I don’t think anybody would accuse them of such, but polling firms don’t exactly have an easy job; in fact it’s quite a tricky business. Even before the scrutiny of the public eye (and the expectations coming with it) setting up and conducting a survey is full of possible pitfalls that can lead even the most honest pollster astray. I’ve often seen, heard and - if I’m honest - experienced the complete shock of voters as one side won while pointing at surveys showing an opposite result. So instead of taking these polls on face value, but without diving into the broader societal/political forces at play, let’s explore the challenges a survey has to deal with to even produce surface level numbers - starting with the fundamentals.

Why Polling is Hard

While the last few years saw an increase in online polling of public sentiment – whether it’s about elections, policies or cultural shifts and trends - the polling industry itself is still slow to change its practical approaches and is persistent in using more traditional methods we might picture. This presents a growing challenge with generational culture shifting around anonymous calls, decades of experience with various robocalls, spam and so forth. With the younger generations preferring other methods of communication, it is increasingly difficult reaching a statistically valid sample size, as depending on who answers, it’s easy to slip into various domains of sampling bias. Now we will focus on how polling firms reach (from their point of view) a scientifically satisfactory sample and how the environment is created to get to the point which is interesting for the public: the numbers and line charts themselves.

Dialing In

The first thing a responsible and adept pollster must settle is who they are going to ask. As asking the entire population (relevant to the topic of the poll) is reasonably impossible, the main task is getting a workable slice that contains all the elements of the full picture - that is to say, a representative sample - which reflects the characteristics and can be used to predict the behavior of the larger population. For the purpose of thematic consistency, let's take an election poll as an example. In this case, the full population we can't reasonably reach is the entire electorate, every person who legally has a vote. We still need to get data we can use to measure support for one party or the other, so for the sample to be representative we need to consider all the nuances (or as many as possible) that make up the actual electorate. This means making sure we have enough responses from - preferably roughly following the societal distribution on the ground - groups such as:

• genders

• different age groups and generations

• regions and counties; religions and ethnicities

• education level and occupation (in broad categories).

This distribution is often based on census or other statistical data but it's ultimately up to the polling firm to use the sources they prefer.

Some, like Quinnipiac and Gallup, use random digit dialing which is exactly what it sounds like, calling random phone numbers. Others use mixed methods, such as Ipsos sets up participation groups (called panels) based on cold-calling and mail invitations or Rasmussen uses - essentially - interactive robocalls. Traditional methods are very much in play, in fact, 6 of our 10 core pollsters still extensively use it. However, we cannot pass by the fact that these surveys often require serious weighting to counteract the low(er) response rates. In practice this can mean that if younger generations are less inclined to pick up the phone, the responses of those who do, have to be given more weight to represent their proportion in the electorate. This doesn't mean they have to be inaccurate as (based on our own analysis) Gallup boasts a mere 2.2% error rate, but demographic corrections have to be applied precisely otherwise the results can be heavily skewed.

Another important element is whether they (technically) start fresh every survey, or do they poll a group of participants repeatedly. The latter option, which is called a panel, and it can create huge datasets like YouGov's MRP methodology, but it can also require considerable weighting as the responders are not filtered based on representative criteria.

A common way of normalizing data on the sample level is introducing quotas, meaning that pollsters will try to reach a predetermined amount or ratio of participants, such as based on the age distribution of the electorate. A number of factors can influence the accuracy of a sample built with quotas, such as whether the participants were selected randomly, by the interviewer, or if they are part of a panel as any of these can introduce their own set of human- or chance-based biases. Using quotas seems to yield positive results though, as YouGov's error rate is only 1.5% in the US (though a high 3.5% in Europe) and Morning Consult is also at the top of our chart with only a 1.8% aggregated error rate!

More complex (and more expensive and difficult) methods can create samples that are 'out-of-the-box' more representative, but these are also reasons why they are more rare. A combination of effective sample creation and weighting is nevertheless practically always needed, so the method a firm uses doesn't inherently determine its value.

Almost needless to say, there are many different ways and combinations of sampling and it's entirely up to the pollster to decide how to approach it.

We also need the responses tagged ('cross-tabulated' or 'crosstabbed', when responses are broken down by the different demographic groups, something polling enthusiasts often want to see as soon as possible) so we can set combinations, like college-educated middle-aged women from the capital; under-25 blue-collar men from rural towns; or retirees in this or that side of the country and so forth as there might be stark differences between how these different 'labels' create a vast diversity of perspectives. Polling firms usually target at least 1000 responses (which is the academically accepted minimum and a balanced between statistical significance and practical feasability); some might have more but that doesn't necessarily translate to better results (YouGov's MRP, based on limited data has a 2.5% error rate which is rouhgly around average). As we established, because of the amount of responses and holes in the sample, we'll have to introduce corrections as some groups may only have a few individuals. The exact methodology of the weighting (sometimes even whether they do it at all) depends on the polling firm. The general idea however, is to make sure responses represent the actual presence of each demographic group - at which point pollsters also have to be careful not to over-weigh a group and disproportionately distort the results (more on this in the next post of the series). Now classic examples of this step going awry are the recent oversampling and over-weighting of college-educated voters, leading to - based on the polls - some surprising results from and around 2016 (a real low point for the industry as most people would note aptly). Sampling and weighting errors are of course not the only reasons for inaccuracies in, for example the polling around Brexit, and we'll dive into concepts such as biases, contradictions, 'shy voters' and 'disappointed voters' (both mythical and actual) elsewhere.

Reaching Out

Once the sample creation method is settled, a pollster also has to decide how they reach out to people. This intersects with the sample creation to a degree, as random digit dialing can often already lead to recording the responses via telephone. This method is very common, and in Hungary for example, most firms rely heavily on phone calls. The other most common method is web-based surveys, used exclusively (such as YouGov or Morning Consult) or in tandem with phone calls (like Kantar or Ipsos). There are still some though who rely on face-to-face interviews (Eurobarometer for example), but for national-level polls it's by far the rarest.

So there we have it - one step closer to a thorough understanding of how polling works, and consequently, where it can fail. However, even with all possible tinkering with sample setups; experimenting with contact methods and tweaking the weighting formulas, the data ultimately still depends on the people polled. Whether they respond at all, whether they are honest with themselves and the interviewer, how much they trust the pollsters, or if they have contradictory preferences shape the data fundamentally. Join us next time, when we'll examine how and what kind of questions are asked and what can be done there to further refine the final numbers.

Zohran Mamdani won the New York mayoral election on the back of progressive policies, such as fare-free buses, rent freeze (which is not unprecedented) and in tandem with the latter investing into increasing supply by building subsidized housing. Of course, questions can be raised: how feasible are these policies, and what practical effects might they have, regardless of intent? While the first question would probably require an evaluation of NY budget and state/federal contributions, mayoral powers and Rent Guidelines Board collaboration, so we’ll instead focus on comparing them to similar policies enacted elsewhere. As he’s not inaugurated yet, we’ll obviously only evaluate his campaign talking points and their feasibility.

Ireland Rent Pressure Zones: Rent Control Without Freeze

Since 2016 Ireland introduced rent control, though not rent freeze, in selected areas of the country, and from 2025 Rent Pressure Zones - where this policy is in effect - were extended to the entirety of Ireland.

Background

In a nutshell, in the early 2000s the Irish economy became over-reliant on its construction industry. Housing prices increased through the roof in the late 90s; followed by tax relief and incentives granted leading to a dramatic increase of housing supply (read more in the Rural Renewal Plan). At its height in the mid 2000s, 12% of the country’s workforce was employed in the construction sector. By 2007, rent prices started to fall, continuing into 2008 (see below) to yields as low as ~3% - comparably renting yields are around ~7% in 2024 and 2025. Demand for property started to decline by 2007, then the bubble burst by 2008-2009 and rent prices fell by ~25% (see below, ~1600EUR in Dublin 16 fell to ~1200EUR by 2010).

As the property market and the construction industry crashed, supply started to decline and as such rent started to ‘bounce back’ as well - noted as early as 2014.

The RPZs

Originally introduced in 2016, a few years after rent prices started their sharp hike up. These were designated areas where rents could not be increased by more than general inflation, or 2% if inflation was higher. A criterion for the designation was that rent increases in the area must have been more than 7% annually in at least 4 of the last 6 quarters. As noted in the 2014 article linked above, that was already the countrywide average, with Dublin providing the brunt of it - see the 3 Dublin areas on the figure below.

Since 2025, all of Ireland is an RPZ, which means rent control rules essentially following the guidelines above.

Between 2016 Q1 and 2017 Q1 rent prices in inner Dublin (Donnybrook in our example) increased by 11%; and by more that 5% in Q1 2021 Q1 and 2022 Q1.

In Naas, a commutable distance city outside of the Dublin metro area (designated RPZ since 2017), where rent prices increased 8% from 22 to 23, and then by another 15% from 23 to 24.

There are of course exceptions when prices can exceed the RPZ rules, such as longer periods of not being rented or substantial changes (such as extensions or large scale renovations) and new building without a history of previous rent prices to be controlled, which could account for some of the periods of steeper increases.
Naas could also be explained by the general trend of moving away from larger cities for ( sometimes theoretically) cheaper housing.

All in all, while it seems like the policy is not ironclad, despite some spikes I’d still deem it successful in flattening - or even breaking - the pattern of consistent increases in rent prices at least compared to the yearly 7% average increase before. Some trends and effects on the housing market ultimately can’t be frozen or controlled completely and when looking at data such as rent prices here, other factors (like exceptions and new supply) must also be taken into account.

Hungary Housing Market: Subsidized Loans

Hungary has multiple types of subsidized loans and mortgages introduced over the years. The first one currently available was introduced in 2015 theoretically to support new families based on the number of new children planned (past children do not count) - sadly data is not available for 2015 but based on later data, a price growth rate of 15% is still considered high comparatively.

One of the latest of such was launched in 2019 summer. This loan grants families who promise to have children a maximum of 10 million HUF (~25k EUR) loan with subsidized repayment conditions (assuming they actually follow through otherwise the subsidies must be repaid). The loans started taking effect in Q4 2019 and we can see the increase in price growth EoY despite increase supply compared to 2018.

The most recent loan option is about lowered interest rates for first time homeowners and while 2025 data is not fully available yet, we can see another sharp increase rate for price growth. The construction of new dwellings also decreased, but I imagine it will start to increase a bit going forward as buyer intent rises. Based on historical trends supply will likely also increase going forward.

There are some other longish running programs and shorter bursts of subsidies but these are the most relevant.

Other things I have to call out:

• the latest price growth rate (at least) is not just alarming in terms of local currency context as according to the Hungarian National Bank (the primary data source for this item) is “an outstanding rate even by European Union standards”

• the rates displayed are country-wide, so the 2021 deep-Covid price growth record was also driven by the spike in countryside property, not the capital itself which only saw a 12% increase (in the Nat. Bank’s words “This was primarily due to the outstanding price increase in the countryside”). A personal side note is that home office work became the standard (during those years) which is why a lot of people took the opportunity of moving away from the capital.

• I also acknowledge that there are other factors in play as well, such as inflation, but that was only 5% officially (though the Hungarian CSO has a tendency to skew data a bit), and growth on the housing market eclipses that number. That leads me to believe that people see property more and more as an inflation proof investment- though that would require a dedicated analysis.

My reading of programs such as subsidies making housing more affordable is that if the supply side is left to the private sector, the subsidies will just be built in the sale values by the private sector without moving the affordability meter much - I could go as far as saying these are, in a way, often subsidies for the sellers, not the buyers (not going into it in details but the HNB also measures overvaluation of housing properties in the summaries, linked in the chart caption below). This is not to say that it’s completely ineffective, done right I imagine it can reach positive results but based on this case study, it needs to be planned and executed with a lot of care and oversight. Ideally construction and ownership of new dwellings being party handled by the government introducing the policy so prices can be first-party controlled.

Budapest Public Transport: Discounted and Free Passes

The full-price adult monthly pass provided by the Budapest Public Transport Service (BKK) costs 9k HUF. Taking the national average net wages of 470k HUF in the business sector (not published separately for Budapest, as per CSO Hungary), making it an already inexpensive way of commuting in the city. The passes with the usual discount for students, pensioners cost 3,5k HUF, about a third of that amount.

Since 2019 there are new discounted tickets:

• parents with small children only need to buy a student-price pass, down from a previous full pass price

• children under the age of 14 travel for free, down from the previous student pass

These are considerable price subsidies as a lot of commuters in the city primarily use public transport, and metro area car-using families also often leave their vehicles in parking lots somewhere in the city or around its limits switching to public transport options.

Standard single ticket prices were also frozen for a long time without increasing with inflation.

A recent (since 2021) change is free passes being offered for children under the age of 14; and because of a competition between the state railway and the BKK, recent tariff changes included the a decrease in price of the monthly Budapest passes from 10k to 9k (rounded). This competition also resulted in a complex web of intertwined revenues with BKK and state railway passes being purchasable through each other’s services and as such both gaining a portion of the profits of their passes bought on the other’s platform.

Anyway, while service is not free of course, these were significant shifts in pricing, reducing revenue of the holding company. Below is a comparison of BKK revenues and expenses over the last 6 years (since the current opposition supported mayor holds office), where we can see that:

• Total aid relatively consistently increased despite the recent disputes over funding with the government; this is mainly used for modernization, renovation and other development efforts, not plugging holes in the budget. This however also means that the BKK’s own capital is (mainly) not used for these purposes. This is relevant because it shows that possible holes are not plugged by specifically through external financial support.

• Total transportation costs increased despite the new discounts and price decreases, which is offset partly by hiking all other non-pass ticket types (single ticket prices increased by 60% for example in the last ~5 years after being frozen for a decade before)

• Total revenue taking a hit from 2020 and more sharply from 2021, but this period coincided with the Covid lockdowns, home office being the standard for most workplaces (for years) and thus less commuting. Tourism also plummeted in 2020-2021 though interestingly that didn’t seem to affect total transportation revenue for that year

• Total revenue rebounding to positive for the first time in the last 6 years

• But own capital decreasing steadily and losing a large chunk in 2021 - again coinciding with Covid and the new subsidies which suggests losses being offset by a loss of capital. This is something to keep an eye out for in the future, but the recent increase in transportation revenue and the positive closing balance doesn’t lead me to be alarmed of the financial health of the BKK.

Looking at other metrics, we can see a similar picture. Capital decreasing by 32%, but ROE rising to positive in 2024 for the first time in the period. The increasing debt ratio together with the decrease of capital also suggest increased costs needing to cover other expenses without overloading the budget. Coming out from the Covid period downturn and the new tariffs and discounts with a marginal increase of debt ratio (compared to just before Covid) is a good sign as there seems to be a lack severe adverse effects of discounted pass pricing where the beneficiaries are primarily the residents of the city.

As before, capital decreases might need to be addressed in due time if the trend continues.

So newly discounted and in some cases downright free passes didn’t seem to leave a dent immediately in the profitability or revenues of the BKK, as the decreases mainly occurred during the Covid years. Since then both revenues and profitability have risen suggesting that these discounts are not detrimental to the BKK’s financial health. Capital decrease is, again, something to monitor but I wouldn’t relate that to the discounts/subsidies (as it’s mainly driven by amortization of assets, and personnel related costs as per the published yearly summaries) given the transportation revenues being constant or increasing.

Conclusion

Some of these comparisons of course are green apples to red apples. Every city is different, and naturally NY is a whole different dimension than Budapest or Dublin, with different means, different tools and different finances. They all face similar challenges though, out of control rent prices, inadequate amount of housing, and a call for cheaper public transport options. While the scale is obviously different, these stories from cities that are comparatively similar in their own countries to New York tell us that a radical 100% implementation of Mayor-Elect Mamdani’s policies is not a given, they might be (might have to be) watered down.

Overall however, they might just work if the commitment is there.

In September 2025, the UK, Canada, Australia and France announced the recognition of the statehood of Palestine. Alongside the obvious political implications of the move, this decision makes the support for a two-state solution a practical reality for these states (compared to the previous rhetorical/political reality). But this is also an interesting event that signals a break from US policy on this matter, and gives the UK and France a level of agency and independence in Middle-Eastern foreign policy not seen in decades. This is more than just a shift - it’s moving into separate diplomatic bedrooms on the matter.

The end of an era

In 1956, President Nasser of Egypt nationalized the Suez Canal Company, responsible for the operation of the Suez Canal and owned by the French and the British. In response, the two former shareholder countries, together with Israel, prepared an invasion of Egypt to remove Nasser and to ensure that the canal remains open for transit. It started on the 29th of October, 1956 and ceasefire was announced barely over a week later on November 6 after economic and political pressure on the UK and France from the international community, most importantly including the United States. The crisis (along with the French defeat in Vietnam and the US involvement even before the French withdrawal) revealed that Britain and France can’t recover their pre-WW2 power and that their foreign policy is subordinate to US interests (which can and will keep them on a short leash if and when needed using their economic and diplomatic might). The normalization of Egypt-Israel relations (the Camp David Accords in 1978 and the following peace treaty in 1979) didn’t even bring these two to the negotiation table.

The Suez Crisis thus marked the end of the era of independent British and French great power and Middle Eastern politics (as well as it being a huge loss of prestige for the now junior countries). Since then, both countries supported US efforts in the region, most notably of course during the Gulf War, the wars in Afghanistan and Iraq (though France dissented in the latter) and the intervention in the Syrian civil war.

European-led initiatives since then also required political, intelligence and logistical support from the US to compensate for inadequate inventory and lesser capabilities. Air-refuel and reconnaissance (including AWACS) capacities were provided during the 2011 intervention in Lybia (as well as operational assistance in establishing the no-fly zone, the US involvement being ‘shaping and enabling’ as President Obama put it) and French troops were airlifted (along with air-refuel and other support) to Mali using US planes during Operation Serval (I could have just said ‘the 2012 intervention in Mali’ but ‘Serval’ just sounds so much cooler) along with the usual intel & recon support. The shortcomings during these European initiated actions serve as yet another reminder of the diminished abilities of Britain and France for independent capabilities in projecting power in the MENA region - whether political or military. For the sake of keeping it high level I won’t go into more details about the specific military forces mobilized, assistance provided or other conflicts such as the conflict with the Houthis starting 2023 (though it would also show the severe differences in commitment and capacities as the US rotated multiple carriers and deployed nuclear submarines for example while the Europeans did not despite the key importance of the the Red Sea in maritime trade and power projection) - we aren’t a milblog after all.

Public support and foreign policy

Switching gears into another but related topic (and to navigate back to what we are more about, polling), a few thoughts on the role of public support in foreign policy. This is an old debate in international relations, with arguments for either side - some believe that foreign policy decisions are too important to leave it to the (uninformed and fickle) public while others are in the position of public support being necessary to provide legitimacy for the actions in this field too as governments going against public opinion will surely be punished (in a democratic system at least). We could find examples for both sides: for the latter, consider the buildup of public support (connecting Iraq to terrorism shortly after 9/11 and the “possession of WMDs” - see Pew for a summary) in the US and it being largely successful in that initially. On the other hand, public opinion turned on the Vietnam war as early as 1968 (Gallup) yet US involvement continued until 1972-73.

Policy vs. Public Opinion

And now we are getting to why it matters now - if and how public support shaped the decision to recognize Palestine.

The UK saw a more-or-less steady support for Palestinian statehood as polled by YouGov: 45% for and 14% against in 2025; where support is only slightly up compared to 2023 from before the conflict was reignited (40% for, 8% against).

The French public seems to show more support historically as this ifop poll from 2012 shows a 66% approval for voting to recognize Palestine in the UN. This support also seems like a rough constant as more recent (2025) ifop surveys show an altogether 69% support (unconditional and conditional together). A BMFTV poll also shows a 53% approval for Macron recognizing Palestine, but that difference could also be chalked up to the president’s (lack of) popularity.

As recent polls show the US is not so different in this sense either, as a 2024 Gallup poll measured a 53% support for recognizing Palestinian statehood - though the party divide is clearly present as 73% of Democrats would approve it whereas it was only 26% among Republicans. The overall support also seems to stagnate zooming out as it reached the high fifties in the 2000s, then dropped to the forties and now it’s back at 50+% (you can examine the polled trends in the linked Gallup summary). A 2025 Ipsos poll even measured 58% support which shows a rising tendency for approval.

And how is this important? Well we can see that governments can readily ignore public opinion if they prefer.
The US didn’t recognize Palestine last time it would have been a majority approved decision under President Bush, while it’s more understandable later as the majority shifted after 2010.
The French and the Britons were more consistent in that regard (although for the latter it didn’t and haven’t reached a majority either way), public opinion didn’t change significantly over the course of the last 15-20 years.

What all this means

Which raises the question of timing, and ties these two topics together.

Considering the history of alignment on Middle-Eastern affairs between the US, France and the UK since the Suez Crisis, the materiel reality and the reliance of the latter two to exert influence even in the immediate Mediterranean area, it’s not a surprise that the Europeans ignored public support for Palestinian statehood for a long while. Now however, despite the material situation being largely unchanged (as European defense industry is still behind the American) France and the UK moved to recognize Palestine. While I wouldn’t expect the diplomatic and military situation to change between the allies drastically, this is a noteworthy rift between the previously aligned policies and could point at a considerable erosion of US soft-power over the European counterparts.

The US public largely considers the Israel-Palestine conflict a critical interest with 90% of the respondents of this Gallup poll considering it at least somewhat important. Sentiments are also shifting though: 68% sympathizing with Israel and 18% with Palestine in 2015, this was 46/33% in 2025.

Considering these, along with the US force posture review and mixed messages around it, it’s not unimaginable that this decision by the UK and France is a part of them trying to assert more influence in the region by probing a more independent diplomatic stance through ending a decades long item of direct US alignment.

If an when others will follow is another question. Germany, the Netherlands and Italy for example have yet to do so, although the latter two laid out conditions/debate for recognition already - but given these recent developments, it’s not at all out of the question that this first official break of policy is only the beginning.

Goals, assists, end-of-season statistics and highlights: they tell a story, sure, but all of these numbers are hardly equal.

Watching football regularly, we often run into conversations about player performance. We compare them using simple metrics like goals/assists; with more complicated ones like xG, xA, or even more abstract values; or simply base opinions on who passes our subjective ‘eye tests’.

For the purposes of a qualitative approach, let’s remove the eye tests for now. The other two main approaches (of simple metrics and advanced values) and their sub-metrics of player comparison run into the issue of not all goals, assists and other contributions being equal. Team performance varies, and opponent strength does so too. Comparing a goal against the 18th placed team in a league versus a goal scored against the 2nd has the same impact on face value, when generally you wouldn’t expect that the two teams put up the same quality of play (while of course upsets do happen).

This is why we built a formula to rank some of the currently established star players, taking into account various aspects, such as:

• whether a player scored the first goal of the match, the team or neither (as the assumption is that scoring the 3rd goal on a 3-0 match has somewhat less impact than opening the scoreline). We scaled the ‘values’ of scored goals up depending on whether one, or both are true and devaluing it slightly if both are false

• minutes played, rewarding impactful substitutes. Because a decisive 15-minute cameo can be more impactful than a full 90-minutes of mediocrity

• the position of the opponent team in the league where the match was played (for example, a goal against Slovan Bratislava would take their Champions’ League position, rather than their domestic league), scaling the entire performance

• the value of shooting: lots of shots (on or off target) results lots of chances created but a high number may only just be padding for the statistics sheet. Nevertheless, on-pitch threat is not ignored, but wasteful shooting is weighted less. The best forwards are, after all, the ones who are not just threatening to the opponent’s goal, but are clinical too

We wanted make sure we don’t miss any key aspects of the game or where a player is expected to contribute; when determining the data basis of the impact calculations we looked at their comprehensive on-pitch presence. This includes, in order of decreasing weighting significance (for the purposes of the impact index):

• Goals and assists, naturally, as these are the most visible metric of a player’s (and especially a forward’s) performance

• Creativity: key passes, dribbles. A player who can unlock or stretch defenses can have indirect impact on the match by setting their teammates up for goals, assists or just by drawing the attention of defenders.

• Threat: shots, times fouled, and runs offside. Pressuring the opponent can lead to mistakes from their part and might result in direct results, even if only with a lucky deflection.

• Defensive contribution: an aggregate of tackles, interceptions, blocks. As tracking back is often expected from forwards today, this also became a real metric of how they impact the structure of their teams.

With these data points, we hold all examined players to the same standard, regardless of the nuances of their individual roles. So Haaland being ‘classic 9’ will directly compete with Mbappe. who’s more of a winger or attacking midfielder. As they play in the same third of the pitch, we think it’s a fair approach to eastablish universal standards in our capacity as neither professional scouts nor fantasy football podcasters.

At this point, I’d also like to highlight that the ‘Impact Index’ is not a ‘scale of 1-to-X’ - it doesn’t have a theoretical highest possible number. Since it’s performance based it’s only limited by whatever a player manages to achieve.

For this inaugural impact index ranking, we looked into the stats of:

• Erling Haaland, forward, Manchester City

• Robert Lewandowski, forward, Barcelona

• Kylian Mbappé, midfielder, Real Madrid

• Harry Kane, forward, Bayern Munich

• Ousmane Dembelé, midfielder, PSG

and examined their numbers in domestic leagues, domestic cups and the UCL

Without further ado, here’s how their performance measures up against each other:

What do you think? Did it line up with your expectations?

Perhaps the most striking takeaway is how our methodology recalibrates and
re-contextualises the contributions of players, without being overshadowed by pure goal tallies (consider that Salah had 47 goal contributions in the Premier League whereas Dembélé racked up ‘only’ 27 in Ligue 1). After all, the game isn't just about who scores and how much, but also about who shapes the game and how. Begone surface level statistics!

This is essentially just a proof of concept; in the future, we’ll look into more in-depth analysis of performance over time through the impact index, and check & compare some other players as well. So, stay tuned and let us know if there's anyone else we should analyze!

Take me to the data

We originally intended to deploy our polling model for the 2026 Hungarian elections only (at least in terms of it being the first and only one this year), but we found a bit of extra time, we decided to try and convert it to the Dutch elections too.

The electoral system & the path to 76

To quickly recap, the Dutch House of Representatives (the Tweede Kamer) consists of 150 seats, distributed proportionally based on vote share and 76 seats are required for a majority. Due to the nature of the electoral system, it’s unlikely that any single party achieves majority (it hasn’t even happened yet in the modern era, last time it was in 1891), so coalition building is standard activity. As far as I know, no coalition plans have been announced in advance, so let’s explore a few scenarios.

Our model - based on July data - indicates a nail-biter race, while PVV remains the favourite despite losing considerable ground. GL-PvdA trails them closely, and we must also note the tight race for 3rd between CDA and VVD. PVV coming away from the elections again as the largest single party is certainly a possibility, even if trends are against them at this point.

Given how quickly the Schoof cabinet fell apart, the observable trends forecast an array of possibly uneasy scenarios for government forming and coalition building as interest and ideology-wise similar compositions will be difficult to achieve. As things stand a right-to-center-right PVV-VVD-CDA coalition, which would essentially be a repeat of the previous one, would likely fall short of gaining a majority (only just, with 71 seats).
While GL-PvdA could have an easier path to 76 when it comes to potential partners - having historical experience with both the CDA and the VVD, perhaps involving D66 - they would first have to overtake PVV. Even then, GL-PvdA would need their partners to pick up an extra 5 seats and a left-to-center-left coalition would also not reach majority (or even 60 seats for that matter).
We must also not forget the smaller players’ potential to become kingmakers as BBB and JA21 (for a conservative government) or SP and PvdD (for a GL-PvdA led formation) would gain about 9-13 seats, which could also be enough to support a minority government (one that just fall short of the majority) from the outside of it.

The shaded areas you see around the vote share lines represent our ‘confidence bands’ showing the full distribution of the polling data. This highlights exactly why we build our own model, to visualize the sometimes incredible distribution of polling and to reduce the noise of this range by weighting the individual results and normalizing the aggregates. While raw polling numbers would leave us confused about results that could fall anywhere in these areas, our model refines it to the most probable outcome (as always, based on what we can measure, and acknowledging that it is not a guaranteed result, upsets and unforeseen twists do happen).

Seat Projection

Based on this data, we also experimented with creating a seat projection using the d’Hondt method (similar to how a live election would perform allocations). Sadly granular vote share data is not readily available from all sources (for example Peil.nl doesn’t even publish them at all, or at least we didn’t find them), so we reverted back to using only our more straightforward polling model, to ensure we include all data we have access to - and we’ll shelve the projection model to use it where the context allows for such.

We believe presenting data is best done in an interactive way, so the full chart is available here for you to explore directly.

We’ll do one more update right before the election to capture as many polls published and to get as clear a picture as possible. Keep an eye out for it on our Github (which updates automatically as we add more data) and we’ll also post a note about the final pre-voting standings here.

August 2025 Update

Adding poll data from August we can see some very interesting changes compared to the last few months’ trends.

PVV widened the gap a bit from a predicted tie to a narrow lead, but the confidence bands still overlap indicating real possibilities for a reversal. VVD’s support seems to have collapsed in August and the simultaneous rise of CDA, JA21 and the gains of PVV could suggest a looming realignment as some voters might have migrated to these parties from VVD’s base. D66 and GL-PvdA didn’t produce noteworthy swings either way.

Of course in a proportional system like the Dutch, it’s not just about who’s on top, but how the rest of the pack performs - specifically those closer in ideology for coalition building - and at the moment things the right-from-center part of the spectrum has more reasons to look forward to election day. We’ll see how the last few weeks shape the field.

September 2025 Update

The race continues to be tight, as CDA might come to second place ahead of GL-PvdA; but the confidence bars overlap so theoretically despite the PVV lead it’s possible for GL-PvdA to tie the first place, or come in second ahead of CDA.

The spectacular collapse of VVD continues as D66 and Ja21 close in.

At this point a 3-way majority coalition seems impossible as both a PVV-CDA-VVD/Ja21 and (in the unlikely event of them winning) a GL-PvdA-VVD/CDA-D66/SP would fall short of a majority.

4-party coalitions are not unprecedented as it’s been the case since the 3rd Rutte cabinet in 2021 and it seems like the most likely option this time as well.

October 2025 Final Pre-Election Update

Key updates following the October additions:

• the Forum voor Democratie overtook the Socialist Party and thus replaced them on our parliamentary chart as the party with the fewest seats still called out individually

• the PVV solidified their lead against GL-PvdA as the latter continued to decline

With the VVD, CDA and (of course) the GL-PvdA stating that they won’t govern with the PVV, coalition building seems to be dead on arrival (assuming this stance holds after the election for the first two). With all the troubles a minority or a demissionary government would mean given recent experience (with neither side having enough seats to pass agenda comfortably) one thing seems to be sure: the saga of this election will not end with the results being announced.

Post election thoughts

So the Dutch general elections passed and boy was it a surprise. The polls unanimously prepared the electorate for a PVV victory and not only did they not win by the polled margin, they ended up coming in second. Also surprising were the results for the VVD over- and the GL-PvdA heavily under-performing. I’m not going to go into a tirade on whether polling is broken or not, I think the polls were fair in terms of aligning with the general expectations and rather a post hoc survey of what changed in the electorate could reveal why it was off the mark (plus I admit I’m not familiar with the Dutch news cycle so there may have been things pointing towards this last minute shift).

Also worth noting that generally the polls were still acceptable; on the first chart here we can see the polled results vs actual and the difference, the accepted margin of error highlighted in green (3% plus-minus, which translates roughly into 4 seats here). We can see that while it was not fully accurate, most were still in the range - except for the VVD and of course D66, the latter of which obviously being a huge election changing miss. The last poll published by Peil just a day before the election was a lot closer and the sharp uptick of D66 shows it on the chart as Peil has the best track record and therefore the highest weight (though the other October polls still skew the aggregation model away from the result).

All in all we also must not forget that polling works with fundamentally human data, often obfuscated by emotions so we shouldn’t expect it to be perfect. While missing the mark on who won understandably will make us raise questions as it should, we must remember that polling is not like looking into a crystal ball. We’ll go into why polling is hard soon in a dedicated series of posts so keep an eye out (and hit subscribe, just do it) for it if you are curious about how it works behind the scenes and just how easy it is to run into obstacles or make mistakes even if they try to do everything right

Take me to the data

Well it’s 7 months until the parliamentary elections in Hungary, which means that we are soon reaching the most intense campaigning period. Polling will pick up as well, and instead of seeing a handful of reports over a quarter, we’ll probably get that many in a month, which will enhance the overall accuracy of both the aggregating model and the seat allocation simulations.

A Quick Recap

Some background information of the Hungarian electoral system for international readers. The parliament consists of 199 seats, of which 106 are won in individual First-Past-The-Post (FPTP, akin to the US or UK and similar systems) electoral districts. 93 seats are allocated via the ‘national lists’ - the threshold of which is 5% vote share, then proportionally distributed using the d’Hondt method, with minority representatives having a preferential quota.
A unique Hungarian twist is the concept of ‘fractional votes’: this is used to assign extra votes to a party’s national list based on the amount of votes in each FPTP district. This is is a bit of a complicated system with two different components, but the gist of it is:

• the ‘loser compensation’: all parties who reached the parliamentary threshold are assigned extra votes to their national list based on their candidates who didn’t win in FPTP districts

• the ‘winner compensation’: these fractional votes are awarded based on the margin of the winner ahead of the 2nd placed candidate (minus one) and adding all votes cast on candidates who lost (so if they won by 10k votes in a district, they get 9.999 extra for the national list).

This is sadly not simulated in our model as this escapes our current capacities but our assumption is that this would have negligible effect anyway and the toss-up seats abstract these post hoc changes adequately for now.

The parties on the ballot include Fidesz, the governing party since 2010 (and having a 2/3 supermajority for almost the entire time); DK, started by a former prime minister and having a track record of participating since 2014; MH the offshoot of the 2018 runner-up, split from them before the 2022 election; and MKKP, a sort of ‘pirate-party’.
The last few months, and years even, were already turbulent on the Hungarian political landscape (more so than usual) with the emergence of the TISZA party in early 2024, a new(ish) center-right formation led by Péter Magyar (formerly related to the current system).
There are a bunch of other smaller parties too but they scarcely stand a chance of having measurable support so we’ll skip them from the introduction.

Vote Share Over Time

By this point in time TISZA (in red) overtook Fidesz (orange) in most polls, but as we can see here, it is far from a race over:

The confidence bands also show a huge distribution of measured support for Fidesz, with a staggering 13% possible difference between highs and lows! Interestingly though, the support for TISZA seems to be gauged much more narrowly with fluctuations being only within the margin of error. We also have to note that the sudden shift in tendencies so far might be a mirage as July has only really seen what we deemed as low quality polls - we’ll see whose lead holds as the August polls are coming in, but we shouldn’t be surprised if a reversal is solidified sooner rather than later.

These swings partly depend on the pollsters perhaps skewing for one side or another as they are often sponsored by the parties, and partly due to a perceived trust in these institutions and the government itself, but our model works hard to normalize the resulting uncertainties beyond simply averaging them out.

We can also observe a slow but steady downward trend for the other two tracked parties as they edge the parliamentary threshold, creating a nonzero chance of them not winning any seats from the proportionally allocated national list. Due to the nature of the FPTP seats, it’s almost guaranteed that only TISZA and Fidesz will win electoral districts which means there is a real possibility of having only these two parties sitting in the parliament for the next cycle. Considering that these two parties are practically diametrically opposed (not to mention the rigid government-opposition divide in Hungary), coalition building seems off the table based on conventional wisdom; and a hung parliament is also not out of the question (see below for very preliminary seat projections)

Simulating The Parliamentary Seat Allocation

We also built our first simulation model tweaking it to Hungary-specific rules and challenges. At it’s core, the simulation is a Monte Carlo test designed to simulate 10.000 elections every time it’s ran. Each of these 10,000 'elections' uses slightly different, but plausible, (defined by thresholds supported by historical trends, such as high and low turnouts being based on historical elections) combinations. By simulating these variations capturing a full spectrum of possibilities. These variables include:

• Shifting turnouts; how many undecided voters might participate and how they could do so; and random deviations within the margin of error.
  For the August data, we simulated undecideds as 28-35% of the full electorate (based on polling trends and adding a margin of error for some wiggle room) and ~70% turnout rate. This could change as new polls narrow the undecided share more.

• National party list seats based on our adjusted national vote share from above. While the nuance of the ‘fractional votes’ are not directly simulated due to our limitations, we believe that the simulation’s broader approach accounts for its effects indirectly with elements such as the shifts above and the clear uncertainties of the toss-up seats.

• FPTP districts are simulated individually, adjusted by factors such as a proportional voting support shift based on the 2022 election results and the current adjusted vote share, the final result in 2022 in each district and small unpredictable local swings. This is perhaps the most challenging and rough part of the simulation as district level polling is not readily available. Nevertheless, the local dynamics we are working with here still offer us the opportunity to predict possible outcomes. Based on the simulation results the seats are categorized ‘safe’ (a clear win with a comfortable margin), ‘likely’ (a close win which still could realistically flip), and ‘toss up’ which is everything else that are too close to call. There are also 4 new districts created first debuting in 2026, which we categorized as toss-up without further simulating them due to lack of historical data and we wouldn’t be confident including more workarounds.

This gives us a valuable estimate of how the parliamentary composition could look like, without veering into the territory of ‘reading tea leaves’ trying to predict unpredictable shifts.

We will follow the events and the data of this election closely and you can expect shorter monthly check-ins when we update our charts including this one above; but if you’d like to check out our full data and latest projection in an interactive format, visit our Github here.

2025 End of August update

Seems like the July reversal didn’t hold at least for now, which is not entirely surprising given how tight the race is between the top 2 parties (and how far the election date still is). The party Tisza regained its lead at the moment and stably hovers slightly above 40% support, while the vote share of Fidesz is doing parkour over the last 3 months (note the 8% difference between July and August).

A lead in vote share however doesn’t automatically mean a commanding lead in parliamentary seats, especially in the Hungarian election system and we can see on the new projection just how slim this margin would be (the simulation is also less impressionable than the aggregation model).

2025 September Update

The flip-flopping of the polls continued, though in September a new pollster, McLaughlin & Associates, entered the landscape without previous presence in Hungary. While the individual poll doesn’t matter much, it is putting our model to work as it has to place it without historical accuracy data. Our opinion is that we’d rather risk overweighting a new pollster than dismissing them just because they haven’t polled this electorate yet. As such we’ll see how the charts change in October when it’s more likely that only established firms publish polls.

2025 October Update

October Hungary 2026 election tracker update time! Currently our only active tracker by the way.

At this point I won’t even call out how much the polled numbers top 2 parties can change month-by-month - it seems we are stuck with this double helix of a chart.

This is partly because of the stark difference between government and non-government aligned polling firms. We don’t specify which is which because our model doesn’t factor in bias for weighting, only rolling accuracy. Now of course in past cycles government-aligned pollsters were more accurate generally (not going to unpack that) so at some point it’s entirely possible that they will miss the mark and thus the aggregation will also be off in the end. Just something to keep in mind.

While we don’t care about alignment generally we also made a choice in excluding a firm, called XXI. Század (21st Century) without previous track record which is way too openly and closely aligned with the governing party. We felt that we couldn’t yet give it the benefit of the doubt and as such treat them the same as any other newcomer, essentially giving them an average rating above established but historically lower quality firms. This is because of the past of the head of the firm - we stand by our choice and may reconsider next time.

2025 October Update

In our last update in 2025, let’s take a short look at the November polls. Probably at this point to no-one’s surprise TISZA took the lead again in the overall adjusted average. As we can see, the race is not any less close; in fact after these position switches the two biggest parties keep staying closer to each other. The other two properly measurable parties are also slipping below the parliamentary threshold, though that doesn’t mean they can’t get there.

There were also new polling firms publishing their snapshots over the course of the last few months, we’ll see how many will continue to do so.

Have you ever looked at US poll and asked yourself how invested you should be? Have you ever obsessed over election models in Germany? Were you ever puzzled why working culture seems to move at a snail's pace despite the circumstances? Debated if Mo Salah is really the most clinical forward in the top 5 European leagues? Have you ever spent way too much time thinking about these on a Sunday morning (maybe even doomscrolling the news)? Well, if you have, you are definitely not alone. And if you haven't, we've dived in so you don't have to. Either way, if you were ever curious about the data behind the headlines, you are exactly who we build Margin of Error for.

The Why and the Who

For a long while, we both spent a lot of time trying to prove the other wrong. Depending on your point of view, (un)fortunately we both have a knack for looking up specific and perhaps obscure data to support our points. This started and has always been a hobby, then, through the sheer amount of weekend hours we invested, gradually became a certain lifestyle.

Then the idea of publishing our thoughts started in 2024, where we - guilty as charged - became way too immersed in the US election cycle. It was really a simple train of thought, trying to make sense of what we are seeing at first, then trying to compare the data with past cycles to ease our anxiety (didn't work), then understanding what the methodologies are and where they went wrong. From this point, one thing led to another and we started projecting our ideas and findings elsewhere too, and since we were already in the mindset of statistics, predictions and the human component therein, we looked at our other interests to find where we could dig deeper. We figured that if we liked thinking about all this, there's a chance others would too. And now we hope to share this appreciation for what the numbers tell us about various aspects of the world, how we get to them and what we might learn.

Besides our downtime, we also have extensive experience with reading data and creating statistical frameworks. We both wrote our theses on topics based on the intersection of psychology and statistics, and our day-to-day work also covers the generation of professional assumptions based on what data we have available.

And so here we arrived to the point of this post, the grand opening of our page. The name Margin of Error has a dual purpose. On the one hand it refers to, of course, the actual inherent statistical margin of error that's part of any data collection that involves humans providing the information. On the other hand it reflects our own core philosophy, that few things human are exact or perfect and there's always a certain margin of error - whether it comes to decision-making, life stuff, or projects such as this. We have to try, and we are committed to doing our best, to read information and come to conclusions, to make decisions and bet on the things we believe in, but we also have to embrace the possibility of missing a swing, and being ready to correct course when we recognize a need. And that's true for adjusting a polling model, changing culture at a workplace based on engagement data, or growing and becoming better in writing introductory articles on newly created blogs.

The What

So what can you expect to read about here? Looking back, I think we narrowed it down to a little bit of everything. More specifically though, as a next step we will be diving deep into how polling works and explaining our very own polling accuracy model. In the future we will publish other, perhaps more grounded topics as well, ranging from sentiment analyses, through workplace culture and trends, to sports (well, really mostly football).

Part of who we are is knowing who we aren't, and part of knowing what to expect from us, is knowing what not to. We aren't writing peer-reviewed academic theses for scientific journals. We offer our takes on the data, our own perspectives based our reading of data. Every interpretation comes with its own margin of error, including ours. Our purpose is to balance informed curiosity with professional talk, in order to make the fundamental topics of Margin of Error more accessible and to bring interesting data to the table: not to define truths, but to create conversation.

This project is more than just a hobby for us, we dedicated and we continue to invest tremendous amounts of free time and passion to making it happen, but this is not our full time job (yet). We carefully consider and digest each and every topic, and we want to challenge ourselves both in terms of what we talk about and how we get to that point. Quality over quantity is our modus operandi. Because of these, we will not post every other day, but we will not disappear in the meanwhile - we will be active on our social media channels with bits and pieces of engagement and we are always here in case you'd like to reach out to participate, or with a question.

The Why Now

We feel comfortable with our designs, our plans for the future and we had ample time to test our ideas, capacities and capabilities. As our home, Hungary is preparing for an election in 9 months it gives us a locally very much relevant opportunity to test our model and reach out to a core audience. We are also well ahead of the US midterms in 2026, the French presidential election in 2027 and the next UK (if we can figure out a way to meaningfully work with their system) and German elections well beyond the horizon so we can prepare for upgrading the scope of the model in due time.

We hope you'll join us for the ride, enjoy the topics we muse about and share your thoughts, ideas or feedback - we will be here and happy to hear from you.

To (a well informed) adventure!

Subscribe now

Share Hunor’s Substack