The academic debate increasingly acknowledges different types of political participation, which are differentiated along several dimensions, such as the level of effort or duration of the participation (Theocharis and Van Deth 2016). Research has described some of the participatory activities—mostly but not exclusively afforded by online communication—as low in effort and civic skills they require from citizens and bearing low risks (e.g., Christensen 2011; Nanz et al. 2022). Another critical dimension differentiating political participation types is how strongly these activities target the political system. These two dimensions are interrelated: Activities closer to the political system (e.g., being a party member) may be perceived as ‘effortful’ by citizens compared to other, less effortful acts such as supporting a cause online or boycotting products (see Micheletti et al. 2011). This has important implications: If a political system relies on an active citizenry that engages in political activities related to the representative delegation of power (e.g., party membership or signing of petitions), it may suffer if people migrate towards activities that are more distant from the legitimation of power because they are easier to engage in (e.g., Christensen 2011; Harrison 2020; Van Deth 2014). However, this ease of participation has also been described as an entry point for less politically engaged citizens and could benefit the political system in the long run (Andersen et al. 2021; Copeland 2014; Stolle et al. 2005).

This study investigates individuals’ participation trajectories over time, i.e., the transformative effects of engagement in one type of participation on others. If we consider the political system as a starting point, it is possible that, over time, citizens’ participatory activities move closer to the political system or further away from it. These developments are usually termed upstream and downstream effects (Holbein and Rangel 2020). If citizens start by participating in less effortful activities distant from the political system (e.g., donating, volunteering, civic messaging) and, over time, engage in participation types more closely related to the political system, we would observe an upstream effect. The protest movement Fridays for Future is an example where, over time, digital, expressive modes of action led to a very explicit voicing of demands directly targeting the governments via demonstrations and personal meetings in many countries. A downstream effect, in contrast, can be detected when citizens participate in high-effort activities closely related to the political system but drift towards less demanding, non-political activities over time. For example, attending street demonstrations may be substituted with checking in to political events online. While some research has investigated the role of electoral turnout for future political activities (e.g., Green and Shachar 2000; Holbein and Rangel 2020) and the interplay of online and offline participation (e.g., Kim et al. 2017; Cantijoch et al. 2016) over time, research about developments in non-electoral participation is sparse. However, a closer investigation of different participation types is necessary to understand citizens’ entry points to participation trajectories.

The study disentangles political participation trajectories over time utilizing an exploratory panel survey (n = 3490) conducted in Denmark. First, it describes why citizens participate in different political activities and the potential consequences for society (Van Deth 2014). Second, the study introduces upstream and downstream effects (e.g., Holbein and Rangel 2020) to non-electoral participation and puts forward scenarios for why citizens may follow different participation trajectories over time. Third, the study tests intra-individual change in participation with the help of random intercept cross-lagged panel models. This is relevant to understand the drifts of participatory actions in democracy, as we argue that participation less able to address the political system directly may cause democratic frustration and weakens the political system in the long run.

Effort and target of political participation

As more ways to engage politically emerge, citizens face a ‘high-choice’ participation environment and apply new selection criteria. Van Deth (2014) differentiates between four types of political participation: Forms of engagement that (1) take place in the political sphere (PP I), (2) are targeted at the political sphere (PP II), (3) are targeted at community issues (PP III), and (4) are non-political but politically motivated (PP IV). This study does not aim to repeat the argumentation for this distinction (see Ohme et al. 2018; Theocharis 2015; Theocharis and Van Deth 2016 for an extensive discussion). Yet, it helps us arrive at an important observation: Activities that demand more resources from citizens are often more closely connected to the political sphere (PP II). At the same time, supposedly less effortful types can be found mostly among activities targeted at the community (PP III) or include non-political, politically motivated activities (PP IV).

Importantly, this differentiation is not the only way to conceptualize political participation (see Gibson and Cantijoch 2013; Gil de Zúñiga et al. 2014; Teorell et al. 2007; Valeriani and Vaccari 2016; Yamamoto et al. 2015). Moreover, significant strands of research have focused on political expressions as acts of political activity (Gil de Zúñiga et al. 2014; Valeriani and Vaccari 2016; Yamamoto et al. 2015). As shown in Fig. 1, PP IV in this study includes multiple acts of political expressions. Following the argumentation by Van Deth (2014) and Theocharis (2015), acts of expression can be understood as political participation if they are “used to express political views and mobilize others” (Theocharis et al. 2021, p. 34). We applied this conceptualization because it allows for connecting participation activities with the political system. Moreover, our study focuses on acts of participation within a democratic framework, i.e., under the assumption that acts of participation do not attack the political system in general. While this current phenomenon of ‘dark participation’ is very important (Inguanzo et al. 2022; Lutz and Hoffmann 2017), it was not the scope of this research and not as prominent during the time of data collection.

Fig. 1
figure 1

Conceptual model

Full size image

Previous research applying the conceptualization of Van Deth (2014) to empirical data has stressed that it should not be assumed that any type of participation is more important than the other as long as studies on the specific effectiveness of participation activities in achieving change are lacking (Theocharis 2015; Ohme et al. 2018). Moreover, the relevance of participation activities also depends on the type of democracy: Community participation may contribute to participatory democracy, online expression to a deliberative understanding of democracy, and voting specifically to an electoral democracy (e.g., Strömbäck 2005). What becomes apparent, however, is that different types of political participation are related to varying extents to the political system. Hence, the frequency with which citizens participate in different types of activities means that they more or less directly address the political system, and their concerns have different chances of being recognized by the feedback mechanisms of democracy.

Why do citizens choose different types of participation to bring their concerns forward? One strand of research introduces collective and selective incentives people may experience despite knowing that chances to reach the desired outcome by an individual action are low (Bäck et al. 2011; Aldrich 1993; Finkel and Muller 1998). A collective incentive would be to believe that a political action contributes to the desired (often public) good. In contrast, a selective incentive can be the individual enjoyment of such acts (ibid).

A second strand of research considers the necessary resources for political participation. The availability of sufficient resources, such as time and civic skills, has been described as one common predictor for political participation (Verba et al. 1995). Given that citizens have a certain ‘participation reservoir’ that does not increase just because choices increase, it is important to understand how they distribute their resources across different participation types. Therefore, this type of argumentation received much attention when the choice of participatory activities increased (e.g., Thorson 2015).

Some participatory activities require fewer resources from citizens, as they can be done either quickly, on the go, with low risk involved, or are even perceived as a hobby, for example, sharing information on social media, boycotting products, or urban gardening (see Micheletti et al. 2011; Nanz et al. 2022; Thorson 2015). These are contrasted by activities, such as demonstrating, contacting politicians, or joining a party, that require more effort, greater civic skills, more resources, bear more risk, and may, therefore, be perceived as more demanding by citizens. Potential risks include a confrontation with authorities, handling unfamiliar or turbulent situations like protests, or not having sufficient (civic) knowledge or arguments needed to engage with others (Cantijoch et al. 2016). However, empirical evidence on the effectiveness of different types of political participation is sparse; hence, evaluating an activity as ineffective just because it demands less from citizens seems unwarranted.

It is not the goal of this study to explain why people participate in specific activities. Rather, we bring forward the argument that, especially in a new participation environment, where many types of political participation involve new digital formats or are more societally accepted than earlier, a greater number of activities exists that involve less effort and risk than others. These activities—by definition and as explained by Van Deth (2014)—are less strongly related to the political system in which they occur.

Towards the background of low rates of political participation in Western societies (see Andersen et al. 2021; Macedo 2005; Putnam 2001; Schlozman et al. 2013), the observation that more demanding activities target the political system more directly than supposedly less effortful activities that are more distant to the system has two important implications: First, it can mean that a political system that requires an active citizenry may suffer if people start engaging in political activities more distant from it because they are easier than others and become more attractive over time. The political system may directly suffer if one activity is substituted by another, but also indirectly because activities that in their mode of action bypass established political decision-making processes and decision-makers can undermine trust and belief in the responsiveness of the political system (Parvin 2018). Second and in contrast to that, this ease of participation has been described as an entry point for less politically engaged citizens (Andersen et al. 2021; Copeland 2014; Stolle et al. 2005), which could benefit the political system in the long run, if starting to participate in less demanding activities would steer citizens closer to participatory acts that target the political system directly. To get a first descriptive idea of these developments, we ask:

RQ1

How frequently do people engage in three types of non-electoral political participation at different times?

Trajectories of political participation over time

Political participation is driven by a rich set of predictors such as, for example, citizens’ political interest, knowledge, and efficacy (see Andersen et al. 2020; Bimber et al. 2015; Jung et al. 2011). But—paraphrasing a statement from Holbein and Rangel (2020, p. 1198)—while much has been done to study what influences political participation, comparatively little has been done to explore what political participation influences. In our case, less is known about how political participation in one type influences participation in other types. This is especially true for whether prior political behavior can have a transformative impact on future political behavior. Hence, we do not study the societal outcomes of participation but focus on the overtime development of participation in different political activities.

Most work on longitudinal participation effects has been conducted on voting. Here, the effects of prior voting on future turnout are well documented (e.g., Green and Shachar 2000; Goldberg and Sciarini 2023; Holbein and Rangel 2020; Jessen et al. 2021; Smets and Hams 2013). Evidence is more mixed for the impact of voting on other types of participation (e.g., Bechtel et al. 2018) or individual attitudes and preferences, such as political interest, citizenship norms, and political efficacy (see Bhatti and Hansen 2012; Holbein and Rangel 2020; Jessen et al. 2021). Less, however, is known about inter-participatory transformative effects in non-electoral participation. So far, research focused on the interplay between online and offline types of participation (Cantijoch et al. 2016; Kim et al. 2017). While this distinction can be valuable, it seems less helpful to investigate how the transformative effects of participation relate to the political system because both online and offline types of participation can target different political spheres (Theocharis 2015).

We, therefore, focus on upstream and downstream transformative effects across three types of political participation that differ in effort and relation to the political system. Different theoretical arguments for upstream and downstream effects exist. Upstream effects from low- to high-effort types of participation are possible because participation types distant from the political system can increase political interest, knowledge, and efficacy (Andersen et al. 2021). Already in 1995, Verba, Schlozman, and Brady argued that “political interest and political efficacy, for example, certainly facilitate political activity, but activity presumably enhances interest and efficacy as well” (p. 271). In addition, through low-effort participation types, citizens may learn skills, feel empowered, and more secure to engage in more demanding activities in the future (Kim et al. 2017; see also Conroy et al. 2012; Livingstone et al. 2007).

Downstream effects can exist if there is a spillover from more to less demanding types of participation (Kim et al. 2017; Vissers and Stolle 2014). This means that people enrich their participation diet with less demanding types of political participation over time to achieve political goals at relatively low added costs, which would resemble a rich-get-richer paradigm. This may come at the expense of participation in more demanding types due to limited available resources in their ‘participation reservoir.’ A second reason can be ‘democratic frustration,’ “when a perceived democratic delivery deficit interacts with a strong democratic expectation or desire” (Harrison 2020, p. 1). If citizens are disappointed that high-demand activities do not lead to envisioned changes, they may withdraw from participation altogether. Less drastic outcomes are also possible: While it may be hard for citizens who show high dedication to political causes to withdraw entirely, they may pursue participation at lower costs and, therefore, engage in types more distant from the political system.

Empirical evidence for upstream and downstream effects in the non-electoral arena is still sparse. In a two-wave lagged dependent variable model, Cantijoch et al. (2016) find little indication that online modes of participation affect offline modes before and after an election. Kim et al. (2017) found such a gateway effect from online to offline modes in a two-wave panel study across 2 years, but only for 16-year-olds, while the opposite was true for 22-year-olds. Nanz et al. (2022), utilizing two-wave models in the context of the Austrian national election, found lagged effects of low on high-effort participation, which resembles evidence for an upstream rather than downstream effect. However, none of these studies involve more than two waves and investigate transformative effects for more than two types of political participation. Our study is one of the first to investigate transformative inter-participatory effects. Using a panel design, we test whether upstream and downstream effects exist between three types of political participation across three waves. Hence, we ask:

RQ2

What is the evidence for upstream and downstream effects between types of political participation over time?

Analytical strategy

To analyze whether political participation develops in an upstream or a downstream process, we focus on intra-individual change causally linked to earlier engagement with different types of political participation. There are several approaches to modeling intra-individual change (within effects). Models in the regression framework are mainly suited to associate within change with within change. For the purpose of this study, however, we need to identify causal links between the level of participation at one point in time and the level of participation of a different type at the next point in time, while also accounting for autocorrelation. Therefore, we decided to follow the random intercept cross-lagged panel models (RI-CLPM) suggested by Hamaker et al. (2015) using the Structural Equation Modeling framework.

In the RI-CLPM, latent factors are added, each predicting one type of political participation across different time points, thus capturing the stable differences between individuals. Using the remaining within individual variance in the structural part of the model, we have each type of political participation at time t predicted by political participation of the same type at t-1 (the autoregressive effects) and by the other types of political participation at t-1 (the cross-lagged effects). With the autoregressive effects, we control for previous deviations from the expected level of participation, leaving the cross-lagged effects as our main test of within individual upstream and downstream effects. Both cross-lagged upstream and downstream effects are added to the model, allowing us to test the upstream and downstream effects simultaneously. With three waves in the panel data, we can predict participation in waves 2 and 3. In our models, we will first test the effects between the waves separately (i.e., in the same model but freely estimating the same cross-lagged effect between the different waves). In a second step, we constrain the effects over time to be equal, allowing us to test whether the effects are consistent over time.

We rely on several measures typically used in comparable structural equation models to assess fit. We report the model chi-square (χ2), but given the large sample, the associated test of a fitting model will likely lead to a significant result (Kenny n.d.). In addition, we rely on the Non-normed Fit Index (i.e., the Tucker–Lewis Index TLI, which should score above .95 for a good fit) and the Root Mean Square Error of Approximation (RMSEA: \(\hat{\varepsilon }\), which should score below .05 for a good fit) and also include the close fit hypothesis PCLOSE, which should be non-significant. To compare the fit of nested models, one usually focuses on the chi-square difference test (χ2D). But again, due to sample size, this test will likely lead to significant results even when loss of fit is negligible. So, in addition, we focus on the RMSEA difference (\(\hat{\varepsilon }_{D}\)), which can be interpreted on the same metric as the original RMSEA (not to be confused with the RMSEA change [Δ\(\hat{\varepsilon }\)], see Savalei et al. 2021).

Method

Sample and field time

The study utilizes a five-wave online panel survey conducted among the Danish population in 2014/2015 (n = 3490), of which only the three non-election waves are used in the analysis. The sample was recruited in collaboration with the online polling company Epinion. 9125 respondents participated in the first wave. The second and fifth waves used in this study were fielded approximately 4 months apart from the respective previous wave (see Fig. 1 for field time). Wave 3 (27 May–15 June 2015) and 4 (19–26 June 2015) were fielded around the 2015 Danish national election and are not included in the analysis since they did not measure non-electoral participation. An election campaign between non-electoral waves can impact citizens’ behavior. Hence, we capture this impact under realistic, democratic conditions since election campaigns are part of democratic all-day life.

In total, 3490 respondents participated in all waves. The overall retention rate between waves was 81%. The average age was 55 years (SD = 17), 52% were female, 41% had finished primary, vocational, or high school, 37% a short- or medium-cycle higher education, and 21% a long-cycle higher education. These data closely resemble population data at the time. Still, due to panel attrition, the sample is slightly older and more educated than the general Danish population (for a detailed overview of the sampling procedure and attrition rates, see Andersen et al. 2021).

Measures

Political participation

The three types of political engagement were measured by asking respondents how often they had participated in various political activities. In Wave 1, we asked about activities in the last 12 months, while Wave 2 and Wave 5 asked about the last 4 months in between waves. We used a five-point scale ranging from never (0) to four times or more (4). The full list of items with descriptives can be found in Appendix Table 2. To operationalize the framework suggested by Van Deth (2014) and its extension by Theocharis (2015), 23 items were assigned to four different types of political participation (see Fig. 1). We included a similar number of digital and non-digital activities undertaken at the local and national level of government. The activities were selected from previous studies and surveys (Ekström and Östman 2015; Ekström et al. 2014; GLES 2013; Portney and O’Leary 2007; Stolle et al. 2005; Yndigegn and Levinsen 2015). Activities target the political sphere (PP II) if they refer directly to political institutions (e.g., visiting a politician, signing a petition) or happen in a state-guaranteed framework of protest (e.g., participating in a demonstration). Activities address issues at a community level (PP III) if they refer to a direct action with immediate outcomes on a local level (e.g., supporting a community’s crowdfunding project or volunteering in a local organization). For non-political but politically motivated activities (PP IV), pertinent considerations and the political purpose must be emphasized (e.g., boycotting products for political purposes or expressing an opinion on social media about a political issue). A confirmatory factor analysis was undertaken based on the data from the first survey wave and revealed a sufficient distinctiveness between the types of political participation (Ohme et al. 2018). Subsequently, a mean scale of the items per type of participation was created, resulting in the final measures used in this study. See Fig. 1 for the full conceptual model and specimens for each type of participation and Appendix Fig. 4 for confirmatory factor analysis results.

Results

Addressing RQ1, we see that people participate mostly in types of participation with lower effort distant from the political system (PP IV). Less often, people participate in activities targeting the local community (PP III) and the political system (PP II). Hence, we see apparent differences in the level of participation across the three types of participation. This finding remains stable over time. Only the change from w2 to w5 can be directly compared since the decrease from w1 to w2 can be explained by the time frames surveyed (last 12 months for w1, last 4 months for w2, and w5). Although some differences become visible in Fig. 2, overall, the development remains stable on an aggregate data level. However, these numbers do not tell us anything about the intra-individual changes between waves.

Fig. 2
figure 2

Frequency of types of political participation over time. The sum score frequency shows the mean across the different items for each participation type

Full size image

RQ2 asked about evidence for upstream and downstream effects on an individual level. In our initial model, we estimated autoregressive and cross-lagged effects between consecutive waves, and the variables at the same time point were allowed to covary. This model, however, did not fit the data sufficiently (see Model 1 of Table 1). In order to achieve fit, we included an effect from PP II in w1 to PP IV in w5 as well. The new model fits significantly better than the original one (χ2D(1) = 53.86, p < .001; \(\hat{\varepsilon }_{D}\) = .17) and has a sufficient fit (see Model 2 of Table 1). The full parameter estimates are shown in Model 2 of Appendix Table 3. Below, we illustrate these results in Fig. 3. The figure presents the upstream and downstream effects separately, which are modeled simultaneously.

Table 1 Fit indices for models to assess upstream and downstream effects
Full size table
Fig. 3
figure 3

Upstream and downstream effect of political participation over time. Upstream (a) and downstream (b) cross-lagged effects. Both figures represent the same model, but for reasons of clarity the upstream and downstream effects are presented separately. For the same reason, the autoregressive effects are present in the model, but merely represented in the figure by the gray paths. The dashed arrow is the added effect to improve model fit. Model fit unconstrained model: χ2(2) = 0.83, p = .659; TFI = 1.00; RMSEA = .00 (90%CI [.00, .03]), PCLOSE = 1.000

Full size image

With regard to the upstream effects, we do not find a consistent pattern across the waves. Between w1 and w2, we find no significant positive upstream effects but a significant negative upstream effect of PP IV on PP II (b = − 0.04, se = 0.02, p = .035). This means that looking at the changes within individuals, frequent participation in PP IV significantly decreased the chances of participation in PP II. Between w2 and w5, we find that PP IV has positive effects on both PP II (b = 0.18, se = 0.06, p = .004) and PP III (b = 0.09, se = 0.05, p = .059), although the latter does not reach conventional levels of significance. This means that individuals with more frequent activity in PP IV are more likely to show increased levels of PP II and (partially) III in the next wave.

With regard to downstream effects, again, the patterns are not consistent across the waves. We do find that between w1 and w2, PP IV is affected by PP II (b = 0.34, se = 0.04, p < .001). Between consecutive waves, we find no other significant effects. PP II does not affect PP III, nor do we find any significant effects between w2 and w5. The effect added between PP II in w1 and PP IV in w5 to improve model fit (dashed arrow) is positive (b = 0.35, se = 0.03, p < .001).

Although the differences in coefficients across the waves suggests otherwise, we still run a model with the effects between the same types of participation to be equal across the waves. With this constrained model we can formally test whether it is not likely that the variations we find in the effects are not mere random fluctuations from consistent effects over time. To test whether the effects between the waves are similar, we therefore ran the model with the effects between the same types of participation to be equal across the waves. The constrained model is close to sufficient fit but has a relatively low TLI value (see Model 3 of Table 1). At the same time, the chi-square difference test (χ2D(6) = 31.99, p < .001) and the RMSEA difference (\(\hat{\varepsilon }_{D}\) = .05) seem to indicate a loss of fit when comparing the constrained model to Model 2, although the chi-square difference test may be significant due to the large sample and the RMSEA is on the threshold of acceptable loss of fit. From this, we conclude it is better to interpret the effects unconstrained between the waves.

Our main finding is that we detect both upstream and downstream effects between PP II and PP IV. But the upstream effect, going from less effortful to more system-targeted forms of participation, occurs between w2 and w5 (with an election in between). In contrast, the downstream effect occurs between w1 and w2 (without an election in between). Added to this, between PP II and PP IV, we find a positive downstream effect between w1 and w2; we also observe a negative upstream effect simultaneously. And between w2 and w5, we find exactly the reverse (although here the negative downstream effect is not significant): A negative downstream and a positive upstream effect between PP II and PP IV.

These findings may indicate some replacement effect, where activities in the form of participation substitute activities in another form. However, a closer look at the descriptives of our data does not corroborate this. We find that between w1 and w2, 10.2% of our respondents increased in PP IV while decreasing on PP II, and between w2 and w5, 7.7% of the respondents increased in PP II while decreasing in PP IV. For these respondents, we can speak of a partial substitution. Overall, however, these numbers are smaller than one would expect if a substitution effect were the main explanation for the upstream and downstream effects we find (i.e., when positive changes in one participation form were associated with a negative change in the other). As such, we cannot speak of a general trend of substitution.

Discussion

This study investigated the upstream and downstream effects of non-electoral participation over time. It contributes to a growing field of research that aims to answer the question of how different forms of political participation evolve and influence each other (Cantijoch et al. 2016; Kim et al. 2017). We find that people participate most frequently in types of participation most distant to the political system and least often in forms that directly target the political system. While this shows preferences among the citizenry, opportunity structures may explain the differences, as it is easier to, for example, constantly boycott specific products than contact politicians.

Are forms of participation distant to the political sphere replacing more traditional and citizen-oriented forms of participation over time, or are they a jumping board for more involvement? By calculating intra-individual changes between types of participation that differ in effort and their distance to the political system, the study contributes to our understanding of the transformative effects previous acts of participation can have on future activities. Three main findings are worth discussing in the context of extant and future work and societal implications.

First, we find no clear indication for only upstream or downstream effects but can establish that both types of effects occur over the course of almost 1 year. Following the conceptualization of participation by Van Deth (2014) to distinguish three types of participation that differ in levels of effort and resources, we find that participation in one type of activity can have transformative effects on other types of participation over time. Importantly, we investigate intra-individual changes and thereby the participation trajectory of each citizen separately. Our evidence for both upstream and downstream effects somewhat contradicts Nanz and colleagues (2022), who find low-effort participation to predict higher effort types over time. Importantly, we find this evidence for transformative effects between participation directly targeted at the political system (PP II) and non-political, politically motivated activities (PP IV). Here, the downstream effects are more substantial than the upstream effects. Participation targeting the local community (PP III) has no transformative effects on other types, nor do other types transform into these activities. This suggests high stability in types of local participation, which may also be explained by the fact that people who have settled in life engage more strongly in these types of activity (Andersen et al. 2021).

These findings indicate that, first, people switch between types of participation over time. This volatility can be good news for democracy, as there seems to be no clear trend towards types of participation that are less demanding and more distant from the political system. Second, only a few people entirely substitute one type of participation with another over time. This may not be surprising, as citizens do not decide on acts of participation based on our conceptualization. The small share of citizens who increase in one type and decrease in another (10% and less) shows that participation trajectories are gradual and rarely absolute.

The second striking result relates to the between-wave differences we find. While we theorized upstream and downstream effects as a linear process, our exploratory approach revealed that there is strong indication for a downstream and negative upstream effect between w1 and w2, with strong indication for an upstream effect between w2 and w5. The two timeframes were separated by the 2015 Danish national election, taking place between w3 and w4. We acknowledge that this is a post hoc interpretation, but the possibility exists that the election has turned citizens’ participation trajectories upside down. Elections are political heydays in the democratic life of citizens, especially in Denmark, with its short and intense campaign time that increases news media use, political interest, and political knowledge among the citizenry (Andersen et al. 2021). This increase in civic skills and interest may explain why participation in low-effort types before the election predicts high-effort participation after the election. However, other seasonal effects, such as higher energy levels and aspiration after a summer break, may also be influential. Future research should, therefore, explore the role of election for a reverse effect in participation trajectories more closely.

Importantly, we find indications for a downstream and even negative upstream effect between the first two waves. Should transition away from system-related activities be the ‘democratic normal,’ this can be problematic because a sustainable, trusted political system needs to be responsive citizens’ input and citizens need to perceive the system to be responsive, too (Almond and Verba 1963). Structural downstream effects may also challenge the legitimacy and functioning of representative democracies. If those voting and holding office differ from those not engaging in this type of political participation, their voices will be less visible and consequential. However, we find little indication for replacement effects of one type of participation with another. Moreover, low-effort types of participation can also be perceived as a ‘buffer’ that prevents people from giving up participation (e.g., if they are disappointed by their participation in more demanding types or have to stop their participation due to a lack of resources or perceived risks). Hence, future research should stay attentive to downstream effects and test the overtime relationship between low-effort activities and no participation at all.

A third, unexpected finding of our study was a very strong lagged effect between PP II in w1 and PP IV in w5. This needed to be included to achieve a significant model fit, which means there is a strong association between these types of participation over time independent of w3 and has, compared to other relations we observe, a lag of one wave. This can mean multiple things: first, it might simply be a statistical artifact. Second, the possibility exists that besides the wave-to-wave changes we explore, a strong underlying current exists. This current only exists for the downstream effect. It outnumbers all other coefficients in strength, suggesting that the 4-month differences between the waves were too short to estimate a more general trend. We do not wish to overload this finding with post hoc interpretations, as literature about the relevant lag to estimate participation trajectories is sparse. Research on electoral participation often uses the 4- or 5-year rhythm between elections (e.g., Green and Shachar 2000; Goldberg and Sciarini 2023; Holbein and Rangel 2020; Jessen et al. 2021, Smets and Hams 2013). Studies investigating non-electoral development use shorter intervals: Cantijoch et al. (2016) and Nanz et al. (2022) rely on pre- and post-election waves with shorter intervals. Kim et al. (2017) examine the change between online and offline participation over a year. Our study now adds that 8 to 9 months is a timeframe in which relevant overtime changes in non-electoral participation can be detected. Future research should investigate the relevant lags for non-electoral participation further.

Limitations

This study has several noteworthy limitations. First, we rely on a single-country study, and patterns we find may not be universal, especially because we use an opt-in panel, which despite being standard in most social science survey research today, challenges the assumption of representativeness (see Baker et al. 2010). Denmark, the country of this study, can be considered a least likely case to experience changes in participation patterns over time. The political stability in the country, a well-functioning welfare state, and high satisfaction with and trust in the political system make especially upstream effects towards more participation targeting the political system less likely. On the other hand, there is an opportunity to use political participation as a self-actualizing means for identity building, as people can afford such expressive types of participation due to high tolerance and undisputed freedom of speech. Future research should, therefore, test for upstream and downstream effects in other countries and political systems.

Second, we use a broad measurement of political participation to capture all the different nuances of activities. However, being able to observe change led us to exclude relatively stable modes of participation, such as membership in parties and other political or community organizations. Such modes may afford even higher commitment and effort by citizens, and we acknowledge that our model cannot account for potential upstream effects that lead to long-term commitments. At the same time, we excluded activities with very low effort, such as liking content on social media, as those were, per definition by Van Deth (2014), not conforming to the necessary bases of political action but can rather be understood as a political expression (see Gil de Zúñiga et al. 2014; Ohme et al. 2018; Theocharis 2015; Vaccari et al. 2015; Yamamoto et al. 2015). We, therefore, cannot test whether downstream effects to such activities or even towards non-participation exist. Future research should investigate to what extent the examined types of political participation are related to acts of political expression over time.

Overall, we find only partial indication that one-directional developments in non-electoral participation exist. This also speaks against the likelihood that participation moves away from the political system in general, potentially causing ‘democratic frustration’ (Harrison 2020). The downstream effects we explore are potentially ‘saved by the election campaign,’ which seems to turn this trend into upstream effects. Staying in the picture of the study title, elections may function as a superfood for salmons to jump up the cascades of political participation. However, some underlying currents that only unfold over a more extended period might complicate jumping the participation ladder.

The consequences of these downstream effects need to be explored further, especially among younger generations; noteworthy examples showing the opposite, such as the Fridays For Future movement, exist (see Boulianne and Ohme 2021). A political system needs to be prepared to receive less direct input from its citizens or present possibilities for how citizens—with low effort—can participate in politics directly. Moreover, we need to study how easy entry into citizen participation can be made possible. Therefore, more academic fisherwomen and -men are necessary to study the journey of the salmons in political participation.