attempts, being bullied, eating disorders.

Teenage pregnancy is an intermediate case. It is a matter of growing public concern and in the long term it can carry significant public costs. Its incidence, however, remains relatively low. And more effective prevention may well lie in better sex education than in raising self-esteem.

Risk of sexually transmitted infections is similarly a matter of public concern. But findings on the role of self-esteem in moderating this risk are mixed, actually pointing in both directions. If there is a case for further research into the effects of self-esteem, there is one here. This is also true for the repercussions of childhood self-esteem upon economic circumstances in adulthood.

As to what can be done to raise or protect self-esteem, measures to improve parenting skills and remove risks of child abuse would appear to offer the best prospects. Self-esteem can be raised through planned interventions. But the case for doing so, and more particularly for diverting public resources into such efforts, has yet to be made.

On the other hand, particular weight was given to published 'meta analyses' of research findings, where these were available. Such analyses are now widely regarded as a more secure basis for deciding what the accumulated scientific evidence actually shows. Rather than simply counting the number of studies reporting a particular result they combine results from all available sources according to strict mathematical criteria. Whereas in the past reviews typically concluded that a relationship, for example between self-esteem and gender, exists or does not exist, this newer procedure allows an estimation of the strength of the relationship. This kind of analysis therefore has particular value for informing policy.