Does unemployment have any impact on party choice? If it does, then in what way? Will the unemployed tend to support parties to the Left or to the Right? Is there perhaps variation in this pattern over time, with the impact of unemployment differing from one historical period to the next? A recent analysis of the relationship between unemployment and electoral performance in five European countries (the United Kingdom, the Federal Republic of Germany, Belgium, the Netherlands and Austria) concludes that this relationship 'is negative as expected, but very weak. Moreover, the relationship is statistically insignificant' (Visser & Wijnhoven 1990, 72). Another result from this study is that in periods of mass unemployment, Conservative parties perform better than Social Democratic parties.

Election consequences of economic conditions have been studied extensively,particularly the USA. Unemployment has been investigated together with such issues as inflation and taxation, with the focus being on the impact of economic concerns upon voting. Two hypotheses have been important in the American debate: (i) an incumbency-oriented hypothesis and (ii) a policy-oriented hypothesis (Kiewiet 1981). The former emphasizes

Resumé

By the late 1980s, the absolute number of jobless in Norway had reached the same level as that during the Great Depression of the 19305. Unemployment in the 1930s is known to have had an impact on party choice. Among those hit by unemployment in the 1980s, however, we no longer find a strong correlation between unemployment and party choice. This situation constitutes the point of departure for this article. The relationship between joblessness and party choice is investigated on the basis of four hypotheses: (i) The incumbency thesis, which suggests that the incumbent party receives support in accordance with fluctuations in economic conditions - i.e. it is punished in bad times and rewarded in prosperous times, (ii) The policy thesis, which predicts that voters concerned about unemployment prefer the Labour Party, (iii) The social composition thesis, which argues that the fluctuations between unemployment and party choice are simply due to changes in the social composition of the unemployed, (iv) The generation thesis, which underscores that the impact of unemployment on party choice varies according to historical generation. Confronted with empirical evidence, the social composition thesis is rejected. Likewise, support for the policy thesis has gradually vanished. Some empirical support is given to the incumbency thesis. In the end, however, the generation thesis appears to be the most relevant.

that the incumbent president or candidate receives voter support according to fluctuations in economic conditions. During difficult times the politicians in power are punished in elections: voters turn against the incumbents and 'throw the rascals out'. In prosperous times, however, incumbent politicians receive increased electoral support. The incumbency-oriented thesis stresses the possibilities of both reward and punishment. Incumbency in government can thus be either an asset or a liability. The policy-oriented hypothesis, by comparison, predicts that - everything else being equal - voters concerned about unemployment will prefer candidates from the Democratic Party. Thus, irrespective of incumbency, Democratic candidateswill voters hit or threatened by unemployment.

A policy-oriented hypothesis regarding the electoral effects of unemployment been defended by D. Roderick Kiewiet (1981). In a study of US presidential and congressional elections from 1956 to 1978, Kiewiet concludes that voters concerned about unemployment tend to support Democratic candidates whether or not a Democrat is already in a position of incumbency. 'Voters who had personally experienced unemployment gave a modest but consistent boost to Democratic candidates in virtually every election' (Kiewiet 1981, 459). However, when the focus is on the electoral effects of economic issues on a broader base, the conclusion is less clear-cut (Kiewiet 1983). When one, in addition to unemployment, includes a range of other issues, such as inflation, taxation, government spending and family finances, a substantial amount of empirical evidence can be found to support both the policy thesis and the incumbency thesis. However, the incumbency thesis seems the more important one for voter responses (Kiewiet 1983, 136).

Another dimension in this debate is the unit of reference: Do economic assessments refer to the individual or to the nation! An electoral decision based on individual economic assessments has been called 'pocketbook voting '. "Sociotropic voting", on the other hand, takes as its point of reference the financial situation of the nation as a whole (Kinder & Kiewiet 1979). Which is more important for voting - the assessment of one's personal financial situation, or the assessment of the nation's economy? On the basis of empirical evidence from the USA, sociotropic voting seems to be more widespread than pocketbook voting (Kiewit 1983). A Norwegian study shows a somewhat mixed picture, though the conclusion clearly leans more towards sociotropic than pocketbook voting (Aardal & Listhaug 1986).

In the following analysis the unit of reference is the unemployed person. The question is: Do the jobless have a deviant voting pattern compared with those in work? If being unemployed is decisive in regard to the party one votes for, then we are dealing with a kind of pocketbook voting. When one deals with national unemployment rates, however, the nation is the

unit of reference. Voters in work who are concerned about unemployment in a national context are thus committed to national interests rather than personal ones. In this case we are dealing with sociotropic voting. The dividing line between sociotropic voting and pocketbook voting - readily associated with altruistic versus egoistic concerns - may be analytically clear-cut. However, in real life the division is blurred and thus difficult to handle empirically.

Irrespective of sociotropic or pocketbook voting, soaring unemployment rates might be expected to harm the government in elections, since not just the jobless, but also people in work who are concerned about the national economy may vote for the opposition. In fact the European experience from the 1980s has run contrary to this suggestion.

The incumbency thesis and the policy thesis are by nature ahistorical. This shortcoming becomes particularly clear when the focus of the analysis extends over a longer period of years, such as in this case, where the impact of unemployment on party choice will be tested from the 1930s to the 1980s. In such a context several problems have to be addressed. The social composition of the jobless may have changed, for example, and a potential transformation of the pattern between unemployment and party choice may be due to the social background of the jobless operating as a prior variable. This prospect gives rise to a third hypothesis, the social composition

A fourth hypothesis may also be formulated - a generation thesis - which uses historical generation as a variable. This thesis takes into account the fact that the rate of unemployment, as well as the political tools perceived appropriate to boost employment, varies over time. In short, the historical context changes. This fact is ignored by the policy oriented thesis. To be more precise, the policy thesis points to the fact that a specific party has a political capital of trust concerning unemployment. This capital, however, is acquired in a certain historic phase and can be wasted in another phase. The political realities are always in a process of change, but cognition of the realities can often lag behind. According to the generation approach the reason is that basic political values and political outlook are formed during an early phase in the life-course. Thus, the glasses which filter political realities are often coloured by the realities of a passed society.

In at least two ways, a generation approach is adequate in a study of unemployment and party choice. First, unemployment affects various age groups unequally. In periods with high unemployment, young people with no work experience seem to have special problems finding a job. Second, youth unemployment is of particular interest when studying the impact this has on party choice. It is during the early years that party preference is established; though this is not fixed for the rest of life, this period of political socialization is still crucial for the establishment of sympathies and

antipathies towards the various political parties. The generation thesis, in sum, is based on two premises: (i) the political context varies from one period to another; and (ii) the most receptive period for impression is during the adolescence.

We thus have four hypotheses regarding the relationship between unemployment party choice: (i) the incumbency thesis, (ii) the policy thesis, (iii) the social composition thesis, (iv) the generation thesis. These alternatives will be tested with data from the Norwegian election surveys in the order mentioned above.

Incumbency Thesis Versus Policy Thesis

The Norwegian Labour Party can be compared with the US Democratic Party as the seemingly most trustworthy party for the jobless. Can we therefore also in Norway observe that Labour gains disproportional support from the unemployed and those concerned about unemployment, irrespective what party is in government? The Norwegian Election Studies make it possible to analyse the relationship between unemployment and party choice on the individual level. The 1957,1969, 1985 and 1989 election studies are used here, all of which have questions about experience of unemployment.¹ All of the questions are retrospective. In the 1957 election survey the question concerned experiences from 'before World War ll', whereas the 1969 election survey concerned the period 'after World War IF, i.e. from 1945 to 1969. The election study from 1985 covers the early 1980s ('during the past four years', i.e. 1981-85) and the study from 1989 the late 1980s ('during the past four years', i.e. 1985-89). One problem attached to such retrospective questions is the possibility of memory lapse, especially when the period in question was a long time ago, as in the 1957 and 1969 surveys. On the other hand, being jobless may have been such a traumatic experience that it is not easily forgotten.

There are no fundamental differences between the 1957 and 1969 questions. ask about both the respondent's own experience of unemployment experiences in his or her immediate family. They have in common, in other words, the somewhat vague term 'immediate family' and hence whatever variations that may ensue as a result of a small or large 'family'. In 1985 and 1989, by comparison, the relevant question related to employment experiences of either the respondent or the respondent's household. The term 'household' is less ambiguous and generally more limited than 'immediate family'.² In order to operate with time-series from 1957 to 1989 given these differences in question formulation, two categories are distinguished: (i) those who either themselves or in the immediate family/household have experienced unemployment and (ii) those without such experience.

Concerning the incumbency-oriented thesis, the crucial question is which government was in position during the various periods of unemployment. As some of the periods in question had more than one government, we shall characterize each one by the government which was in office longest and thus most readily associated with responsibility for the situation. There are four relevant periods for this study: the first, before 1940, is mainly a period with a non-socialist government;³ the second, between 1945 and 1969, is predominantly a Labour period; in the third, from 1981 to 1985, a non-socialist government was in power; and the fourth, from 1985 to 1989, was predominantly a Labour period.

Table 1 presents party preferences among those with and without experience unemployment in the various periods. The parties in the centre of the political spectrum - Liberal (V), the Agrarian Party/Centre Party (Sp), Christian People's Party (KrF), and the New Liberal Party (DLF) - are treated as one party bloc, the middle parties. Added are data from 1990, based on ten aggregated surveys conducted by MMI (Markeds- og Mediainstituttet, with a question about hypothetical voting in a parliamentary where those registered as unemployed are recorded as a separate category.

Does the pattern observed in Table 1 support the incumbency-oriented thesis or the policy-oriented thesis? The data clearly show the marked impact of unemployment before World War II and in the subsequent years (1945-69). Three-quarters of those with experience of joblessness - either for themselves or in their immediate family - supported Labour or parties to the left of Labour, as against somewhat less than half among those who had not experienced unemployment. This disproportionately high support for the Left took place at the expense of support for both the Conservatives and the Middle parties. It provides backing for the policy-oriented thesis, as the government in power during these periods varied. However, in the 1985 election the connection between unemployment and high support for Labour/low support for the Conservatives weakened, and by the 1989 election - and in the surveys from 1990 - it had disappeared for Labour altogether, and nearly so for the Conservatives. Although the Conservative Party still has somewhat greater support among those who have not experienced the difference is not statistically significant in either 1985 or 1989.

From the 1960s to the 1990s the parties on the outer flanks, both on the Left (the Socialist Left - SV) and on the Right (the neo-liberal Progress Party), increased their strength in general, and among those who have experienced unemployment in particular. In the 1989 election these two parties both established a position above 10 percent. In this election, as well as in the previous one in 1985, these parties obtained disproportionately high support among the unemployed.

Can the change in pattern during the 1980s be interpreted as support for the policy or incumbency thesis? The propensity of the unemployed to support Labour has gradually diminished in the elections from 1957 to 1989. Prior to the 1989 election, those with an experience of joblessness tended to give disproportionate support to Labour irrespective of whether Labour was in power or not. The policy thesis, in short, receives empirical support here. In the 1985 election, however, the empirical base for the policy-oriented thesis was weakened, and the 1989 election represents a definite break. For the 1989 election the policy-oriented thesis regarding Labour can be rejected.

But does the 1989 election support the incumbency-oriented thesis? The pattern of voting among those who have experienced unemployment does differ somewhat from those without such experience. The jobless seem to have an affinity for the Progress Party and the Socialist Left. In a way, both these parties enjoy one privilege in common: they have always been in opposition without governmental responsibility. They have never ruled and thus never erred. Thus, part of the explanation behind the apparent attraction these two parties have for the unemployed may be connected to this factor and as such provides empirical support for the incumbencyoriented The most striking feature in Table 1 is, nonetheless, the overall disappearance of a strong relationship between unemployment and party choice.

Another perspective on the issue of policy-versus incumbency-oriented voting is provided by whether voters believe that unemployment at any time would be worse if another government (either non-socialist or Labour) had been in power? This perspective includes sociotropic voting - so far disregarded, i.e. people in work who react against unemployment by voting for the opposition. In every election survey in the period from 1965 to 1989, the Norwegian Election Studies have posed a question relevant for such a test.⁴ The proportion of voters who consider that the present government has done a better job than any aspiring government could have managed has thereby been identified. From this perspective, if confidence in a Labour government is systematically higher than in a non-socialist government, then the policy-oriented thesis is supported. If, on the other hand, confidence is systematically higher for an opposition government, then the incumbency-oriented thesis would be supported.

The curve in Figure 1 indicates the share of voters with the most confidence in the current government (marked on the curve) regarding the unemployment problem. There is support for the policy-oriented thesis in the two pronounced drops in confidence following the governmental shift from Labour to non-socialist governments (1965 to 1969 and 1981 to 1985). However, neither of the two insignificant increases from a non-socialist government to Labour (1969 to 1977 and 1985 to 1989) gives the same

clear-cut evidence. Thus, the empirical picture seems somewhat ambiguous. The 1985 to 1989 elections are particularly important in this regard, as these are two elections where unemployment appears as a salient issue. Compared with the 19605, trust in government has now declined substantially,the between a non-socialist government and a Labour government being insignificant. It is reasonable to suggest, therefore, that Figure 1 and Table 1 present the same conclusion: the voters' trust in Labour as a party capable of conducting an employment policy significantly better than an alternative government has dissipated.

The Social Composition Thesis

An alternative explanation for both the decreasing support for Labour among the jobless in particular and the disappearance of the strong correlationbetween choice and joblessness more generally is a change in the social composition among the unemployed. In the 19305, the working class was perhaps especially hard hit by unemployment, in contrast to the situation in the 1980s, when more of a cross-section of the working populationseems

lationseemsto have been affected. This would mean that the observed linkage between unemployment and party choice can be attributed to the social background of the unemployed. In attempting to test this assertion empirically, it is difficult to control for 'occupation' among the unemployed. But since we are operating with retrospective questions, most of those who have experienced unemployment (either themselves or in their immediate family/house hold) are registered by occupation.

There is a further problem, however. As occupation and unemployment are not always measured at the same point in time, occupational category may have changed. To guard against this pitfall, the occupation composition among those with and without experience of unemployment (white- and blue-collar workers, self-employed, farmers/fishermen) has been checked. The general pattern remains the same in the distributions in 1957, 1969, 1985 and 1989, but the predominance of blue-collar workers among the jobless has weakened from 1957 to 1989.⁵. In other words, unemployment seems to strike with a somewhat less distinct profile in the 1980s than in the 19305. If it was possible to control accurately for social background measured by occupation, the trend in the data - from strong to weak, or non-existent, correlation between party choice and unemployment - would certainly be softened somewhat, but the main pattern would, as far as our data can be analysed, remain basically firm.

An alternative test of the influence of social composition among the jobless is to take blue-collar workers and divide them according to unemployment. blue-collar workers with experience of joblessness vote Labour more often than those without such experience? Figure 2 shows that this was the case up to the 1985 election, when the two curves converge and cross. In fact, in the 1989 election blue-collar workers with no experience of unemployment voted Labour more often than those who did have some experience with unemployment. The message from Figure 2 can thus be interpreted as a rejection of the social composition thesis, i.e. that the end of the disproportionate support for Labour among the unemployed was caused by a stronger representation of the middle class among the jobless. Hence, both the social composition and policy-oriented thesis lack enduring support and must be rejected. The incumbency thesis, however, has received some tentative support. This is the status before the final step, the test of the generation thesis.

The Generation Thesis

In order to test the generation thesis it is first necessary to review the
variation in unemployment during various historical epochs. Second, on
the basis of this variation the population is divided into various generations

according to unemployment rates in order to investigate the impact of unemployment on party choice in various generations. Thereafter, the two generations hardest hit by unemployment - the depression generation and the second post-war generation - are compared. As the impact of unemployment on party choice is quite different between these two generationsthe 'why' is addressed.

Variation in Unemployment Rate

The beginning of this century up until 1920 was a period of prosperity and low unemployment in Norway, culminating in a boom period from 1917 to 1918. The 19205, however, saw the onset of a period characterized by mass unemployment and strikes among manual workers. Then in the early 19305, an explosion in unemployment rates occurred in Norway as well as throughout most of the world. Unemployment figures peaked in 1933, when one-third of all trade union members in Norway were registered as unemployed.⁶

In the 1933 election, Labour was on the offensive against its main foe,
the Conservative Party, a parliamentary supporter party for the minority

government derived from the Agrarian Party (later the Centre Party). According to Labour, unemployment was the responsibility of the nonsocialistgovernment. initiative and the market economy had proven a failure. The capitalistic system, which at one moment would hire people, would fire them the next. Now the time was ripe for fundamental change. The government needed to play an active role both as a regulator of the economy and as an independent economic actor who raised large state loans.⁷ Labour's slogan from 1933 was 'everyone at work'; the solution was an expansive economic policy characterized by increasing state intervention and funding. Providing work for the unemployed was expected to have positive and widespread economic consequences. Spending power would increase as more people earned money. In fact, Labour's policy was formed along lines advocated by Key nes, even though the name was not mentioned.⁸

The Conservative camp rejected the slogan 'everyone at work' as misleading, instead that the result would be 'everyone on the dole'.⁹ Labour's expansive policy, it was argued, would lead to a higher level of taxes with devastating burdens on free enterprise, which in fact held the key to the solution of the problem. The correct strategy would be to cut taxes in order to provide better conditions for free enterprise so that it could prosper and hire instead of fire workers. The conflict centred around the classical dividing line between Left and Right: the necessity of regulating the economy versus the demand for a free market economy.

After the 1933 election, unemployment decreased in Norway and in 1935, supported by the Agrarian Party, the Labour Party came into office. In light of these and subsequent events, it was easy for people to associate Labour with being the party which had known the best remedy for economic recovery and which possessed the capacity to end mass unemployment. The massive support for Labour among those who had experienced joblessness the 1930s is undoubtedly related to the political situation which existed during the Great Depression when a non-socialist government was in office. In addition, the long post-World War II period, characterized by full employment up through the 19705, was also a period in which Labour was in power most of the time. This historical context enabled Labour to build up a reservoir of trust with respect to unemployment, upon which the policy-oriented thesis is based.

This long period of low unemployment, however, came to an end. Figure 3 shows the development in the rate of unemployment from 1945 to 1990.¹⁰ The curve indicates a period of stability from 1945 to 1980, with some seasonal variations. But the 1980s mark the start of a new period with a higher level of unemployment. During the early 1980s there was an upsurge, followed by a temporary fall in the mid-1980s. Then at the end of the 1980s and onset of the 19905, unemployment levels climbed past the 100000

mark, or 4.9 percent of the labour force. Since that time unemployment has continued to climb, and as of 1992 about 6 percent of the labour force is unemployed (those temporarily engaged in jobs initiated by official labour market authorities are not included).

During the 1980s, in short, joblessness again surfaced as a serious societal problem in Norway. In 1986 Labour had succeeded the non-socialist coalition headed by the Conservative Party leader Kåre Willoch. Under the new Labour government, unemployment had escalated, but was to a certain degree accepted - on the understanding that this period of high unemployment would be short. Labour had, according to its own version, started the recovery from the devastating effects brought about by the policies of the previous non-socialist government. In this temporary situation, was seen as a necessary evil. But by 1989 unemployment had for the first time surpassed the level from 1933. And in contrast to the 1933 election, the party now had no clear and optimistic plan for solving the problem. As the 1989 election campaign opened, Labour was in power and was thus seen as responsible for the high employment rates. Historically, this meant a break with the past.

To the voter, the well-established image of the Labour Party as the guarantor of full employment could easily appear crushed. As we have seen, this suggestion is empirically substantiated (see Table 1 and Figure 1). The empirical support for the policy-thesis has eroded.

By the late 1980s the social context had also changed substantially. The

heyday of industrial society had passed and been replaced by what is often labelled as the 'postindustrial society'. Labour and the Conservatives were no longer main contestants with highly divergent economic policies and prescriptions. It is not surprising, then, that the impact of unemployment on party choice may differ from 1933 to 1989. Indeed, the social position of the jobless is also different today. The establishment of the welfare state has given the unemployed a more secure material base, thereby weakening or cancelling the link between joblessness and material need. On the other hand, such basic material security has not removed the social problems attached to unemployment. In addition, the relationship between party choice and unemployment has not been completely dissolved. The two parties on the outer flanks - the Socialist Left and the Progress Party - receive disproportionate support among the jobless, an observation we have attributed to the incumbency thesis.

Unemployment and Generation

The large variation in the rate of unemployment - as well as the shift in what is regarded as an appropriate political tool in the fight against unemployment - creates a good position for the formation and demarcation of different generations. Not surprisingly, in the sociological literature on generations frequent mention is made of the 'depression generation', i.e. those who had their formative years during the Great Depression of the 19305. If the formative years coincide with a turbulent period, characterized for example by recession and unemployment, this 'Zeitgeist' may colour one's political outlook. Receptivity to impressions during adolescence is a central premise in the theory of political generations. Another central element in the generation approach is - as earlier mentioned - that during an early phase in an individual's life-course, a set of deep-seated values is established. Basic values are formed early in life, partly because young people are especially sensitive to new impressions - precisely because they lack experience. They meet the world with what Mannheim calls 'fresh contact' or a 'fresh eye'.¹¹ The formative values reflect impressions and events, forming a 'natural view of the world' which functions as a point of reference through which later experiences are related and interpreted. New information is assimilated or rejected according to this 'natural view of the world.'

US political scientists have shown that the Great Depression did in fact have an impact on party choice. An analysis of the 1952 and 1956 presidentialelection Campbell and his colleagues (1964) uses the term the 'depression generation'. The cohorts that grew into adulthood during the Great Depression and the early days of the New Deal era in the USA established a seemingly lasting loyalty to the Democratic Party.¹² In retrospect,we

spect,wecan recognize that the high level of unemployment was important in forming the depression generation in two ways. First, many of this generation experienced for themselves the negative effects of unemployment.Second, - its causes and consequences - during these years was a highly debated topic, an issue of first order. The pros and cons brought up in the course of these discussions provided arguments for supporting the various party alternatives.

The demarcation of a political generation is dependent on important historical events rather than on a fixed interval of years. If nothing changes in the societal framework, if there is a traditional society which rejects or suppresses every new trend, then it is impossible to distinguish between different generations. For a new generation to be moulded, there have to be important events which give a specific historical-social consciousness to certain cohorts. A political generation is, as a group, exposed to the same impressions or events: these can last for many years, or a shorter interval. Each generation receives a distinctive imprint from the social and political events of its youth.

A society in transition and confronted with fundamental changes meets the conditions for the appearance of differing political generations. Thus, partly on the basis of variations in the rate of unemployment, Norway's population may be divided into five political generations. The two oldest groups are commonly labelled the 1905 generation and the depression generation respectively. Both experienced the recession in the 19305, but these hard times hit them at different stages of their life-course and have thus had unequal effects. The 1905 generation experienced the critical years - their formative years - in a period with low unemployment, the years before the Great Depression. The label '1905 generation' refers to the most important event in Norway at the beginning of the century: not World War I - during which Norway remained neutral - but the dissolution of the union with Sweden, effective as of 1905. The depression generation, by contrast, was socialized into a society characterized by a high rate of unemployment. Consequently, we can expect unemployment to have made an imprint on their political orientation.

Those born after World War II comprise two additional groups, the first post-war generation (the '1968' generation) and the second post-war generation. The first post-war generation had their adolescence in a newly established affluent society, marked by the belief that progress was the norm. The period was characterized by full employment, so young people could choose their education without anxious consideration for demands from the labour market. The second post-war generation acquired the right to vote in the late 1970s and the early 1980s. This was concomitant with the swing to the right, to which in fact the second post-war generation contributed. During their adolescence the second post-war generation

presumably took the affluent society for granted, just as the first post-war generation had taken the societal progress for granted. This generation was maturing into adulthood in the mid-19705, when international economic trends shifted. The oil crisis of 1973 marked a turning point, followed by persistent and long-lasting unemployment trends. However, Norway as an oil-producing country initially managed to protect itself against this international trend, and sustained during the 1970s a situation of almost full employment. But with the 1980s came a break with this period of full employment, and joblessness appeared as a seemingly ineradicable evil. Thus the second post-war generation found itself confronted with a labour market which had a limited demand for labour just at the moment when this generation was finishing its years of schooling.

Finally, immediately prior to the two post-war generations, we find the rebuilding generation, i.e. the cohorts who reached adulthood just after World War II in a period when ideologies were declared 'dead'. The rebuilding generation had their formative years in a period marked by rationing and shortage of goods, but with 'everyone at work'.

The exact demarcation of the birth-years of the cohorts comprising the various generations is ultimately based on judgement. Although the dividing may thus be set somewhat arbitrarily, certain cohorts in every generation, it can be argued, represent 'core troops' which exemplify the most typical features of the generation. In short, the choice of division points between the generations in terms of birth-year can be debated, but the dividing lines appear to be valid, at least with respect to the core troops (cf. Table 2).

The four Norwegian Election Studies give a rough measure of unemployment generations and variations over time. As noted previously, the wording of the questions differs somewhat, making it impossible to register exact levels. Inasmuch as our interest is concentrated on the impact of unemployment, however, and not unemployment as such, this deficiency is not of crucial importance.

Table 2 indicates experience of unemployment at various times by generation.The partly confirm what we may reasonably expect from more reliable statistical sources. Three out of four distributions in Table 2 reveal a generation profile. That this is not the case for the 1969 distribution indicates that unemployment in the post-war era (1945-69) was a marginal problem which often lasted only a short period. When unemployment increases, then those young people looking for their first job are especially vulnerable. Consequently, before World War II it was the depression generation who reported the highest level of unemployment experiences, whereas during the 1980s the second post-war generation held the number one position in the ranking of high unemployment. During the 1980s the lowest unemployment was registered by the depression generation - which

must be related to the fact that many of this generation had by then retired. In studying the impact of unemployment on party choice, these two generations - the depression generation and the second post-war generation - are of particular interest since they have been especially hard hit by unemployment.

Let us take a look at the impact of unemployment on party choice in each of the various generations. For simplicity we concentrate on Labour and the Conservatives, the two main parties. In Table 3 it is shown that the impact of the Great Depression was especially clear-cut among those who experienced unemployment in those days: in this social group there is a pronounced tendency to support the Labour Party and not the Conservatives. 1957 election study, for example, shows that the Conservatives' position in the depression generation was almost exclusively due to the scant support from those with experience of unemployment before World War II; of these individuals, 5 percent voted for the Conservatives 78 percent for Labour (see Table 3).

In Table 1, a systematic pattern in the 1957, 1969 and 1985 elections is observed with Labour gaining disproportionately many voters from those who had experienced joblessness, and the Conservatives disproportionately few. This pattern is the same when we control for generation, at least until more recent times. Concerning the 1989 election, however, Table 3 clearly

shows that Labour no longer has this strong position among the jobless in
the two post-war generations.

The Depression Generation versus the Second Post-war Generation

Table 4 compares the relationship between joblessness and party choice in the depression generation and the second post-war generation. In the 1980s Labour totally lost its former strong appeal to the unemployed. In fact, the new Progress Party was the party that attracted most voters from among those who had experienced unemployment in the second post-war generation. the empirical base for this group is somewhat meagre (109 respondents) this position for the Progress Party will not be stressed unduly, but more generally the Progress Party seems to appeal to the jobless from younger age brackets. Consequently, our suspicion that age or generation was operating as a prior variable with the capacity to cancel the relationship between unemployment and party choice can for the moment be rejected.

Discussion and Conclusion

So far the disproportionate support for the Progress Party and the Socialist
Left has been interpreted in light of the incumbency-oriented thesis. Other
possible interpretations ought to be mentioned, however. In a review

article about right-wing extremism in post-war Europe, Klaus von Beyme differentiates between various waves of the development of right-wing extremist parties. He comments: 'A third phase of right-wing extremism was caused by unemployment and xenophobia at the end of a long prosperousperiod' 1988, 11). However, von Beyme does not discuss why this relationship between unemployment and the right-wing extremism party has arisen. There is, of course, a tradition of linking unemployment with political extremism. Extremists appeal to discontented voters. In this connection we could briefly mention that in Sweden, too, the recently launched New Democracy Party (a sister party to Norway's Progress Party) in its first election in 1991 gained strong support among the jobless.¹³

New Democracy Party as well as the Progress Party ardently oppose state-interventionist policy in the economy. To boost employment, they defend a market policy similar to the position held by the Conservatives in the 19305. This political programme for employment apparently attracted voters to the Progress Party in the 1980s. However, only half a century ago, this same programme deterred voters from supporting the Conservatives.This offers a platform for an alternative interpretation. Support for state interventionism in the economy has declined. In the 1969 and 1985 election study, respondents were asked whether they agreed or disagreed with the statement: 'If the government gets more influence over the operationsof banks and industrial enterprises, full employment will be more easily ensured'. Responses to this question are presented in Table 5, where 'experience of unemployment' is also introduced as a control variable. Two conclusions can be drawn from the results displayed: (i) belief in the necessity of state regulation in the economy has fallen from 1969 to 1985;

and even more important, (ii) belief in state intervention is, in 1985, independent of experience of unemployment. This was not the case in 1969, when those who had experienced being jobless were significantly more supportive towards state regulation of the economy. The majority of those who in the 1980s experienced unemployment did not regard governmental regulation of the market forces as a suitable instrument for ensuring full employment. When the large majority of voters feel that the politicians lack the tools to solve the unemployment problem, then unemployment no longer seems a salient issue of consequence for the election strength of the various parties.

The impact of unemployment on party choice is obviously related to the political debate. With good reasons Visser & Wijnhoven (1990) underscore ideological discourse in their discussion of why Conservative governments have not been punished electorally for mass unemployment. Conservatives have criticized a state-interventionist policy and defended a marketoriented In this connection, having lower inflation has been emphasized, and rising unemployment has been regarded as a necessary condition for reaching that goal.

Further, it has been stressed that, in the long run, profitable capital accumulation rather than state intervention in the economy has the potential raise employment. The point is that a certain degree of unemployment perceived as unavoidable in the market economy, and state intervention in the economy would only make bad matters worse. Consequently, responsibility for full employment is placed on the politicians. Once the question is reduced to how the market is functioning, then politicians can be easily excused.

In Norway, the neoliberal Progress Party appears as the most ardent defender of market economy and offers a prescription on how to solve unemployment: more market and less state will speed up the economy. The party's appeal among the jobless can thus make sense. However, the Progress Party also wishes to restrict financial benefits to the unemployed. Is then a jobless person who votes for the Progress Party an example of the irrational voter? Not necessarily. The unemployed of the depression generation associate the expansive state - in the role as a forceful actor in the economy - with full employment and recovery from the Great Depression. contrast, the young unemployed in the 1980s readily perceive the expansive state as a troublemaker. What was medicine for one generation becomes poison for another.

Much as was the case in the US, unemployment during the Great Depression in Norway had a clear-cut impact on party choice. By the 1980s, however, the correlation between unemployment and party choice had nearly disappeared. Voters' trust in Labour as guarantor of full employment has gradually vanished. Consequently; the policy-oriented thesis has been

undermined. However, only to a restricted degree has the rejection of this thesis been replaced by the incumbency thesis: the two parties which gain a somewhat disproportionate share of voters from the unemployed, the Progress Party and the Socialist Left, have never ruled and thus enjoy a privileged position vis-å-vis the present and former incumbent parties.

However, there are alternative interpretations in addition to the incumbency and the policy thesis. A third thesis - the social composition thesis - has been rejected. This thesis asserts that the end of the strong support for Labour among the jobless is due to a change in the social composition of the unemployed with a sharp decline in blue-collar workers.

A fourth thesis, the generation thesis, is based on the assumption that the most receptive period is in an early phase in the life-course. In this period, basic values are formed and party preferences are established, if not fixed, for ever. The generation approach, however, is irrelevant in a traditional society which rejects or suppresses every new trend. Changes must occur in the societal framework. Concerning unemployment, there have indeed been changes in Norway. Unemployment rates have varied, as has what is regarded as appropriate political solutions to boost employment. This is an obvious reason for the empirical support which the generation thesis receives. The impact of unemployment on party choice does vary according to generation. The pattern between joblessness and party choice is indeed different in the two generations hardest hit by unemployment, the depression generation and the second post-war generation. For the depression generation, support for Labour among those hit by unemployment the 1930s was massive. That is not the case for those in the second post-war generation who experienced unemployment in the 1980s. One reason is that in the post-industrial society Labour and the Conservatives no longer the main contestants with respect to labour market policy. The neo-liberalist alternative represents a challenge to old solutions with its own recipe for solving the unemployment problem - more market, less state. In the late 1980s this message would seem to have an appeal to the jobless in the younger age brackets.

Some reservations to this conclusion should also be mentioned. We have tested electoral impact by mapping party votes and have disregarded nonvoting.Various investigations have, however, substantiated that the jobless are more likely to be absentees at the polls (Banks & Ullah 1987, Lipset 1981). This is in line with an observation that unemployed persons can become apathetic and feel disintegrated from social networks in working life. Non-voting is of course also an electoral impact of special importance if the absentees hide party preferences which deviate from those of the jobless who vote. Unfortunately, we lack empirical data to test such a proposition. In addition, there is the possibility that unemploymentalso a negative impact on voting turnout among people in

work: if the politicians cannot manage to solve such a serious problem,
why bother to vote?

Finally, a methodological reservation should be mentioned. Cross-sectional have been used from five different time-points. With longitudinal data - or a panel study - we could have followed the voter from one election to the next and distinguished between party choice before and after unemployment. In this way we could have differentiated between the shortterm long-term unemployed.¹⁴

Despite such deficiencies in our data, there is good reason to trust the main conclusion. The developmental trend seems indeed clear-cut, and another recent study of five European countries confirms that the Norwegian is not unique (Visser & Wijnhoven 1990). Perhaps the explanation is simple enough: no party has managed to convince the jobless that it holds the key to higher employment. A feeling of inevitability accompanies today's mass unemployment. Indeed, this gives no optimistic prognosis for the years to come. Concerning the future, however, a caveat is in order: if the unemployment problem increases and the long-term jobless accumulate, conclusion based on elections in the post-war period up to 1989 can easily become invalid.

---

NOTES

1. The 1957 question: 'Before World War 11, it was, as you know, difficult to get a job. Can you remember if your father, yourself or any of your immediate family were unemployed for any length of time in this period?' The 1969 question: 'Have you or any of your immediate family been unemployed for more than 2-3 months after World War II?' (i) Yes, myself; (ii) Yes, immediate family; (iii) No; (iv) Cannot remember. The 1985 and 1989 question: 'Have you or any in your household been unemployed or experienced major problems in getting a job during the last four years?' (i) Yes (myself or my household); (ii) Problems with getting a job; (iii) No; (iv) Retired, not working.

2. Another factor concerns age. The lower age-limit for respondents in 1985 and 1989 was 18 years of age. This means that four years earlier the youngest ones were only 14 years of age, with no possibilities for personal experiences from unemployment. But as the question also concerned unemployment in the household, it is not necessary to include only those old enough to have a job. In the 1957 and 1969 election study, respondents were 20 years of age or older. As mentioned, in 1957 the question related to experiences from 1940 and earlier. One reasonable demand is that the respondent must be old enough consciously to register unemployment. For this reason, we have chosen to exclude from further analysis those who were nine years of age or younger in 1940. For the 1957 study, in other words, only cohorts born in 1930 or earlier are the object of analysis.

---

2. Another factor concerns age. The lower age-limit for respondents in 1985 and 1989 was 18 years of age. This means that four years earlier the youngest ones were only 14 years of age, with no possibilities for personal experiences from unemployment. But as the question also concerned unemployment in the household, it is not necessary to include only those old enough to have a job. In the 1957 and 1969 election study, respondents were 20 years of age or older. As mentioned, in 1957 the question related to experiences from 1940 and earlier. One reasonable demand is that the respondent must be old enough consciously to register unemployment. For this reason, we have chosen to exclude from further analysis those who were nine years of age or younger in 1940. For the 1957 study, in other words, only cohorts born in 1930 or earlier are the object of analysis.

3. This was a period in which there was a non-socialist government, with two exceptions: a Labour Party government was in power for two weeks in 1928 and again from 1935 to 1940.

4. The question was: 'In recent years a Labour government (or a non-socialist government) been in power. Let us have a look at the policy which has been pursued in various fields. Suppose that we in recent years instead had a non-socialist government (or a Labour government): Do you believe that the unemployment rate would have been (i) higher; (ii) much the same; (iii) lower; (iv) don't know, impossible to answer?'

5. In 1957 the proportion of blue-collar workers among those with and without experience of unemployment was 61 percent {with) and 39 percent (without) respectively. The corresponding figures from the 1985 election study were 45 percent (with) versus 33 percent (without), and in 1989 40 percent (with) versus 31 percent (without).

6. This percentage is basedon figures from those tråde unions which had available statistics. These statistics are often regarded as the most reliable, but there are also pitfalls in the interpretation of such figures. Not every tråde union has an arrangement with unemployment insurance and is thereby forced to have files with reliable statistics. In addition, when the statistics are basedon tråde unionists, not only are employees without tråde union affiliation excluded, but also youth who for the first time apply for a job and have never been confronted with the possibility of joining a tråde union. Moreover, tråde union statistics provide only a partial indication of the total amount of unemployment. To get closer to the exact figures we have to rely on a variety of not altogether adequate material (cf. Bull 1979). One Norwegian statistician, for instance, estimated the number of unemployed in 1933 - defined as persons actively trying to find jobs without succeeding - to be 101 500 (Tuveng 1946, 86), or approximately percent of the labour force. (Basedon the corresponding figure in the 1930 census, 1 227284, weighted for the increase in the population from 1930 to 1933 (1.5 percent), it is possible to estimate the total working force in 1933 at 1245 659.) But even this figure seems to be misleadingly low. Among other things, those who were working reduced hours are excluded.

7. See for example editorial articles in the Oslo Labour daily Arbejderbladet, 3 and 14 October 1933.

8. However, two Norwegian economists did play a role, Ole Colbjørnsen and Axel Sømme. They presented a three-year plan which concluded that it was possible to create 150000 new jobs by an active economic policy that involved raising large governmental loans.

9. From the Oslo conservative daily. Aftenposten, 14 October 1933.

10. From 1945 we have reliable statistics on unemployment: official statistics have recorded everyone who is registered as jobless at labour exchange offices.

11. Which phase in the life-course is the most impressionable period in constituting a basis for one's formative years? Although Mannheim's essay is not empirically anchored, he did touch on the problem. The lower limit, the start of the receptive period when new impressions are easily assimilated, is by Mannheim stipulated to around 17 years of age. The upper limit, the time when the 'natural view of world' has become established in more or less final form, is in a footnote set to approximately 25 years of age. The reference is to a French linguist (A. Meillet), who has claimed that one's spoken language - pronunciation and vocabulary - ceases to change in a substantial way after the age of 25 (Mannheim 1980).

12. The loser was the Republican Party. With the New Deal era came a massive shift away from the Republican Party: 'The program of welfare legislation of the New Deal and the extraordinary personality of its major exponent, Franklin D. Roosevelt, brought about a profound realignment of party strength, which has endured in large part up to the present time' (Campbell et al. 1964, 276). More recent studies, like that by Glenn & Hefner (1972), provide further evidence on the lasting impact which the Great Depression and the New Deal had on party choice.

---

12. The loser was the Republican Party. With the New Deal era came a massive shift away from the Republican Party: 'The program of welfare legislation of the New Deal and the extraordinary personality of its major exponent, Franklin D. Roosevelt, brought about a profound realignment of party strength, which has endured in large part up to the present time' (Campbell et al. 1964, 276). More recent studies, like that by Glenn & Hefner (1972), provide further evidence on the lasting impact which the Great Depression and the New Deal had on party choice.

13. According to an Exit Poll conducted by Sveriges Radio, 13 percent among the jobless voted for New Democracy compared with 6 percent who voted for the party in the sample.

14. Denmark has experienced a longer period of large-scale unemployment than Norway. In the 19705, mass unemployment in Denmark seems to have insignificant impact on party choice. During the 1980s, as the long-term unemployed accumulated, they began to appear as a group with a rather distinct party profile, preferring parties of the outermost left wing (Goul Andersen 1989, 191; Svensson & Togeby 1991, 73).

ACKNOWLEDGEMENTS

Data sets from The Norwegian Election Studies 1957, 1969, 1985, 1989 used in this article were made available by the Norwegian Social Science Data Service (NSD). Professor Henry Valen has been in charge of the election studies. The 1957 and 1969 surveys were conducted by the Norwegian Gallup Institute and the 1985 and 1989 surveys by the Central Bureau of Statistics. Neither the Norwegian Gallup Institute, the Central Bureau of Statistics nor NSD are responsible for analyses or interpretations of the data. That responsibility rests entirely with the author.

REFERENCES

Andersen, J. Goul 1989. 'Social klasse og parti', in Elklit, J. & Tonsgaard, 0., eds., To
folketingsvalg. Vælgerholdninger og vælgeradfærd i 1987 og 1988. Århus: Politica.

Banks, M. H. & Ullah, P. 1987. Political Attitudes and Voting Among Unemployed and
Employed Youth', Journal of Adolescence 10, 201-216.

Beyme, K. von 1988. 'Right-Wing Extremism in Post-War Europe", West European Politics
11, 1-18.

Bull, E. 1979. 'Klassekamp og fellesskap 1920-1945', in Mykland, K., ed., Norges historie vol.
13. Oslo: Cappelen.

Campbell, A., Converse, P. E., Miller, W. E. & Stokes, D. E. 1964. The American Voter:
An Abridgement. New York: John Wiley & Sons.

Elder, G. H. 1974. Children of the Great Depression. Chicago: University of Chicago Press.

Glenn, N. G. & Hefner, T. 1972. 'Further Evidence on Aging and Party Identification', Public
Opinion Quarterly 36, 31-47.

Kiewiet, D. R. 1981. 'Policy-Oriented Voting in Responses to Economic Issues', The American
Political Science Review 75, 448-459.

Kiewiet, D. R. 1983. Macroeconomics & Micropolitics. The Electoral Effects of Economic
Issues. Chicago: University of Chicago Press.

Kinder, D. & Kiewiet, D. R. 1979. 'Economic Discontent and Political Behavior: the Role
of Personal Grievances and Collective Economic Judgements in Congressional Voting',
American Journal of Political Science 23, 495-527.

Lipset, S. M. 1981. Political Man. The Social Bases of Politics. Expanded Edition. Baltimore,
Maryland: Johns Hopkins University Press.

Mannheim, K. 1980. 'The Problem of Generation,' in Kecskemeti, P., ed.. Essays on the
Sociology of Knowledge. London: Routledge & Kegan Paul Ltd.

Schuman, H. & Scott, J. 1989. 'Generations and Collective Memories', American Sociological
Review 54, 359-381.

Svensson, P. & Togeby, L. 1991. Højrebølge? Århus: Politica.

Tuveng, M. 1946. 'Arbeidsløshet og beskjeftigelse i Norge før og under krigen'. Sak og
samfunn nr. 13. Bergen: Chr. Michelsen's Institute.

Visser, W. & Wijnhoven, R. 1990. 'Politics Do matter, but Does Unemployment? Party
Strategies, Ideological Discourse and Enduring Mass Unemployment', European Journal
of Political Research 18, 71-96.

Aardal, B. O. & Listhaug, O. 1986. Economic Factors and Voting Behaviour in Norway 1965-1985.
Institute for Social Research, Working Paper 1986:4.