|This report presents the results of a study of retirement behaviour among Norwegian state employees aged 50 to 70 years. The purpose of the study is to document any trends in the average age of retirement of state employees from 2001 to 2007, map the specific retirement routes used by state employees, and to analyse variation in the timing and choice of retirement route by gender, education level, and between occupational groups and different branches (services) within the state sector. The study draws on data combined from two sources: a governmental employee register and a register for a occupational pension (superannuation) program for state employees. The employee register provides information about all governmental employees by 1st October each year from 2001 to 2007. The occupational pension register provides information about the take up of pensions from 2001 to May 2008. Retirement behaviour is thus investigated over a succession of 12 month periods (from October one year to October the following year) plus the shorter period from October 2007 to May 2008. The obligatory retirement age for most state employees is fixed at the age of 70, but they have a right to retire and take out old age pensions three years earlier from the age of 67. Some groups of state employees face lower age thresholds at 60, 63 or 65, known as special age limits. The option to voluntarily leave three years earlier than the age limit also applies to these groups. The majority of state employees also have the option to leave from age 62 as part of a negotiated early retirement pension program known as AFP, which is more beneficial to take out, however, from age 65. All groups can be granted a disability pension at any age up to 67, when claimants are forced to take out an ordinary old age pension. The analysis applies three types of statistical methods. Survival curves indicate at which age a theoretical cohort leave state employment and/or take up a pension. Average retirement age indicates the average age of those retiring, in terms of leaving employment and taking up a pension. Also discrete time (proportional odds) models indicate how retirement ages have changed and vary between occupational groups and services. This is an alternative method which provides better statistical control for the age composition of the employees and those retiring. Various types of retirement routes are also investigated using competing risk (multi nominal) discrete time models. By and large, the different statistical methods provide consistent results, with a slight exception when comparing state branches/ services and some individual years. The age distribution of state employees is almost uniform from age 50 to 60. The number of employees drops sharply, however, in the age brackets above 60. This drop reflects that many employees start retiring from their early 60s and (in addition) that cohorts born after Wold War II are much larger than those born up to 1945. Survival curves show that some retire already at age 57, typically men working in the police force and the military services. Most other groups can retire only from age 62. More men than women retire before 62 but more men than women also tend to work after 66. Women retire most typically between 62 and 66 years. Surviving curves also show that people retired later in 2006/2007 than in 2001/2002. In particular, fewer people retired in the age bracket 62-64 in 2006/2007 compared to similar age groups five years before. Also average retirement age among retirees and discrete time models indicate that people retired later over the study period. Average retirement age increased from 61.3 in 2001/2002 to 62.4 in 2006/2007. The retirement age increased more strongly among women than men. In 2001/2002 women retired earlier than men, in 2006/2007 women retired later than men did. The analysis distinguishes between 13occupationalgroupsand 22 governmental services. The largest occupational group is â€˜executive officers', people who are responsible for dealing with various types of applications. Universities and higher education is the largest service. Police and military personnel retire earlier than any other occupational group, whereas university/college teachers and researchers retire latest. Among services, it is also the police who retire earliest whereas the Ministry of Foreign Affairs has the oldest retirees. Linear regression models indicate that the very low retirement age among police and military personnel is explained by the high frequency of special age limits which apply to these groups. Retirement age is positively correlated with education level; people with higher education retire later than people with less education. The high retirement age at universities/ colleges is only partially explained, however, by the high education level in this service. Retirement age vary more between occupational groups than between services. This is even more clearly when analysing both factors in a single statistical model. Much of the occupational level variation is explained by special retirement ages, however, which tends to be granted to occupational groups rather than services. Some occupational level variation is also explained by education level. When controlling for special retirement ages and education level, occupational level and service level variations are of similar magnitude. The composition of occupational groups changed over the study period. This compositional change in occupations can help explain why average retirement age increased in the period. This factor can only explain a minor part of the increased retirement age, however, estimated to 24%. Also increased education level could contribute to higher retirement age but this factor appears to work via the changing composition of occupations. Retirement age developed differently within occupational groups and services. Clerks delayed their retirement more strongly than any other occupational group. But this groups was also substantially reduces in numbers, and may not be fully comparable throughout the fire year period. Also executive officers, advisers (in governmental departments), and university/ college teachers and researchers increased their retirement age substantially. Technicians and military personnel were the only occupational groups reducing their retirement age. Retirement age also developed differently between services but these results tend to be sensitive to variation between individual years. Still, the Labour and Welfare Administration (NAV), the Public Road Administration and governmental departments increased their retirement age substantially over the period. The data cannot fully identify all available types of retirement in terms of pension programs, and even less so when analysing how retirement changed over the study period. Disability retirement has seemingly decreased more strongly over the study period than the take up of other pensions. One should be careful to note here that some services, particularly many manual jobs, were privatized and moved out of the governmental sector during the study period. Some employees continue working also after taking up a pension, typically known as flexible retirement. Before 62 years of age such flexibility is more frequent among women than men, typically by combining a 50% disability pension with a 50% job. Above 62 years the gender gap in flexible retirement narrows. Flexible retirement increased over the study period. There is only moderate variation in flexible retirement between occupational groups and services. The report investigates retirement behaviour using various methods, including the average retirement age among retirees and discrete time (proportional odds) models for everyone at risk for retirement.By and largethe twomethods give similar results, particularly when comparing occupational groups (adjusted r=0.965). There is some variation between the two methods, however, when comparing governmental services (adjusted r=0.85).