Economic security as Labour's electoral foundation

Substantively, having a higher income, greater savings, a secure full-time job or being retired, and owning property (particularly outright) are all associated with lower probabilities of being subjectively economically insecure. As is a person’s belief that they can rely on the UK benefits system if needed. In contrast, having childcare responsibilities and unsecured debts, as well as having a disability, not owning property and being in insecure work or lacking employment at all, are all associated with a greater likelihood of insecurity.

So economic insecurity can stem from multiple different sources, and it is a broader concept than simply income and employment (and so on) alone. But how might it help us explain contemporary British politics?

From the graph, one can make out the patterns of volatility and fragmentation that have characterised the British party system in recent years. That is, there are substantial vote flows into and out of the parties during the 2019–2025 era (indicative of people switching their party preference regularly), and both the ‘left-’ and ‘right-wing’ party ‘blocs’ have become more evenly balanced between multiple parties as support for the Conservatives and (eventually) Labour has declined.

The Conservatives had a historic challenge in 2024 with a dramatic fall in their vote share, from 43.6% in 2019 to 23.7% in 2024, their worst ever performance. This equates to a change from 33% in 2019 to 19% in 2024 when expressed as a proportion of the whole electorate. Extending our analysis to May 2025, we see that the Conservatives have lost around two-thirds of the vote coalition that they gathered to win the 2019 General Election, with most of these voters going to Reform UK, either in 2024 or the year following.

This inflow, along with gains among previous non-voters, has helped propel Reform into first place, with support rising from around 2% of the electorate to 23% between 2019 and 2025. This is made easier for Reform because the 2 largest parties’ support has been ebbing away, making ‘first place’ a lower bar.

While the ‘right-wing’ bloc has split in 2 between 2019 and 2025, the ‘left-wing’ bloc has become even more fragmented. The 2024 Labour vote resulted from modest increases in support between 2019 and 2024 (from 25% to 27% as a proportion of the electorate), originating across the ideological spectrum.

British Election Study data showed that, among the relatively small number of voters switching to Labour between 2019 and 2024, 35.8% were from the Conservatives, 27.4% were from the Liberal Democrats, 5.9% were from the Greens, and 11.9% were from ‘others’ (with the remainder being previous non-voters). This was a very small total increase in vote share, but one which delivered Labour a very large majority because of its fortuitous geographic distribution (Miori and Green, 2025), which arose from the split of the Conservative vote on the right, tactical voting among progressive voters, and 2 unpopular incumbent governments: the devolved Scottish Nationalist Party Government in Scotland, and the Conservative Government in Westminster.

Though Labour made modest gains between 2019 and 2024, they have shed substantial numbers of supporters — almost two-fifths when expressed as a proportion of the electorate that previously supported them — during their first year in office.

Given the ‘bloc’-based nature of politics post-Brexit (Griffiths, Fieldhouse et al., 2025), these losses have mostly been to other socially liberal, centre-left parties (the Greens and the Liberal Democrats), but they have also lost some voters to Reform UK and, in particular, to ‘undecided’. Whether these undecided voters — equivalent to almost 1-in-5 potential voters as of May 2025 — ultimately return to their 2024-era parties, opt for a new political home, or abstain entirely will have a big influence on the outcome at the next election.

This cumulative ‘vote flow’ diagram is good for getting a sense of the degree of overall fragmentation and volatility in contemporary British politics, but to really get a sense of where both major parties are losing voters since 2024, we visualise the direction of losses to Labour and the Conservatives ‘losses’ expressed as a percentage of their previous vote share. These are shown in Figures 3 (for Labour) and 4 (for the Conservatives).

Starting with Labour, in Figure 3, we can see that the party has lost over half (51%) of its 2024-era supporters as of May 2025. With roughly even proportions going to the Greens (8.2%), Liberal Democrats (9.4%) and Reform (7.9%). Very few Labour voters have defected to the Conservatives (2.3%), as we might expect given that party’s continuingly weak showing in opinion polls.

Optimists within Labour might point to the fact that the single largest proportion of defectors have become ‘undecided’ (18.9%) rather than defecting to a specific rival party. As we will later show, combating economic insecurity may be key to winning back these (and other) defectors between now and the next election.

Figure 4 shows that, while Labour’s vote has fragmented in many directions – left, right and centre – the story for the Conservatives is very much one of ‘revolt on the right’. Like Labour, the Conservatives have lost over half (51.1%) of their 2024-era supporters, but these voters have mostly gone in one direction: to Reform UK.

Reform UK have picked up 3-in-10 2024-era Conservatives (29.8%), whereas only 1-in-25 have gone to either of the Greens, Labour or the Liberal Democrats. As with Labour defectors, a substantial number have also switched to being ‘undecided’ (15.8%), and it may be these supporters that could prove most liable to be won back in the coming years.

The context is important here. It is exceptional and unprecedented for a party that has lost an election after being in government to continue to lose even more support than it gains. These 2024–2025 patterns are continuations of patterns that took place between 2019 and 2024, when the Conservatives started to shed supporters to Reform UK.

As we demonstrated in our previous report (Nuffield Politics Research Centre, 2024), the Conservatives were heavily punished when they were in office, with the party’s losses (to both Reform UK and Labour) being highest among economically insecure ex-supporters.

Clearly, however, a large part of the overall story of the Conservatives between 2019 and 2025 is the rivalry with Reform UK and the split of the right-wing vote (which, as we shall show, also owes a lot to continued grievances over rates of immigration). If we zoomed out to look at the total story between 2019 and May 2025, our numbers would show that the Conservatives have lost fully 72% of those who supported Boris Johnson’s government at the 2019 election, with almost half of these previous supporters (44.8%, or 62% of all defectors) now opting for Reform UK.

As our earlier overall ‘vote flow’ diagram (Figure 2) showed, whatever few gains the Conservatives have made from other parties or non-voters is clearly not enough to compensate for this haemorrhaging of support on the right.

Another way to look at this continued split is to consider the composition of Reform’s vote over time (not displayed). In May 2025, 41.9% of Reform’s post-July 2024 gains — that is, excluding those already voting for the party at the last election — came from 2024 Conservative voters, 16% from Labour and 32.6% from non-voters.

If we zoomed out to take a longer perspective that would encompass the entirety of Reform’s growth since 2024 (as visualised in Figure 2), we would see that of Reform’s May 2025 supporters, well over half (58.9%) voted for Boris Johnson’s Conservatives in the December 2019 General Election. Almost another quarter were those who did not vote in that election (22.2%), and only around 1-in-10 (9.7%) were Labour supporters in 2019.One must go back multiple elections, to the Blair era, to see substantial numbers of current Reform’s supporters, via UKIP and the Brexit Party, originating from Labour defectors, because of the way the Labour and Conservative parties’ voters have changed over time (Griffiths, Green and Fieldhouse, 2024).

These patterns — of broad splintering of Labour’s support, and the continued ‘split on the right’ — are outcomes that have their foundation in the ‘Brexit realignment’ that has organised voters into 2 electoral blocs (Griffiths, Perrett et al., 2025): a Conservative-Reform supporting bloc that supports or supported ‘Leave’ in the Brexit referendum and is broadly right-of-centre and more socially conservative; and a Labour-Liberal Democrat-Green and nationalist supporting bloc that supports or supported ‘Remain’ in the Brexit referendum and which is broadly left-of-centre and more socially liberal.

Labour’s 2024 vote was predominantly comprised of the latter bloc, which explains the dominant direction of vote losses to the left, or to undecided, as opposed to Reform UK. The Conservatives’ 2019 and 2024 votes were predominantly comprised of the former bloc, which explains the dominant direction of vote losses to the right.

Indeed, the strong alignment of ‘Leave’ and socially conservative voters with the Conservatives following their stronger rightward shift post-Brexit, meant that their vote, if they lost support as they did during their period in office between 2019 and 2024, would most naturally split right with competition from another right party, which Nigel Farage supplied in Reform UK. This right-wing split and the rise of Reform are consequences of the Brexit realignment under May and Johnson, and the subsequent implosion of support for that government.

Sceptics might highlight that a large proportion of both Labour and the Conservatives’ July 2024-era supporters — 19% and 16%, respectively — are ‘defecting’ to ‘undecided’ rather than to any other named party. Although we can see evidence for ‘bloc’-based patterns of defection when looking at those naming a preferred other party, it would clearly complicate this picture if the large number of voters that Labour is losing to ‘uncertainty’ actually tended to prefer right-wing parties. Indeed, this might give some indication that these voters were on the way to Reform. Likewise, the argument about party blocs would also be weakened if many of the Conservatives’ losses to ‘uncertainty’ actually preferred more centrist or left-wing options to Reform.

One way to get a sense of where these currently undecided voters are likely to end up is to use their responses to various ‘propensity-to-vote' (PTV) questions embedded in the British Election Study. Respondents were asked, ‘How likely is it that you would ever vote for each of the following parties?’ They then rated each of the larger UK parties from 0 (‘Very unlikely’) to 10 (‘Very likely’). Don’t knows were excluded.

Figures 5 and 6 below visualise the average PTV score assigned to the Greens, Liberal Democrats, Labour, the Conservatives and Reform (aligned roughly left-right as far as sociocultural issues go), among currently undecided voters who supported Labour (Figure 5) or the Conservatives (Figure 6) at the last general election.

As can be seen, observing the self-declared, most-likely destination of the currently undecided — were they to support any party — shows that the patterns of bloc-based defections that we have identified remain corroborated. Ex-Labour supporters assign a far higher PTV score to parties on their left, in sociocultural terms, than parties on the right. Indeed, currently undecided, ex-Labour voters give a particularly low PTV score to Reform, suggesting it is highly unlikely that Nigel Farage can make many further gains from Labour's 2024-era coalition.The roughly 1-in-5 voters that Labour has lost to ‘uncertainty’ since July 2024 are not Reform UK voters in disguise.⁴

Conversely, of course, among undecided ex-Conservative voters, Reform is clearly more popular than Labour, the Liberal Democrats and (in particular) the Greens. The Conservatives have more to worry about to their right than to their left here, which was the lesson from the vote flow graphics above.

Finally, of course, it is notable that among both sets of ‘defectors’, the party of origin gains the highest ‘PTV’ score (that is, currently undecided ex-Labour or, in particular, ex-Conservative voters are more likely to say they will vote for that party than any other). This is a further reminder that not all of these defectors are lost forever, and both major parties could win back this support with the right offer.

To what extent might a harder position on immigration help win these voters back? This is important for our report, because we are focused on understanding the relative weighting of people’s economic insecurity and their immigration preferences on the current patterns of switching. If Labour’s smaller losses to Reform are characterised by anti-immigration attitudes, but their much greater losses to undecided, the Greens, and the Liberal Democrats are comprised of voters with more liberal attitudes, Labour’s difficulties with dealing with the immigration issue become more apparent.

In Figure 7, we present the distribution of attitudes to immigration among 2024 General Election Labour supporters according to their present (May 2025) vote intention. Immigration attitudes are measured using a question in the NPRC/JRF Economic Insecurity Study that asked respondents, ‘How much would you support or oppose a government making it easier for immigrants to come to Britain to work?’

We distinguish between those who would support (or strongly support) the proposal, those who would oppose (or strongly oppose) it, those who neither support nor oppose it, and those who say ‘don’t know’. We visualise the attitudes of ‘Labour loyalists’ who continue to support the party as of May 2025 and ‘Labour defectors’ who moved to supporting other parties, being undecided or intending to abstain.

Observing the attitudes of those moving from Labour to specific rival parties is difficult due to the relatively small subsamples available in our JRF/NPRC panel (from which the immigration measure is derived). This forces us to group parties together in broad groups. Accordingly, we then also present the distribution of attitudes among those defecting to ‘left-liberal’ parties (that is, the Greens, Liberal Democrats, SNP or Plaid Cymru), to ‘right-conservative’ parties (that is, the Conservatives or Reform UK) and to ‘Undecided’. As shown earlier, in Figure 3, these groups are comprised of roughly 19%, 10% and 19% of July 2024 GE Labour supporters who were re-interviewed in May 2025.

The results speak to the difficulty that Labour may have in seeking to mobilise current and past supporters on the basis of the immigration issue. While, overall, those voters whom Labour have lost since July 2024 tend to be somewhat more anti- than pro-immigration (by 39% to 26%), remaining Labour voters tend to be slightly pro-immigration on average (that is, they split 33% to 25% in the other direction).

Furthermore, there are large numbers among both groups (36% and 42%, respectively) who are either undecided or neutral on the issue. While it is not the case that current or previous Labour supporters are dramatically in favour of increased rates of immigration, nor are they easily pigeonholed as staunch opponents.

Things are more interesting still if we look at the attitudes among those defecting in different directions (left, right or undecided). While, as we might expect, those defecting to Reform UK or the Conservatives tend to overwhelmingly disagree with proposals to further liberalisations of the immigration system, opposing the statement by 73% to 9%, the more numerous Labour defectors to ‘left-liberal’ parties or to ‘undecided’ tend to be more liberal or at least moderate on the issue.

Defectors to ‘left-liberal’ parties (that is, Liberal Democrats, SNP, Plaid Cymru and the Greens) show more support than opposition (42% to 22%), and the undecided are also less uniformly hostile (22% support to 32% oppose). Furthermore, over one-third of ‘liberal-left’ defectors, and close to half of all currently ‘undecided’ defectors, are simply neutral or undecided on the immigration question. This clarifies the dilemma for Labour.

While it is conceivable that signalling strong opposition to immigration might help the party claw back some losses to the Conservatives or Reform UK, it is unlikely to be a panacea or even a particularly effective stopgap in regaining the heavier losses to left-liberal parties or the undecided, or even maintaining the supporters sticking with it for the time being.

This helps to understand where voters are going to and where they have come from, but it does not explain why. This is the question we turn to next.

How many ‘lost’ voters, between July 2024 and April 2025 (coinciding with Wave 3 of our survey), might Labour have been spared, had their economically insecure former supporters instead have behaved like their more economically secure counterparts?

Figure 9 visualises this counterfactual, drawing on both the raw differences between the rates of defection among both insecure and secure 2024-era Labour supporters as well as the relative sizes of these groups as a part of Labour’s overall vote share.

While losses to the Conservatives and to abstention may have been marginally higher if the ‘insecure’ group instead behaved exactly like the ‘secure’ segment of their electorate, Labour’s overall retention rate among its 2024 General Election supporters would have been about 4 percentage points higher thanks to the cumulative effect of stemming defections to the Liberal Democrats, Greens, Reform UK, other minor parties and ‘undecided’.

The 4 percentage points saved would be equivalent to saving roughly 8% of Labour’s total vote loss between July 2024 and April 2025 (51%). These numbers would increase to 4.9 percentage points and 9.6%, respectively, if Labour supporters who were ‘neither secure nor insecure’ or undecided on their economic insecurity also had the lower defection rate of the ‘secure’ group.

We can actually interrogate these counterfactual scenarios even more robustly. Our panel data allows us to examine within-person change, such as the positive change of a respondent moving from reporting being insecure in Wave 1, to either a positive ‘secure’ category, or a middle category of neither secure nor insecure in Wave 3. It also allows us to examine those people for whom there has been no improvement or a decline in reported economic security. The latter is particularly damaging for a recently elected government, especially if hopes had been raised that things were going to get better, and if those hopes were dashed within the first year of Labour’s tenure.

Accordingly, Figure 10 presents the rates of defection from Labour — both overall and to specific parties, undecided or abstention — by April 2025 among the party’s supporters at the July 2024 General Election who: (a) moved (from March 2024 to April 2025) from being economically insecure to no longer being insecure; (b) remained not economically insecure throughout both time periods; (c) were continuously insecure in both time periods; and (d) moved from being not economically insecure into insecurity. These trajectories represent the journeys taken by 14%, 52%, 25% and 8% of Labour’s total 2024-era coalition, respectively.

Due to the relatively small samples of respondents at play, we do not distinguish between being ‘economically secure’ and being ‘neither secure nor insecure’ among our group of ‘not economically insecure’. However, logically we should still expect to see that Labour suffers progressively greater rates of defection as we move from groups (a) to (d). This would be in general accordance with the logic of reward–punish models (Tilley, Neundorf and Hobolt, 2018) in the political science subfield of ‘economic voting’. That is, voters should be liable to reward parties who preside over improvements in their welfare and punish those who preside over deteriorations.

As such, we should expect Labour voters to generally have higher rates of loyalty to the party when they are not experiencing economic insecurity, but particularly when they have experienced a recent improvement that might be perceived as linked to the actions of government. Conversely, loyalty should be lowest when voters feel insecure and especially when their transition to this state came only after the party took office.

Happily, for expectations of a (reasonably) rational electorate, this is in fact what we see. Rates of defection (including to indecision) are lowest among the group of Labour supporters that moved out of insecurity during the party’s first year in office (45%). Although it is true that rates of abandonment are only marginally higher for those who remained non-economically insecure (46%), these rates of defection are substantially lower than among 2024 Labour voters who have remained continuously insecure during Labour’s government (59%) or who actually transitioned into insecurity under Labour’s tenure (64%).

While the aggregate amount of defection presented in this figure is largely due to substantial numbers of Labour voters moving to ‘undecided’ (as highlighted in the earlier vote flow diagrams), it is notable that Labour supporters who experienced an improvement in their security show the lowest likelihood of defection to Reform UK. They are around 2.5 times less likely to defect to Nigel Farage’s party than those who remained insecure during Labour’s first year in office.

To understand the cumulative effect that improving the security of the insecure supporters or preventing others from becoming insecure might have had for Labour’s vote share, we turn to Figure 11. This again presents a visualisation of the counterfactual: how would Labour’s vote share improve if insecurity among former supporters was ameliorated or prevented from occurring?

As with Figure 9, it is based both on comparing the rates of defection among 2 groups and also their relative size as a proportion of Labour’s July 2024-era total support. Specifically, we show how many fewer losses — as a percentage of Labour’s overall April 2024 vote — would have been prevented had (a) all consistently ‘insecure’ supporters behaved like formerly 'insecure’ supporters who moved out of insecurity, and (b) had all supporters who moved into insecurity behaved like those who remained continuously non-insecure.

In counterfactual scenario (a), Labour’s vote share among its previous (July 2024) supporters would be 3.3 percentage points higher; in scenario (b), the retention rate would be a further 1.5 percentage points higher. This ‘saving’ of almost-5 percentage points in Labour’s 2024-era vote share would be equivalent to preventing almost 10% of Labour’s defections between July 2024 and April 2025 — would result from stopping losses to the Liberal Democrats, Greens and Reform, in scenario (a), and mostly by preventing losses to ‘undecided’ in scenario (b).

Accordingly, based on Figure 11, a naïve estimate of one of the benefits to Labour, had they improved feelings of economic security, would be to have prevented up to 10% of their losses during their first year in office. Of course, the actual difference could be even larger if feelings of security were maximised to the fullest, beyond the constraints of our ‘insecure’ versus ‘not insecure’ simplifications, and if insecure supporters of other parties moved towards Labour to reward the party for alleviating their financial anxiety.

However, even this conservative estimate would have substantial implications for the result at the next general election, given the increasingly competitive nature of recent local and by-elections.

In sum, our data shows that economically insecure voters are more likely to have abandoned Labour since the 2024 election. This direct effect of economic insecurity on defections from Labour remains consistent in statistical models which consider demographics and broader economic evaluations, and it is consistent across the party choice spectrum in 2025.

Taken together, these comparisons of insecure versus secure voters (a ‘between-person’ analysis) indicate that the insecure are more likely to have defected from Labour since the last election, compared to the economically secure.

As our data is a panel, which obtains responses from the same individuals on repeated occasions, we are also able conduct a ‘within-person’ analysis of the direct effect of economic insecurity. In other words, we are able to test whether 2024 Labour voters become more likely to defect from Labour as they become more economically insecure.

To do so, we estimate the average increase in the likelihood of defecting from Labour to different parties (expressed as percentages) that occurs when a person becomes more economically insecure. We calculate these effects for defections from Labour to Reform, the Greens, Lib Dems and undecided, and plot them below in Figure 12.⁶

This demonstrates the possible consequences for Labour of failing to address the issue of people’s feelings of economic insecurity. Additionally, estimating the effect of changes in people’s economic insecurity on their vote switching provides greater confidence that economic insecurity has a causal effect on Labour defections by ruling out alternative explanations.

Results from these models show that as 2024 Labour voters become more economically insecure, they become more likely to defect from Labour to Reform, the Lib Dems and ‘undecided’. Specifically, an increase of just under 3 points on the 0–10 economic insecurity scale is associated with an increase of 3.6 percentage points in the likelihood of defecting to Reform, an increase of 2.3 percentage points in the likelihood of defecting to the Lib Dems and an increase of 2.7 percentage points in the likelihood of defecting to ‘undecided’.

Among 2024 Labour voters in our survey who participated in 2 or more waves, 25.6% experienced a change of 3 or more points on the economic insecurity 0–10 scale on at least one occasion. This shows that the changes in economic insecurity represented in Figure 12 are not rare instances but instead a frequent experience among Labour voters, with potentially far-reaching political consequences. Taken together, this evidence highlights the importance of addressing economic insecurity if Labour are to stem its tide of defections since the 2024 election.

2. ‘Indirect effect’ of economic insecurity on Labour’s handling of national economy

As explained in the introduction, the electoral importance of ‘the economy’ is not only via a person’s own economic circumstances and their worries or their lack of worries about those circumstances. It is also about the national economy as a signal of government competence and whether things are going in the right direction (Lewis-Beck and Stegmaier, 2019). That is helpful information because it gives voters greater, or lesser, confidence in the Government overall and provides a competence signal that transfers to other policy issues (Green and Jennings, 2017), while a growing economy makes fiscal and spending decisions easier.

The association between evaluations of a government’s handling of the national economy tends to be strongly related to vote choice. Indeed, for the defections that have taken place between 2024 and 2025 for Labour, we find in our statistical analysis that evaluations of Labour’s handling of the national economy have a substantial effect on defections to all destinations (see Table 6 in the Appendix), independent of economic insecurity. Put differently, perceptions of how Labour is handling the economy are very important for its electoral prospects, alongside people’s experiences and feelings of economic insecurity.

Critically, these experiences and evaluations of the national economy are interlinked. Among those respondents in our sample who are economically insecure, ratings of Labour’s handling of the national economy are significantly more negative. This can be seen in Figure 13, which shows the relationship between economic insecurity and evaluations of Labour’s handling of the economy. Specifically, we plot the percentage of insecure and secure respondents at each wave who evaluate Labour’s handling of the economy as ‘Very Bad’ (the most negative category) at Waves 2 and 3 of the JRF survey (October 2024 and April 2025).

This comparison shows that insecure respondents are more likely to evaluate Labour’s performance as negative. At Wave 2, 40.1% of insecure respondents evaluate Labour’s handling of the economy as ‘Very Bad’, while only 22.8% of the economically secure do so — a gap of 17.3 percentage points. At Wave 3, the gap between the insecure and secure is similarly sized, at 17.7 percentage points.

Next, in Figure 14, we again plot the percentage of respondents who evaluate Labour’s handling of the national economy as ‘Very Bad’, but this time break the sample down by whether respondents were Labour voters in 2024 or voted for other parties in 2024.

As the plot shows, 2024 Labour voters are less likely to evaluate Labour’s economic performance as negative, but the gap between the secure and insecure remains. Specifically, at Wave 2, 18.2% of insecure Labour voters evaluated the Government’s handling of the economy as ‘Very Bad’ — 12.4 percentage points more than economically secure Labour voters. At Wave 3, the equivalent figure was 22.8%, compared to only 10.3% of economically secure Labour voters, representing a 12.5 percentage point gap.

While these figures represent simple percentages, statistical analysis suggests to us that the association between economic insecurity and evaluations of Labour’s handling of the national economy is very likely causal, and that insecurity leads to more negative handling evaluations rather than the other way round. We derive this conclusion in 3 ways.

First, we note that the relationship between economic insecurity and evaluations of Labour’s handling of the national economy holds for people who voted Labour in 2024, and those who did not, and so is not reducible to partisan effects. Labour’s 2024 voters rate Labour less negatively overall, but the distinctions between the insecure and secure persist.

Second, we find evidence that economic insecurity is associated with economic handling evaluations at the within-person level, which provides stronger evidence that this relationship is causal (see Table 5 in the Appendix).

Third, we assess the direction of these assessments in a full panel model from an earlier period, and these confirm that economic insecurity leads to negative assessments of handling of the national economy more strongly than vice versa.

We can only run these analyses where both are measured over 3 or more waves of a survey, and this was only true for the British Election Study panel, in month-years December 2022, May 2023 and May 2024, and therefore the handling evaluations relate to the Conservative Government’s handling of the economy. We report the path diagrams in Appendix Figure 21, which shows the statistical relationships between a person’s economic insecurity and their evaluations of the Government’s handling of the economy at the subsequent wave, along with the reverse relationship (that is, the relationship between government handling evaluations and insecurity at the subsequent wave).

Comparing these ‘cross-lagged’ effects allows us to test whether it is economic insecurity that is driving changes in economic handling evaluations, or vice versa. Conducting this comparison, we do not find that assessments of the handling of the national economy lead to changes in feelings of economic insecurity, but we do find that experiences and feelings of economic insecurity have some subsequent effect on evaluations of handling of the national economy.

This provides evidence that economic insecurity is ‘causally prior’ to evaluations of government performance on the economy. In other words, when people become economically insecure, this insecurity subsequently prompts them to evaluate government economic performance as more negative — not the other way round.

3. Economic insecurity amplifies the impact of Labour's handling of the national economy on Labour defections

The measurement of economic insecurity gives us an insight into people’s worries. This is a particularly powerful concept because a person’s worries will make them attentive and responsive to other signals (Albertson and Gadarian, 2015), heightening their awareness and their need for other aspects of the economy, or immigration, to be handled better.

We anticipate that economic insecurity will amplify the relationship between evaluations of Labour’s handling of the economy and the tendency to defect from Labour. When people are worried about their household finances, we expect that they will pay greater care and attention to government handling of the national economy, because evidence shows that worry prompts people to seek out information (Bonifai, Malesky and Rudra, 2024; Valentino et al., 2008).

Second, feelings of economic insecurity likely raise the salience or importance of the economy to an individual, which means that the insecure will place greater emphasis on economic concerns when choosing who to vote for, as previous academic studies would suggest (Fournier et al., 2003; Green and Hobolt, 2008). These consequences of economic insecurity, we argue, will increase the likelihood that people who evaluate Labour’s handling of the national economy ‘punish’ the Government by defecting to other parties, an ‘indirect effect’.

Our analysis supports this argument. In statistical tests, we identify an indirect effect of economic insecurity on how important Labour’s handling of the national economy is to different individuals and their tendency to defect from Labour. Put differently, we find that economically insecure individuals are more likely to defect from Labour when they evaluate Labour’s economic handling as negative and, again, we find this effect is a broad one for defections to the left and right.

To test this ‘indirect effect’, we combine measures of economic insecurity and Labour’s handling of the national economy into a multiplicative ‘interaction’ term (that is, the individual-specific product of economic insecurity and economic handling scores),⁷ which allows us to assess whether people who are insecure are more likely to defect from Labour on the basis of their rating of Labour’s handling of the national economy than people who are economically secure.

Figure 15 displays ‘average marginal effects’ (AMEs), which are the average difference in the likelihood of defecting from Labour to different parties (expressed as percentages) between a person who evaluates Labour’s handling of the economy as negative, and one who evaluates it as more positive. Specifically, we calculate the difference in the likelihood of Labour defection associated with a difference of just under 1 point on the 5-point economic handling scale (for example, between a person who evaluates Labour’s economic handling as ‘Fairly Bad’ and one who evaluates it as ‘Very Bad’).⁸

Importantly, we calculate these effects for both economically secure and economically insecure voters, which allows us to compare the 2 groups. Each plot in Figure 15 is labelled with the difference between the ‘average marginal effects’ for economically secure and insecure individuals. If a difference is denoted with an asterisk, the difference is statistically meaningful, and the greater the number of asterisks, the stronger the difference.

This shows that, among 2024 Labour voters, those who evaluate the Labour Government’s handling of the national economy as negative are significantly more likely to defect to Reform UK, the Greens and the Liberal Democrats when they are economically insecure. The difference between the secure and insecure for Reform defections is 5.8 percentage points; for the Greens, it is 3.5 percentage points, and for the Lib Dems, it is 2.9.

These differences remain significant for Reform, Green and Lib Dem party defections with additional statistical controls added to the model, as we show in section A3 of the Appendix. Economic insecurity therefore has a significant and broad indirect effect on voting via its amplification of concerns about Labour’s handling of the economy, and it makes people hold more negative evaluations of Labour’s handling of the economy.

Additionally, for Reform, Greens and ‘undecided’, we find evidence of a within-person effect of negative evaluations of Labour’s handling of the economy on defections among the economically insecure, but not among the secure. In other words, as economically insecure individuals become more negative about Labour’s handling of the economy, they become more likely to defect. This provides more robust evidence for the indirect effect of insecurity on voting behaviour and further emphasises that Labour can provide a buffer against negative assessments of national economic performance if it takes further steps to address individual and household economic insecurity.

4. Economic insecurity amplifies the impact of Labour's handling of immigration on Labour defections

Our analysis also examines the degree to which feelings of economic insecurity amplify effects for evaluations of Labour’s handling of immigration. If economic insecurity acts as a signal of poor government performance and makes voters more open to the possibility of supporting alternative parties — as we have previously argued (Nuffield Politics Research Centre, 2025) — then economically insecure voters could be more likely to defect on the basis of immigration-based grievances.

Views on immigration are associated with vote choice in Britain (Griffiths, Fieldhouse et al., 2025). This is most likely to be the case when the issue of immigration is highly salient (that is, important) in the minds of voters, as is currently the case in Britain (Dennison and Geddes, 2018; YouGov, 2025a). In other words, immigration attitudes drive a process of ‘sorting’ where voters choose to support the party that best aligns with their views on immigration, or that they deem most likely to deliver their immigration policy goals. We propose that economic insecurity could amplify this immigration-based sorting, prompting voters to place greater weight on immigration when making voting decisions.⁹

Here, we include a multiplicative term for economic insecurity and Labour’s handling of immigration and examine whether this amplification relationship exists on different vote defections, a further indirect effect. These models allow us to test whether economically insecure individuals are more likely to defect from Labour when they evaluate Labour’s handling of immigration as negative, relative to the economically secure.

If economic insecurity acts as a signal of poor government performance and makes voters more open to the possibility of supporting alternative parties — as we have previously argued (Nuffield Politics Research Centre, 2025) — then economically insecure voters could be more likely to ‘punish’ the Government by supporting another party when they evaluate its performance on immigration negatively. Throughout this analysis, we control for evaluations of Labour’s handling of the economy to ensure that we are properly isolating and analysing the consequences of immigration handling evaluations.

Figure 16 plots ‘average marginal effects’. Here, we plot the average difference in the likelihood of defecting from Labour to different parties — Reform, the Greens, the Lib Dems and ‘undecided’ — between a person who evaluates Labour’s handling of immigration as negative, and one who evaluates it as more positive. Specifically, we calculate the difference in the likelihood of Labour defection associated with a difference of 1 point on the 5-point immigration handling scale (for example, between a person who evaluates handling of immigration as ‘Fairly Bad’ and one who evaluates it as ‘Very Bad’).¹⁰

Again, these probabilities are calculated in percentages. For each party-destination, we calculate these ‘average marginal effects’ for both economically secure and insecure individuals and label the size and statistical significance of these differences in the plot. By doing so, we are able to test whether economic insecurity increases (or amplifies) the likelihood of defecting when an individual evaluates Labour’s handling of immigration as negative.

Figure 16 shows that, among 2024 Labour voters, the effect of assessments of Labour’s handling of immigration on defecting to Reform UK (by 4.9 percentage points) and the Lib Dems (by 2.8 percentage points) are greater among the economically insecure than the economically secure. In section A4 of the Appendix, we subject this relationship to more rigorous testing.

We find that the effects reported in Figure 16 remain consistent when accounting for the additional controls, demographics and wider assessments of the economy (whether GDP is getting higher or lower), that we reported earlier. So, as with Labour’s handling of the economy, economic insecurity amplifies the importance of evaluations of immigration for defections.

We assume that ‘handling’ can mean different things to different people; either that immigration is not being sufficiently controlled, or that immigration policy (and rhetoric) is too stringent. For example, among 2024 Labour voters who switched to Reform UK by April 2025, a full 67.9% of these respondents report that they ‘strongly oppose’ making it easier for immigrants to come to Britain to work, whereas among 2024 Labour voters who switched to the Liberal Democrats by April 2024, the equivalent proportion is only 11.7%.

We thus deem it likely that those individuals who evaluate Labour’s handling of immigration as negative and defect to the Lib Dems (amplified by economic insecurity) do so because they view Labour’s immigration policy as too hostile to immigrants, whereas the equivalent process of defection to Reform is likely driven by a perceived failure to reduce immigration numbers.

In section A4 of the Appendix, we conduct a within-person test of this relationship. In other words, we test whether people become more likely to defect from Labour as their negativity about Labour’s handling of immigration increases, and if this relationship is amplified by economic insecurity. We find no evidence that economic insecurity amplifies the within-person effect of evaluations of Labour’s handling of immigration on defections.

This latter finding is consistent with the idea that immigration attitudes influence vote choice via a sorting-based mechanism. We argue that, for many people, immigration preferences remain stable, but prompt people to switch parties to ones that align more closely with their immigration preferences (and that economic insecurity amplifies the likelihood of choosing a party in this way). This means that immigration preferences can influence vote choice without people changing their views on immigration over time, which possibly explains the lack of within-person effects.

These indirect effects of economic insecurity provide a broad way of understanding why helping people feel more economically secure could strengthen Labour’s electoral foundations. Not only are those who feel more secure less likely to defect from Labour, but helping people feel more secure would help improve evaluations of Labour’s handling of the national economy, and dampen (rather than amplify) the electoral risks of Labour’s handling of the national economy and of immigration. On the flipside, our findings suggest that Labour’s electoral prospects will worsen if its voters feel increasingly economically insecure between now and the next general election, a situation that would boost Reform UK and the other smaller parties in multiple ways.

Two further insights are important. First, Labour clearly faces a ‘net-negative’ problem in the way it is losing support both on the left and right due to economic insecurity. It is not broadening its coalition on one side while it loses support on the other; Labour is simply suffering losses due to economic insecurity across the board.

Furthermore, economic insecurity is amplifying immigration problems for Labour on the political right, and losses on the economy to political parties on the left and right. It is not just a case of focusing on the economy or on immigration: they are intertwined for Labour’s competition with Reform UK, as well as for Labour’s competition with the Liberal Democrats.

Figure 17 highlights how the relationship of immigration attitudes to vote defections for Labour is asymmetric in strength. 2024 Labour voters who are anti-immigration are significantly more likely (by 19.5 percentage points) to have defected to Reform and ‘undecided’ (by 15.3 percentage points) than those who are supportive of immigration. However, 2024 Labour voters who are anti-immigration are no more likely to defect to the Lib Dems, and 5.3 percentage points less likely to defect to the Greens.

In section A1 of the Appendix, we show that these effects are consistent with additional controls. These results highlight that immigration is an issue that involves trade-offs for Labour. If the Government shifts toward a more anti-immigrant position, it risks losing supporters to parties on the left; if it shifts toward a more pro-immigrant position, it risks losing supporters to Reform.

In contrast, the relationship of economic insecurity to vote defections is broad across vote destinations, albeit less pronounced. Compared to the economically secure, the economically insecure are more likely to defect to Reform (by 5 percentage points), the Greens (by 7.2 percentage points), the Lib Dems (5.9 percentage points) and ‘undecided’ (by 5.4 percentage points).

These results show that addressing economic insecurity does not carry the same trade-offs for Labour as the issue of immigration, making voters feel more economically secure would stem the tide of Labour defections in all directions. Recall that economic insecurity is also influencing evaluations of Labour’s handling of the economy, and amplifying that effect for immigration, so the above direct effects for economic insecurity are not the ‘total’ effect of economic concerns on vote defections.

Figure 17 plots between-person effects, where we compare the likelihood of defecting from Labour between people who support and oppose immigration, and between the economically secure and insecure. Next, we compare the ‘within-person’ effects of immigration attitudes with those of economic insecurity over the 3 panel waves of JRF data. In other words, we test whether people become more likely to defect from Labour when they become more opposed to or supportive of immigration over time and compare this to the ‘within-person’ effects of economic insecurity.

Specifically, we estimate the average increase in the likelihood of defecting from Labour to different parties (expressed as percentages) that occurs when a person updates their immigration attitudes to become more opposed to immigration. We calculate these effects for defections from Labour to Reform, the Greens, Lib Dems and undecided, and plot them below in Figure 18.¹²

Figure 18 shows that, in contrast to the broad within-person effects of economic insecurity on defections from Labour (to Reform, the Lib Dems and ‘undecided’), the within-person effects of immigration attitudes are much more specific.¹³ Becoming more opposed to immigration over time leads 2024 Labour voters to defect to Reform UK, but not to any other party. An increase of slightly more than one point on the 5-point immigration attitudes scale (for example, moving from ‘Somewhat opposing’ to ‘Strongly opposing’ making it easier for immigrants to come to Britain to work) is associated with a 2.65pp increase in the likelihood of voting for Reform, but has no impact on voting for other parties.

In sum, the consequences for Labour of an increasing anti-immigrant hostility appear less severe than they do for changes in economic insecurity. Our evidence indicates that changes in immigration attitudes drive Labour defections to Reform alone, whereas changes in economic insecurity prompt voters to defect from Labour to Reform, the Lib Dems and ‘undecided’.

Furthermore, as we have previously explained, the issue of immigration involves trade-offs for Labour, whereas addressing the issue of economic insecurity will have broad benefits for Labour’s support. To provide another way of showing this, we compare the ‘within-person’ effects of immigration attitudes and economic insecurity on overall or general Labour defections. This model estimates the average increase in the likelihood of defecting from Labour to any other party or to ‘undecided’ (expressed as percentages) that occurs when a person changes their immigration attitudes or experiences a change to their economic insecurity.

In other words, rather than examining defections to specific parties, this model estimates the general effect of economic insecurity and immigration attitudes on defections to all other parties and ‘undecided’. We plot the results of this model in Figure 19.

The results of this model also allow us to understand the overall consequences of changes in insecurity and immigration attitudes for Labour’s vote share. Specifically, based on the results of this model, if Labour was to increase the economic insecurity of its 2024 voters by 10 percentage points, we would expect a 1.88 percentage point increase in their vote share among this group. In contrast, a 10pp reduction in anti-immigration sentiment among 2024 Labour voters would be expected to increase Labour’s vote share by only 0.65pp — an effect that is statistically indistinguishable from 0, according to our model.

Finally, we should note that the ‘within-person’ effects displayed in Figure 18 are smaller than the ‘between-person’ differences shown in Figure 17. This is unsurprising because the probability of a person changing their voting preferences is smaller than the probability of differences existing between people. However, this does not mean that the ‘within-person’ effects identified in Figure 12 are any less consequential.

By revealing the role played by economic insecurity in driving voters away from Labour to Reform, the Lib Dems and ‘undecided’ since the 2024 election, and the role played by immigration attitudes in driving voters away from Labour to Reform, they provide convincing evidence of the factors explaining Labour’s dwindling support.

Immigration attitudes drive Conservative losses

In this section, we repeat the above analysis for Conservative defections, examining how economic insecurity and immigration attitudes explain Conservative defections since the 2024 election. We concentrate on the 2 destinations to which the Conservatives have experienced their most numerous losses since 2024: Reform UK and ‘undecided’. Among 2024 Conservatives, 20.8% defected to Reform UK by April 2025, 13.8% to ‘undecided’, and 60.8% still intended to vote Conservative.

Figure 20 presents predicted probabilities for Conservative defections to Reform and ‘undecided’ based on economic insecurity and immigration attitudes, the equivalent relationships to those regarding Labour defections that we presented in Figure 18. In other words, we test whether being economically insecure or opposed to immigration is associated with abandoning the Conservatives since the 2024 election. The analysis for economic insecurity shows that it has no relationship with Conservative losses: the insecure are just as likely to defect to Reform or ‘undecided’ as the economically secure, as shown by the small and non-significant differences for each destination in the left-hand plot.

In contrast, immigration attitudes are strongly related to Conservative defections. The right-hand plot in Figure 20 shows that, since the most recent election, 2024 Conservative voters who are opposed to immigration are much more likely to have defected to Reform UK than those who are supportive of immigration (by 16.8 percentage points), and less likely to have switched to ‘undecided’ (by 6.28 percentage points).

We suggest that anti-immigration voters are less likely to switch to ‘undecided’ because Reform is picking up this support. The underlying models used to calculate these predicted probabilities and a detailed summary of our analysis are presented in Table 10 of the Appendix.

In section A5 of the Appendix, we present additional tests of these relationships. Here, we show that the relationship between opposition to immigration and defecting to Reform among 2024 Conservative voters is robust to the inclusion of all additional controls (as described above), whereas economic insecurity continues to be unrelated to Conservative defections.

It is not surprising that voters are not defecting from the opposition party because they feel economically insecure, partly because the Conservatives already lost voters in 2024 due to feelings of economic insecurity, and partly because we would expect vote losses on the economy to affect the governing party at the time. However, it makes sense that the Conservatives are losing voters on the issue of immigration when faced with a party on the right that is growing in viability (the perception that it is a winning force) and competing on the issue of immigration.

We also note that we observe no within-person effects of either economic insecurity or immigration attitudes on Conservative defections. Given that voters on the right are likely already strongly sorted on immigration attitudes, we view the absence of within-person effects for immigration attitudes as unsurprising.

Conservative losses to Reform dangerous for Labour, with Labour’s splintering support

One reason we focus on both political parties is because their fates are so intertwined. It would be simple but misleading to focus just on one issue — economic insecurity and handling of the national economy — and to focus only on Labour’s losses. This would place insufficient focus on the other sides of Labour’s difficulties, and insufficient focus on the remarkable volatility currently happening in British politics, in which we see both parties of government or recent government losing electoral support.

A focus on economic insecurity would certainly make Labour’s problems less severe, and it would do so through 4 routes, as outlined above, both direct and indirect. However, that should not ignore the role of immigration on Labour’s 2024 voters, which is also significant, although we also know that greater economic security would have the effect of dampening the impacts of Labour’s handling of immigration.

More broadly, the issue of immigration and Conservative losses are both headaches for Labour, as well as a headache for the Conservatives. Labour won its large majority in 2024 in large part because the right vote was split between the Conservatives and Reform UK. But if right-wing voters’ current abandonment of the Conservative Party for Reform continues, Reform can overtake the Conservatives and overtake Labour in almost 200 Labour constituencies where Reform is either second or in close contention (Green and Miori, 2025).

With fragmentation and all parties being on lower vote shares, being in ‘close contention’ is made far easier and much more frequent. This is made even more likely because of Labour’s broad splintering, and that can be best explained by economic insecurity as Labour’s electoral foundation, which should not be wholly separated from immigration.

Figure 21 indicates that economic insecurity is positively associated with defecting from Labour. Across all 4 party destinations — Reform, the Greens, Lib Dems and ‘undecided’ — economically insecure individuals are more likely to have defected from Labour than those who are economically secure (the plots are labelled with the difference in the probability of defecting between the insecure and secure). These differences are statistically significant in all models, as indicated by the asterisks.

Next, to test whether the associations between economic insecurity and Labour defections are not driven by unobserved factors, we estimate additional models that control for sociodemographic characteristics (education, age, income and gender) and perceptions of the national economy (measured by perceptions of GDP growth), alongside those covariates included in the baseline model – immigration attitudes and wave fixed effects. We present the results of these models in Table 3, alongside the results of the from the ‘baseline’ model used to calculate the predicted probabilities in Figure 21.

As economic insecurity is standardised before inclusion in the model, the coefficients represent the estimated effect of a 1 SD increase in economic insecurity. As these results indicate, the relationship between economic insecurity and Labour defections remains robust to the inclusion of these additional controls. Across all models, economic insecurity is positively and significantly associated with Labour defection to all destinations, although the relationship between economic insecurity and defections to undecided is only marginally significant when controls are included.

Results are LPM coefficients from models with random effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1SD increase in the predictor). Robust standard errors are provided in parentheses.

*** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

The models presented in Table 3 also estimate the relationship between immigration attitudes and defections from Labour. As with economic insecurity, our measure of immigration attitudes is standardised before inclusion in the model and therefore represents the estimated effect of a 1 SD increase in this variable, allowing us to compare the influence of economic insecurity with that of immigration attitudes on the probability of defection. As we discuss in the main report, the relationship between immigration attitudes and defections from Labour is asymmetric, varying across party-destinations.

Among 2024 Labour voters, opposition to immigration is positively associated with defecting to Reform and ‘undecided’, and negatively associated with defection to the Greens. In other words, 2024 Labour voters who are opposed to relaxing immigration restrictions are more likely to now support Reform or be ‘undecided’ in their vote choice, and are less likely to now support the Green Party.

These relationships are robust to the inclusion of additional controls: education, age, income, gender and perceptions of GDP growth. Results therefore suggest that opposition to immigration has substantial but partial consequences for defections from Labour, helping explain defections to Reform and to ‘undecided’, but not to other parties.

The fact that results are consistent across models which include additional covariates provides further support for the claim that both economic insecurity and immigration attitudes have ‘direct effects’ on defections from Labour. To provide an additional test of these relationships, we estimate the within-person effects of economic insecurity and immigration attitudes on Labour defections to Reform, the Greens, the Lib Dems and ‘undecided’. We do so by estimating linear probability models for each specific party-destination that include both individual and wave fixed effects.

These models use data from all 3 JRF waves, which allows us to measure within-person change in vote choice (that is, defections) from the 2024 election to October 2024 and to April 2025.¹⁴ The results are presented in Table 4. Coefficients represent the estimated change in the probability of defection associated with a 1 SD increase in economic insecurity or immigration attitudes.¹⁵

Results are LPM coefficients from models with individual fixed effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 SD increase in the predictor). Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05.

Results indicate that within-person change in economic insecurity is positively and significantly associated with defections to Reform, the Lib Dems and ‘undecided’. In other words, as 2024 Labour voters become more economically insecure, they become more likely to defect to these 3 party-destinations (but not to the Greens). In comparison, within-person change in immigration attitudes is positively associated with defections to Reform, but not to any other party-destinations.

As individuals become more opposed to immigration, they become more likely to defect from Labour to Reform, but not to other parties. Within-person effects are therefore consistent with the overall trend observed in previous models. The consequences of economic insecurity for Labour are broad, with our results indicating that insecurity leads voters to abandon Labour for parties across the political spectrum. Immigration attitudes, in contrast, have narrower implications, prompting former Labour voters to defect to Reform (and possibly to undecided) but not to parties on the ideological left.

By analysing the effects of within-person change in economic insecurity and immigration attitudes on defections, we are able to rule out all alternative explanations that are stable over time (‘time-invariant confounding’), such as values, personality, family influences and many sociodemographic characteristics. This provides stronger evidence that the relationships between (a) economic insecurity and defections across party destinations and (b) immigration attitudes and defections to Reform represent causal effects.

Section A2: ‘Indirect effect’ of economic insecurity on Labour’s handling of the national economy

First, we examine the relationship between economic insecurity and economic handling evaluations at the within-person level. To do so, we fit 2-way fixed effects models (wave and individual fixed effects) using data from JRF Waves 2 and 3. These models provide an assessment of whether within-person changes in economic insecurity from October 2024 (Wave 2) to April 2025 (Wave 3) predict within-person changes in economic handling evaluations over the same time period.

The results, which are presented in Table 5, indicate that economic insecurity is positively and significantly associated with handling evaluations. In other words, as individuals become more economically insecure, they become more likely to evaluate Labour’s handling of the economy as negative.

Results are LPM coefficients from models with individual fixed effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 SD increase in the predictor). Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05.

Next, we assess the question of causal direction. While the results presented in Table 5 provide evidence of a within-person relationship between economic insecurity and economic handling, they do not inform us about the causal ordering of this relationship — whether changes in economic insecurity then lead individuals to update their handling evaluations, or whether changes in handling evaluations lead to changes in economic insecurity. To disaggregate between these 2 possibilities, we fit cross-lagged panel models (CLPMs) and random-intercept cross-lagged panel models (RI-CLPMs) to assess the bidirectional relationship between economic insecurity and evaluations of government economic handling.

Due to reasons of data availability (RI-CLPMs require 3 or more waves of data), we have to fit these models using data from the British Election Study. Specifically, we use data from Waves 24 (December 2022), 25 (May 2023) and 26 (May 2024), meaning that these models assess the relationship between economic insecurity and government economic handling during a period of Conservative government. The sample used in these models is restricted to those individuals who provide multiple observations for both the economic handling measure and economic insecurity measure (that is, those who provide data for both measures at 2+ waves).

Figure 22 plots the results from these models. Specifically, we display the autoregressive paths (that is, the relationship between one measure at wave T and the same measure at the next timepoint) and the cross-lagged paths (the relationship between one measure at wave T and the other measure at the next timepoint). The cross-lagged paths provide us with the necessary information to understand causal direction. The upper panel plots the CLPM results, and the lower panel plots the RI-CLPM results. Statistically significant paths are drawn in black, with non-significant paths drawn in red.

CLPMs can be interpreted as modelling between-person effects. In other words, they test whether an individual who is economically insecure at one wave is estimated to evaluate government handling of the economy as negative at the next wave, and vice versa. Results from the upper panel indicate first that lagged economic insecurity has a positive and statistically significant association with negative evaluations of government handling of the economy. In the reverse direction, lagged evaluations of economic handling are positively associated with economic insecurity, although the size of these coefficients is somewhat smaller.

The RI-CLPM can be interpreted as modelling within-person effects. In other words, this mode tests whether an individual who becomes more economically insecure at one wave (a within-person change) then adopts a more negative evaluation of government economic handling at the following wave (a within-person change). As the RI-CLPM focuses on within-person effects, it provides a more causally robust estimate of the relationship between variables.

As results from the lower panel show, the only statistically significant cross-lagged path is from economic insecurity in May 2023 to government handling evaluations in May 2024. This provides support for the claim that economic insecurity is causally prior to evaluations of government handling of the economy.

Section A3: The ‘indirect effect’ of economic insecurity on the amplification of effects of Labour’s handling of the national economy on Labour defections

Table 6 presents results from models which test the ‘indirect effect’ of economic insecurity as a moderator of the relationship between evaluations of Labour’s handling of the national economy and Labour defections. The models presented in this table make use of data from JRF Waves 2 and 3, with the sample restricted to 2024 Labour voters. Specifically, we estimate random effects linear probability models for each party-destination: defecting to Reform, the Greens, Lib Dems and ‘undecided’.

The outcome variable in each model is a dummy coded 0 if the respondent reports the continued intention to vote Labour at the corresponding JRF wave and coded 1 if the respondent reports the intention to defect to the party of interest. Standard errors are clustered by individual to account for the non-independence of observations.

To assess whether economic insecurity moderates the relationship between evaluations of Labour’s handling of the economy and defecting, we include an interaction term between economic insecurity and our measure of economic handling evaluations, alongside the standalone terms for both economic insecurity and handling. We report results for each party-destination from a ‘baseline’ model that controls for immigration handling and wave fixed effects, and a ‘covariate’ model that also controls for sociodemographic characteristics (income, gender, age and education) and perceptions of GDP growth.

As the insecurity and economic handling variables are standardised, the standalone coefficients for insecurity and handling represent the estimated effect of the corresponding variable when the other variable in the interaction is held at its mean, providing meaningful interpretations of the standalone effects of insecurity and economic handling evaluations.

Results are LPM coefficients from models with random effects. Coefficients represent standardised effects. Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

As indicated by the economic handling * insecurity interaction term presented in Table 6 above, economic insecurity amplifies the relationship between evaluations of Labour’s handling of the economy and defections from Labour to Reform, the Greens and Lib Dems. These results remain robust to the inclusion of controls (that is, the results in the ‘Covariate’ columns), although the interaction term in the Lib Dem model drops to p < 0.1 when covariates are included. These interactions are displayed as average marginal effects in Figure 15 of the report.

To provide a more robust test of this ‘indirect effect’, we test whether economic insecurity moderates the within-person effect of economic handling evaluations on Labour defections. To do so, we fit hybrid models (AKA random-effects within-between models; Bell and Jones, 2014). Hybrid models are, in effect, random effects models in which time-varying covariates are decomposed into within-person and between-person components (Bell, Fairbrother and Jones, 2019).

We take the economic handling scores for each individual at Waves 2 and 3 of the JRF panel and decompose them into a person-specific mean (which captures the between-person component of the economic handling measure) and a change score, which captures the deviation from that person-specific mean at each wave (the within-person component of the economic handling measure) (Howard, 2015). The within-person coefficient is equivalent to that from a model including fixed effects, allowing us to rule out all time-invariant confounding in the estimation of the within-person effect of economic handling.

To estimate the indirect effect of economic insecurity as a moderator of economic handling evaluations at both the between-person and within-person level, we include interaction terms between economic insecurity and both the between-person and within-person component of economic handling.

We follow the same estimation strategy as for the models presented in Table 6, first estimating a baseline model that only controls for wave fixed effects, before next estimating a covariate model that controls for sociodemographic characteristics and GDP perceptions. We present the results of these models in Table 7. As the economic handling measure is standardised, the within-person coefficients represent the estimated change in the probability of defecting associated with a within-person increase of 1 SD in the handling measure.¹⁶

Results are LPM coefficients from hybrid models. Coefficients represent standardised effects. Robust standard errors in parentheses. *** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

Results indicate that economic insecurity moderates the between-person effect of economic handling evaluations on Labour defections to Reform, the Greens and Lib Dems; and moderates the within-person effect of economic handling evaluations on Labour defections to Reform and ‘undecided’.

However, to better understand these interactions, we calculate and plot the average marginal effects of a one-standard deviation increase in the economic handling measure (corresponding to a more negative evaluation of Labour’s handling of the economy) among the secure and insecure (±1 SD from the mean on the economic insecurity scale). Given our interest in the within-person relationships here, we only plot the within-person effects. These AMEs are presented in Figure 23.

The above plot shows the differences in the within-person effects of economic handling evaluations between the secure and insecure. Among the economically insecure, a within-person increase in negativity about Labour’s handling of the national economy is positively associated with the likelihood of defecting to Reform, the Greens and ‘undecided’ (although the difference in AMEs for Green defections is not statistically significant).

Among the economically secure, within-person changes in evaluations of Labour’s economic handling do not increase the likelihood of defection, and are negatively associated with the likelihood of defecting to Reform. These results provide further support for the argument that economic insecurity amplifies the consequences of negative evaluations of national economic handling on Labour defections.

Section A4: ‘Indirect effect’ of economic insecurity on the amplification of effects of Labour’s handling of immigration

In this section, we test the ‘indirect effect’ of economic insecurity as a moderator of the relationship between evaluations of Labour’s handling of immigration and defecting from Labour since the 2024 election. We follow the same strategy as that used to assess whether insecurity moderates the effects of economic handling evaluations, with one exception models now include the interaction term of economic insecurity and our measure of evaluations of Labour’s handling of immigration, rather than the economy. We present results from both the baseline and covariate models below in Table 8. The interactions from the baseline models are displayed as average marginal effects in Figure 16 of the report.

Results are LPM coefficients from models with random effects. Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

As indicated by the immigration handling * insecurity interaction term, economic insecurity amplifies the relationship between evaluations of Labour’s handling of immigration and defections from Labour to Reform and the Lib Dems, but not to the Greens or ‘undecided’. These results remain robust to the inclusion of controls (that is, the results in the ‘Covariate’ columns).

To provide a further test of this interaction, we assess whether economic insecurity moderates the within-person effects of immigration handling evaluations on Labour defections with a hybrid model. As with the hybrid modelling process described in section A3, we decompose our measure of evaluations of Labour’s handling of immigration into between-person and within-person components and estimate a model that includes interaction terms between economic insecurity and both the between-person and within-person scores.

Results from these models are presented below in Table 9. As the immigration handling measure is standardised, the within-person coefficients represent the estimated change in the probability of defecting associated with a within-person increase of 1 SD in this measure.¹⁷

Results are LPM coefficients from hybrid models. Coefficients represent standardised effects. Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

As can be seen in the above table, the within-person effect of immigration handling is not statistically significant for defections to any party, nor is the interaction term of economic insecurity and the within-person effect of evaluations of Labour’s handling of immigration. This is consistent with the idea that any defections that occur as a result of evaluations of Labour’s performance on immigration are driven by sorting on immigration. In other words, rather than a change in immigration evaluations leading to a change in party preference, voters are instead changing parties to ensure alignment with their (stable) immigration evaluations.

Section A5: Economic insecurity, immigration attitudes and Conservative losses

We fit random effects linear probability models to assess whether economic insecurity and immigration attitudes explain Conservative losses since the 2024 General Election. We focus exclusively on defections to Reform and to ‘undecided’, because the vast majority of Conservative losses since 2024 have been to these 2 destinations (of 2024 Conservative voters in our data, 60.8% still intended to vote Conservative by April 2025, 20.8% had defected to Reform and 13.8% had defected to undecided).

These models use data from JRF Waves 2 and 3, restricting the sample to 2024 Conservative voters. The outcome variable in each is a dummy which captures defecting from the Conservatives to the party-destination of interest (for example, the outcome variable in the ‘undecided’ model is coded 0 if the respondent reports the continued intention to vote Conservative at the corresponding JRF wave and coded 1 if the respondent reports that they are now undecided).

The baseline model controls for wave fixed effects only; the covariate model adds sociodemographic controls (age, gender, education and income) alongside controlling for GDP perceptions. Standard errors are clustered by individual to account for the nonindependence of observations.

The measures of both economic insecurity and immigration attitudes are standardised before inclusion in the model, meaning that the coefficients presented in Table 10 represent the estimated effect of a one-standard deviation increase in the corresponding variable. These Conservative defection models thus mirror those used to model Labour defections in section A1 and the earlier part of the report.

Results are LPM coefficients from models with random effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 SD increase in the predictor). Robust standard errors are provided in parentheses.

*** p<0.001, ** p<0.01, * p<0.05, ✝ p<0.1

Results indicate that economic insecurity is unrelated to defections from the Conservatives to either Reform or ‘undecided’. In contrast, opposition to immigration is positively associated with defection to Reform and negatively associated with defection to ‘undecided’.

In other words, 2024 Conservative voters who oppose relaxing immigration restrictions are more likely to have now defected to Reform and less likely to have defected to ‘undecided’, relative to those with more supportive attitudes toward immigration. Results from the baseline model are used to calculate the predicted probabilities displayed in Figure 11 of the report, but Table 10 shows, results remain broadly consistent when additional controls are included.

Following the approach used to examine the importance of economic insecurity and immigration attitudes to defections from Labour, we also test the within-person effects of economic insecurity and immigration attitudes on defections from the Conservatives to Reform and ‘undecided’ since the 2024 election. To do so, we estimate linear probability models for Reform defections and ‘undecided’ defections that include both individual and wave fixed effects.¹⁸

The results are presented in Table 11. Coefficients represent the estimated change in the probability of defection from the Conservatives associated with a 1 SD increase in economic insecurity or immigration attitudes, at the within-person level.

Results are LPM coefficients from models with individual fixed effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 SD increase in the predictor). Robust standard errors in parentheses.

Results indicate that, at the within-person level, neither economic insecurity nor immigration attitudes are associated with Conservative defections to Reform or ‘undecided’. The absence of within-person effects for immigration attitudes on Reform switching can perhaps be explained by a sorting of the British electorate on immigration attitudes (and Brexit preferences, with which they are strongly correlated).

Switching from the Conservatives to Reform is switching within the right-conservative bloc of British party politics, and individuals within this bloc likely already hold anti-immigration views (Griffiths, Perrett et al., 2025). Switching from the Conservatives to Reform is therefore potentially driven not by a change in immigration attitudes, but by a change in the party that an individual perceives as best representing their immigration attitudes and policy goals.

Section A6: Relationship between economic insecurity and immigration attitudes

Over the course of this report, we have analysed economic insecurity and immigration attitudes as independent, side-by-side influences on voting; and tested whether economic insecurity amplifies the relationship between immigration handling evaluations and vote choice. However, based on evidence from several studies that an individual’s personal financial circumstances have little direct influence on immigration attitudes (Hainmueller and Hopkins, 2014), we do not anticipate that economic insecurity would directly influence immigration attitudes. In this section, we test this expectation.

To do so, we analyse the within-person effect of economic insecurity on attitudes toward immigration using all 3 waves of data from the JRF Panel. Our measure of immigration attitudes – the outcome variable in these models – is the 5-point scale which asks respondents how much they support ‘Making it easier for immigrants to come to Britain to work’, from ‘Strongly oppose’ to ‘Strongly support’. We regress this measure on economic insecurity, along with wave and individual fixed effects. We estimate this model with 2 samples: first, all JRF respondents; and second, 2024 Labour voters. The results are presented in Table 12.

Results are LPM coefficients from models with individual fixed effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 standard deviation increase in the predictor). Robust standard errors in parentheses.

*** p<0.001, ** p<0.01, * p<0.05.

As these models estimate within-person effects, they allow us to understand whether individuals become more opposed to immigration as they become more economically insecure. In both models, the coefficients for economic insecurity are near zero and not statistically significant. Thus, among both the total JRF sample and Labour voters exclusively, within-person changes in economic insecurity have no effect on immigration attitudes, indicating that people do not become more hostile toward immigrants as they become more economically insecure.

Next, we test the relationship between economic insecurity and evaluations of Labour’s handling of immigration at the within-person level. In other words, we test whether individuals evaluate Labour’s handling of immigration as more negative when they become more economically insecure. To do so, we regress evaluations of Labour’s handling of immigration on economic insecurity, along with wave and individual fixed effects.

As with immigration attitudes, we estimate this model with 2 samples. We begin by testing it with all JRF respondents and then testing the same relationship among 2024 Labour voters. As the handling measure is only included at Waves 2 and 3, these are 2-wave models. The results are presented in Table 13. As with immigration attitudes, we find no evidence that within-person increases in economic insecurity lead to more negative evaluations of Labour’s handling of immigration.

Results are LPM coefficients from models with individual fixed effects. Coefficients represent standardised effects (that is, the change in the probability of defection associated with a 1 standard deviation increase in the predictor). Robust standard errors in parentheses.