Risk and uncertainty are ubiquitous in life and play a role in almost every important decision that an individual makes. Inter-temporal choices - decisions which involve trading off present and future costs and benefits - are similarly pervasive and important. How individuals perceive risk and their willingness or not to take risks along with their ability to conceptualise the future and their preference over the timing of costs and rewards, will shape a range of life choices and behaviours. These choices and behaviours will have important implications for current and future health, wealth and happiness. Therefore, understanding risk and time preferences is closely linked to the aim of predicting behaviour and identifying individuals in danger of making sub-optimal choices that may damage not only their own social and economic outcomes but those of their families and the next generation. Similarly, the study of altruistic behaviour and how it relates to characteristics and impacts upon individual health and economic outcomes is important for understanding how and why individuals' outcomes differ.

Understanding the processes that lead to engagement in risky behaviours is extremely policy relevant since these activities begin in adolescence and tend to cluster together, increasing the risk of poor adult outcomes, particularly health outcomes. For example, work by Alan Emond and co-authors (forthcoming) shows the clustering of risky behaviours amongst the young people in ALSPAC and their correlation with road traffic accidents. Similarly, work by the MRC's Social and Public Health Sciences Unit shows that drinking, smoking, illicit drug use and risky sexual behaviour are among the major health problems affecting young people in the UK and re-affirms that these behaviours tend to cluster together. Moreover, there is a dramatic tracking of health behaviours from adolescence into adult life.

From a policy perspective it is necessary to understand the decision making processes that lead some people to take up risky activities and also to identify those individuals who are at risk of starting to engage in these behaviours. With respect to risk there is a distinction between two elements that feed into the decision making process: firstly, there is the understanding of the risk, and secondly, there is risk preference - the tastes that lead an individual to behave in a certain way, in full knowledge of the level of risk. Similarly in the domain of time-preferences, there are two elements: the ability to conceptualise the future and plan accordingly; and secondly an individual's preferences over the timing of costs and benefits, having considered and accounted for the future as well as the present.

Currently, ALSPAC contains information on observed risky behaviours undertaken at different ages - for example, smoking, binge drinking, unprotected sex, drug and solvent use, non-use of seatbelts. There are also recorded scores at different ages on numerous scales capturing something of an individual's preferences - locus of control, self-efficacy, impulsivity, executive function, sensation seeking and the gambling related cognition scale. However, there are currently no direct measures of attitude to risk or time-preference. The measures that we propose are the natural complement to this existing battery of instruments recorded at earlier ages which capture some element of risk attitude and time preference. Moreover there are currently no measures of altruism in ALSPAC.

Including measures of risk- and time-preference at age 21 would allow comparison with the earlier indicators and track the development of preferences as children age. It would be possible to assess how well these early less direct measures are able to predict later risk- and time-preferences and identify children vulnerable not only to engagement in risky behaviours but also to making other health and economic decisions that are likely to have negative consequences.

Other datasets - for example the British Household Panel Survey - contain questions on risk attitude but do not have anything like the breadth of information on behaviours and background factors that are available in ALSPAC. Therefore including direct measures of risk- and time-preferences in ALSPAC would provide a unique opportunity to examine the extent to which these preferences are correlated with a wide variety of risky behaviours. The existence of a gradient in these preferences by socio-economic status could also be examined.

Similarly, though other UK datasets measure altruism, they lack the breadth and depth of ALSPAC. Including the "Dictator Game" question in the ALSPAC survey to elicit altruism will also allow an assessment of the robustness of alternative (less intensive) survey questions. For example, numerous "Big-5" modules capture "agreeableness", which is related to altruism. Compared to these, dictator games offer a revealed preference alternative to measuring people's altruism which is felt to be more reliable since individuals are incentivised to reveal their preferences. However, the two measures have not been compared in a systematic way in a large-scale survey.

We would use the results of the dictator game to understand more about variation in altruism across the population and the extent to which it may be linked to an individual's characteristics and background.

In addition, we propose that the parents of the ALSPAC children are also asked to answer the "Multiple Price List" and "Dictator Game" questions, allowing computation of the intergenerational correlation in these characteristics affording an insight into how culturally or genetically they are transmitted from one generation to the next. This would help to shed more light on the origin of the intergenerational correlation in many behavioural and economic outcomes. Moreover, comparison of the correlation between parents' parameters and earlier child measures - for example, sensation seeking - would enable assessment of the stability of the intergenerational correlation in these preferences and whether there is a social gradient in any time-variance in the relationship. The inclusion of these time and risk-preference measures will thus open up a large number of avenues for future research that will have important policy implications.

The risk, time-preference and altruism measures that we propose have been shown to have predictive power for actual behaviour, accurately capturing the underlying parameters of risk, time-preference and altruism. To date this sort of validation has not been possible in the UK, however inclusion in ALSPAC would show how these measures correlate with observed behaviours. For example, does an individual's general risk attitude predict behaviour with respect to diet, exercise, and health investment? Does the extent of patience differ between those who do and do not engage in risky behaviours? Are risk and time-preferences in young adults significant drivers of adult health outcomes? Similarly, what is the role of altruism in determining outcomes such as altruistic acts (such as money donations and volunteering) and how important is it compared to other (e.g. economic) factors. We will also look at the extent to which altruism may relate to other individual outcomes such as employment and wages and also partnership and personal relationships to try and understand whether "nice" people do better or worse in the labour and marriage markets.

There exists a vast literature spanning economics, pscyhology, psychiatry and neuro-science, concerning the ways in which to elicit risk and time preferences and assess the extent of an individual's altruism. One way of doing this is to use "Multiple Price List" questions. In the domain of risk preference, an example of the MPL experiment is the following: the respondent is presented with a table containing 20 rows with each row offering the choice of playing a lottery or taking the safe option of a guaranteed amount. In each row the lottery is the same: with a probability of 0.5 the respondent wins £100, with probability 0.5 the respondent wins £0, however the safe option increases from row to row. In row one the safe option is £0, in row two the safe option is £3.33, thereafter the safe option increases in increments of £3.33 such that in row 20 the safe option is £63.33. If subjects have monotonic preferences then they should prefer the lottery up to a certain safe option value before switching and preferring the safe option in all subsequent rows of the table. The switching point for an individual informs us of their risk attitude: for example, only risk lovers should opt for the lottery when the safe option is greater than £50, while the risk averse will switch at row 16 or before.

In order to elicit time preferences, a MPL experiment can be run to uncover individual's discount rates. For example, the respondent is presented with a table containing 10 rows with each row offering the choice between taking £100 in one month's time or a higher amount in seven month's time. In row one the payment at the longer delay is £102.50, in row two the payment at the longer delay is £105 and thereafter the longer delayed payment increases in £2.50 increments until in row 10 the payment at the longer delay is £125. These payment values represent annual interest rates on the £100 ranging from 5% up to 50%. The point at which the individual prefers to wait for the longer delayed payment rather than having the sooner payment provides information about their discount rate. For example, if an individual prefers the £100 in one month to £110 in seven month's time but prefers £112.50 in seven months' time to £100 in one month then we can infer that their discount rate over a six month window is between 20% and 25%. The length of each delay can be varied - for example 3, 12 and 24 month delays between the sooner and the later payment - and may include having £100 today as the sooner option.

In both the risk lottery MPL and the discounting MPL it is proposed that the respondents are informed in advance that at the end of the module, a number of respondents will be selected at random (for example 1 in 100 respondents) to receive the reward that they opted for in one of the rows of one of the MPL games, again this would be selected at random. The selected respondents will then receive a cheque posted out to them immediately in the case that a row from the lottery game was randomly chosen to determine the pay off, or they will have a cheque posted to them at the sooner time or the later time according to their choice in the selected row if it was a row from the discounting game that was randomly selected. The relevant literature suggests the use of real money payment to randomly selected respondents results in incentive compatible responses and truth telling in these MPL games.

For altruism the approach is to use a Dictator Game. In these games, individuals are asked to choose from a range of different charities. They are then given a specified sum and asked how much they would keep for themselves and how much they would donate to charity. The amount that they keep is then given to them (a small proportion of individuals are actually chosen at random to receive the money). This is a straightforward, quick and reliable way of obtaining a measure of altruism and the real payments incentivise the individuals to truthfully reveal their preferences.