Research · The Government's colours

The Government's colours on screen: consistent and easy to recognize

If every state website uses the same visual identity, how much does it matter that the Government's blue looks exactly the same everywhere?

On a new phone, on an old laptop or on a printed sheet, the same colour never looks perfectly identical. The useful question is a different one: what needs to be fixed, and where, so that people recognise it?

I started from the two base colours in the Government's visual identity guide, #004990 and #58595B, and checked how they behave on screens, in different conditions and in print. The situation described is as of 22 October 2026. Here you will find:

• what the guide says and how it relates to the colours in the law;
• how much the two colours change on screen and in print;
• when the change becomes a problem for people;
• what role colour plays in recognising a state website;
• how other states define their colours;
• what I propose for a design system for the state.

The calculations are simulations made with the standard colour formulas, not measurements on devices. This is a personal analysis, meant as a starting point for a discussion with the General Secretariat of the Government (SGG). The technical details are in the blocks that open; the text can be read without them.

0. Summary in eight points

1. The same colour does not look identical on every screen, and it does not need to. On ordinary screens, the two colours change only a little. They change a lot only in sunlight, on e-paper and in accessibility modes.
2. The Government's current blue is a good choice for the screen. It reads clearly as blue and has good contrast on a light background.
3. The bigger problem is in the documents. The same “PANTONE 280C” has three different screen values: in the SGG guide, in the flag law and at Pantone.
4. The blue and the grey should not be used together to convey information. They have almost the same lightness, and in greyscale they become almost the same.
5. The dark theme needs lighter variants. On a dark background, both colours have too little contrast.
6. In print, you have to say which paper the colour is calculated for. The guide's CMYK value matches the screen on coated paper, but not on uncoated paper.
7. Colour helps recognition, but it cannot prove that a website is official. That is the job of the .gov.ro domain, HTTPS and the official banner.
8. The recommendation: a single reference for each colour, with documented variants for each medium.

1. The starting question

One of the values of a design system for the state is that people recognise the institution and know that a website belongs to the Government. From this, it seems to follow that the two base colours must look the same on any screen, old or new.

The research shows that the first part is right: colour helps people recognise a well-known brand. The second part cannot be met, and it is not needed either. The right wording is “colours that are consistent and easy to recognize”:

• the same reference value everywhere;
• documented variants for each medium;
• a measurable tolerance where the state controls production.

2. What the SGG guide says

The April 2022 guide gives the colours on page 6 (“Accepted colours”). The main colours are “accepted for colouring the logo background”, and the secondary ones are “for adjacent graphic elements”.

Three things follow from this:

• The colours are designed for print first. The RGB values are conversions of the print colours, not choices made separately for the screen.
• #58595B is “80% black”, a shade of print black. On screen it is very close to Pantone Cool Gray 11 C.
• For websites, the guide uses other shades of blue (pages 21–26): the header bar is #024476 for central government, #0764AB for town halls and county councils and #027AD4 for prefectures, and the framing lines are #005EA4.

2.1 The same “blue”, three different colours

The blue “PANTONE 280C” appears with different values in different documents. Pantone warns that its screen values are “computer simulations” of the print colours.

The difference is measured with ΔE₀₀ (CIEDE2000), against #004990. Below 2, a difference is visible only with the samples side by side; above 5, anyone can see it.

The blue #004990 is lighter than the Pantone value and reads better as blue on screen; #012169 looks almost black. It is a good choice for the screen, but it would be clearer if the guide called it “the Government's screen blue”, not the equivalent of 280C. For the flag, the values in the law remain, and they are mandatory.

2.2 Pantone, CMYK and paper

The guide mixes three ways of describing colour, Pantone, CMYK and RGB, without saying when each one is used or which paper it is calculated for. I simulated the values in the guide with the print profiles that European printers use:

What this shows:

• On coated paper, the guide's CMYK and the screen blue are practically the same colour. The guide therefore already has a consistent pair for screen and print.
• On uncoated paper, such as office paper or letter paper, no CMYK mix reaches this blue. It comes out lighter and duller. Here you need a value of its own, accepted as such, or Pantone ink.
• The label “PANTONE 280C” is the part that does not fit. The 280C ink is darker than #004990 and seems to be outside the CMYK gamut even on coated paper: the closest mix stays at a difference of about 9. The value 100/72/0/18 is Pantone's 2001 conversion; Pantone has changed it several times since then.
• Other states also give CMYK next to Pantone. Germany gives separate values for coated paper, uncoated paper and newsprint. None of the guides checked names the print profile.

3. What a colour code means on screen

• A hex code means sRGB. The sRGB standard (IEC 61966-2-1) describes a reference screen, with a D65 white and 80 cd/m², viewed in a room lit at 64 lux. On that screen, #004990 is a single, measurable colour.
• The browser has to convert the colour for the real screen. The CSS Color 4 specification requires this. The conversion is called colour management.
• Who does it, in October 2026:
  • Mac, iPhone and iPad: the whole system.
  • Windows 11: only with the “Automatically manage color for apps” option switched on, on compatible graphics cards and screens. Without it, colours reach the screen unconverted.
  • Android: from version 8.1. Even so, many phones start in the “vivid” colour mode, not the accurate one; on Pixel phones, “Adaptive” is the default mode.
  • Firefox: on Windows and Linux, it does not convert CSS colours by default.
• Display P3 and OKLCH colours allow more saturated shades on new screens. They do not, however, make a colour look the same everywhere.

4. How the two colours look on different screens

I simulated idealised screens, with the standard formulas (sRGB, CIELAB, CIEDE2000). The samples below show what the eye would see in each situation. They are displayed on your screen, so they already look slightly different than on someone else's.

On ordinary screens, the difference from the reference stays small, below 4. It grows on old or unadjusted screens, to between 6 and 10, and becomes large only in sunlight and on black-and-white e-paper, above 18. For comparison, in print the ISO 12647-2 standard accepts a difference of 3.5 for the process colours, and big brands usually ask for 2.5–3.

5. Where it really matters

So far I have checked how much the colour changes. For a design system, the important question is a different one: when does the change become a problem for people?

5.1 The blue and the grey have almost the same lightness

The contrast between #004990 and #58595B is only 1.27 : 1. They differ only in hue.

On black-and-white e-paper, in a phone's greyscale mode, on a photocopy or for some colour-blind people, the difference almost disappears. The two colours should not be used next to each other as the only way to convey information, for example blue text on grey or a blue/grey pair of states.

5.2 The dark theme

On a dark background, both colours are below the WCAG thresholds: 4.5 : 1 for text and 3 : 1 for components. Other systems solve this with a variant for the dark theme:

• GOV.UK has, in its app, a separate blue for the dark theme (#37A3FF);
• Material Design uses a different tone of the same colour in each theme.

The example #578ED3 keeps the hue of #004990 and is only lighter. The final variant should be chosen together with the SGG.

5.3 Colour blindness

Red-green colour blindness affects about 8% of men and 0.5% of women. For them, dark blue stays dark blue. For people with tritanopia, which is very rare, the blue becomes a dull blue-green. The grey does not change.

For some people, the difference in perception is bigger than the one produced by an ordinary screen. They too must be able to recognise an official website, through the logo, the text and the banner. WCAG requires the same thing in success criterion 1.4.1: colour cannot be the only way to convey information.

5.4 Accessibility modes

• High contrast in Windows replaces the website's colours with the colours chosen by the user.
• Colour inversion (iOS, Android) turns the blue orange.
• Colour filters and greyscale mode change all colours.

These are modes used on purpose by people with low vision. A design system for the state must let them work, not get around them.

6. Colour helps people recognise a site, but does not prove it is genuine

Recognition and authenticity are different things. Colour helps people recognise a public service, but it cannot prove that a website belongs to the state.

• Colour memory works in categories. Remembered shades are pushed towards the centre of the category and towards more saturation (Bartleson 1960; Bae et al. 2015).
• The link between colour and brand appears only for very well-known brands (Kimura et al. 2013).
• Many people do not even recognise their own shade. In a study at Clemson University, only 45% of participants picked the university's orange out of four Pantone orange shades.
• The statistic “colour increases brand recognition by 80%” has no verifiable source. It comes from a sales brochure and compares colour with black and white, not shades with each other.
• Colour alone does not tell states apart. The SGG blue is at ΔE₀₀ 4 from the Dutch government's blue (according to the values published by the community that uses it), at 7.5 from that of the US design system and at 8.5 from the European Union blue.
• A fake website easily copies the colours and the logo. On 20 March 2025, the National Cyber Security Directorate (DNSC) warned about a website imitating the portal of ANAF, the national tax agency. The advice was to check the sender and the domain, not the look.
• What other states use as a sign of trust:
  • USA: a banner, “An official website of the United States government”, which explains the .gov domain and HTTPS;
  • Singapore: a mandatory banner, which cannot be restyled;
  • Germany: “Digitale Dachmarke”, with a bar showing the flag, the federal eagle, the gov.de domain and the KERN design system, explicitly designed against fraud;
  • Australia: the Scamwatch guidance tells citizens that the logos on a website “can just be copies”.
• I found no study comparing how people recognise official websites by colour, by domain or by banner.

The proposal on this site already includes the official banner, alongside the SGG header.

7. How other states define their colours

None of the guides found gives Display P3 values, sets a tolerance for screens or promises identical rendering. Apple's interface guidelines say plainly that colours look different in sunlight and in dim light.

8. The recommendation: colours that are consistent and easy to recognize

Up to this point, the research describes what exists. From here on, it is my proposal: how these conclusions could become a design system for the state.

Each point below has a visual example. The examples start from what the design systems cited in the project already do: GOV.UK (United Kingdom), USWDS (USA), DSFR (France), Designsystemet (Norway) and Material Design. The new values are proposals to discuss with the SGG, not decisions.

8.1 A single reference for each colour

Each colour has a record: the sRGB code, plus the OKLCH and CIELAB values, so that it can be checked with an instrument, and the contrast on the backgrounds used. This is how GOV.UK and USWDS present their colours.

8.2 Colours named by role, not by code

Pages use the name of the role (“brand”, “link”, “secondary text”), and the role gets a value for each theme. If a code has to be adjusted, it changes in a single place. GOV.UK asks teams to use the “brand” role, not the code. DSFR gives its blue, “Bleu France”, the same name in both themes: #000091 in the light theme and #8585F6 in the dark one.

8.3 Documented variants for each medium

The same role has a written value for each medium. Where the value cannot be set from the office, the guide says how it is set.

8.4 Print: CMYK with a named profile, Pantone only where it helps

For print, the guide could give for each colour:

1. The reference: a measured CIELAB value, under the standard D50 light, with a tolerance: at most 2 on the colour proof, at most 3 in production. CIELAB is a public standard (ISO/CIE 11664) and needs no licence.
2. CMYK values for named profiles:
   • coated paper: FOGRA51 (PSO Coated v3), with FOGRA39 accepted, because many printers still use it;
   • uncoated paper: FOGRA52 (PSO Uncoated v3);
   • optionally, a value for newsprint, as Germany does.
3. Pantone optional, not mandatory, only where the printing process justifies it: long offset runs in one or two colours, or uncoated paper, where CMYK falls short.
4. The grey stays 80% black, with a CIELAB value and a minimum size for text. Printed as dots, the grey can break up in small letters, especially on uncoated paper.
5. Print files are handed over with the profile embedded, so that the printer knows which paper the colours are calculated for.

Signs and other physical materials (films, paints, textiles) use other colour codes. For them, the guide would give a RAL code or the manufacturer's code, an approved physical sample and a measurable tolerance.

8.5 Tolerance only where the state controls production

In print, the state orders the work and accepts delivery of it, so it can require a tolerance against the physical sample. The example uses a difference of at most 3, close to what print accepts (ISO 12647-2) and to what big brands ask for. For citizens' screens, a tolerance makes no sense.

8.6 The blue and the grey do not convey information on their own

The blue and the grey have almost the same lightness, so a state marked only by colour is lost in greyscale. The state is also written in words, possibly with a symbol. WCAG requires this in success criterion 1.4.1.

8.7 Trust rests on what can be checked

Colour helps you recognise the website, but the banner tells you how to check it: the .gov.ro domain and HTTPS. The example reuses the banner text from the demonstration, following the model of USWDS and Singapore. For citizens, the simple message remains: “check the domain, not the colours”.

8.8 A scale for the Government's blue, in version 3 of the guide

A version 3 of the SGG guide could put all the shades of blue on a single scale, with the same hue and steps of lightness. Scales of this kind exist in USWDS (steps from 5 to 90), in Designsystemet and in Material Design. In USWDS, the difference between steps tells you the contrast directly: 50 steps means at least 4.5 : 1.

Step 60 is #004990 itself, unchanged, as in Designsystemet, where the brand colour is kept exactly in the scale. Step 40 is the variant for the dark theme.

The scale serves central government and the prefectures. Local authorities can make their own scale, using the same method.

8.9 Three levels, three ways of using the Government's identity

The Constitution and the Administrative Code divide the administration into three levels, and the visual identity can follow the same division. The SGG guide already does this in part: it has separate header colours for central government, for town halls and county councils and for prefectures.

What follows from this:

• Central government uses the Government's identity as it is. The .gov.ro domain is mandatory for it, as it is for the prefectures (Government Decision (HG) no. 1480/2008).
• The prefecture represents the Government in the county, not the local authority. It is not subordinate to the town halls and the county council, nor are they to it, and one of its duties is to check that their acts are lawful. The county appears in the name because it is the prefecture's territory. Prefecture websites already look like this: the Government's logo, GOV.RO, the name of the county and its coat of arms. In France, prefectures use the state's identity, with the name of the territory, while regional and departmental councils remain outside it.
• Town halls and county councils keep their own identity. No act requires them to use the Government's identity; their legal sign is the coat of arms, approved by Government Decision (Law no. 102/1992, HG no. 25/2003). The DigiLocal programme requires its model only for websites funded through the programme. The system gives them the same components, and they apply their colour through the roles from 8.2. This is how NL Design System works in the Netherlands. In Germany, the common digital brand (Digitale Dachmarke) does not touch the identity of the federal states and the municipalities.
• For all three, these stay the same: the page structure, the components, the official banner and the contrast rules. They are what show that the website is part of the same system.

9. What I could not confirm

• The current Pantone values. Pantone now shows them only in its paid Pantone Connect app. #012169 is the value from the pantone.com archive from 2017 and 2021.
• The default colour management mode in Windows 11 24H2. There are only unofficial sources.
• How Chrome handles CSS colours without a profile. I found no official source.
• The Dutch government's official values. They are accessible only with an account.
• The texts of the EN 12966 and ISO 12647-2 standards. They are paid for; the values come from sources that cite them.
• The CIELAB value of the PANTONE 280C ink. Pantone gives it only in its paid app. The comparisons with 280C use its screen value (#012169); the exact result can be found by measuring a printed sample.
• The FOGRA51 and FOGRA52 profiles. The print simulations use FOGRA39 and FOGRA29, which were available. I did not find out which profile printers in Romania usually ask for.
• The colour blindness simulations. They differ depending on the model used: between 2 and 10 for red-green colour blindness, between 14 and 20 for tritanopia.

10. What this means

We cannot make the Government's blue look identical on every screen and in every condition. None of the other states' guides I checked promises that.

We can, however, define a single reference colour, clear variants for the media that need them and rules that preserve contrast and accessibility. Instead of an identical reproduction, which is impossible, we can ask for consistency where the state controls the result: in documents, in code and in print.

Colour helps people recognise a public service, but it should not be the only sign. The visual identity, the .gov.ro domain, the official banner and the common structure of services work together.

That is why the goal should not be “the same colour everywhere”, but colours that are consistent and easy to recognize, on screen and in print.

Sources

• The visual identity guide of the Government of Romania, version 2, April 2022, pages 6 and 21–26: https://identitate.gov.ro/ro/ghidul-de-identitate/
• Law no. 75/1994, Annex 1, as amended by Law no. 176/2023 (Official Gazette (Monitorul Oficial) no. 550 of 20.06.2023)
• ICC, the sRGB registry (IEC 61966-2-1): https://registry.color.org/rgb-registry/srgb
• W3C, CSS Color Module Level 4: https://www.w3.org/TR/css-color-4/
• Microsoft, colour profiles and Advanced Color: https://learn.microsoft.com/en-us/windows/win32/wcs/advanced-color-icc-profiles
• Microsoft, Auto Color Management: https://learn.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range
• Mozilla, the gfx.color_management.mode setting: https://hg.mozilla.org/mozilla-central/raw-file/tip/modules/libpref/init/StaticPrefList.yaml
• Mozilla, oversaturated colours on wide-gamut monitors: https://bugzilla.mozilla.org/show_bug.cgi?id=1488619
• Android, colour management: https://source.android.com/docs/core/display/color-mgmt
• Google, screen colours on Pixel phones: https://support.google.com/pixelphone/answer/6111557
• MDN, forced-colors: https://developer.mozilla.org/en-US/docs/Web/CSS/@media/forced-colors
• MDN, color-gamut: https://developer.mozilla.org/en-US/docs/Web/CSS/@media/color-gamut
• RTINGS, panel types: https://www.rtings.com/monitor/learn/ips-vs-va-vs-tn
• DisplayMate, colour gamut in ambient light (Galaxy S10): https://www.displaymate.com/Gamut_47S.html
• Dain, Kwan and Wong 2016, colour consistency on phones: https://pubmed.ncbi.nlm.nih.gov/26974937/
• Notebookcheck, Dell G15 5510: https://www.notebookcheck.net/Dell-G15-5510-laptop-review-Budget-gaming-laptop-with-the-RTX-3050.629381.0.html
• E Ink, Kaleido 3: https://www.eink.com/brand/detail/Kaleido3
• Watchfire, calibrating LED displays: https://www.watchfire.com/led-basics/digital-billboards-with-true-color-calibration
• Pantone, 280 C (2021 archive): https://web.archive.org/web/20210920054232/https://www.pantone.com/color-finder/280-C
• Apple, the interface guidelines, colour: https://developer.apple.com/design/human-interface-guidelines/color
• Bartleson 1960, memory colours: https://doi.org/10.1364/JOSA.50.000073
• Bae, Olkkonen, Allred and Flombaum 2015, why some colours seem easier to remember: https://doi.org/10.1037/xge0000076
• Kimura et al. 2013, colour memory for well-known logos: https://pmc.ncbi.nlm.nih.gov/articles/PMC3707833/
• The “Clemson Orange” study, 2019: https://www.flexography.org/industry-news/2019-gary-hilliard-fqc-scholarship-recipients-research-clemson-orange-brand-color-consistency/2
• The origin of the 80% statistic: https://www.insights4print.ceo/2019/02/color-increases-brand-recognition-by-80-the-real-contents-of-the-loyola-study-revealed/
• W3C, WCAG 2.2, success criterion 1.4.1: https://www.w3.org/WAI/WCAG22/Understanding/use-of-color.html
• GOV.UK, the brand blue: https://brand.design-system.service.gov.uk/colour/govuk-blue
• GOV.UK, the app colours: https://brand.design-system.service.gov.uk/colour/app
• USWDS, the official banner: https://designsystem.digital.gov/components/banner/
• Singapore, the official banner: https://designsystem.tech.gov.sg/components/masthead
• The German federal government, Digitale Dachmarke: https://www.bundesregierung.de/breg-de/aktuelles/pilotphase-digitale-dachmarke-breg-2325616
• Canada, colour values in the Federal Identity Program: https://www.canada.ca/en/treasury-board-secretariat/services/government-communications/federal-identity-program/technical-specifications/official-symbols/colour-values.html
• Regulation (EU) 2014/821, Annex II: https://www.legislation.gov.uk/eur/2014/821/annex/II
• Pantone, Pantone Connect prices: https://support.pantone.com/en/pantone-connect-pricing-by-currency
• Pantone, how often printed swatches should be replaced: https://www.pantone.com/eu/en-eu/articles/technical/how-often-should-you-update-pantone-graphics-guide
• The removal of the Pantone libraries from Adobe programs (2022): https://pxlnv.com/linklog/pantone-libraries-adobe/
• X-Rite, how many Pantone colours can be reached with CMYK: https://www.xrite.com/blog/pantone-extended-gamut-guide-helps-printers
• The CMYK conversions of PANTONE 280 over the years: https://printplanet.com/threads/pms-280-conversion-to-cmyk.19416
• ICC, the FOGRA39 profile: https://www.color.org/FOGRA39.xalter
• ICC, the PSO Coated v3 profile (FOGRA51): https://registry.color.org/profile-registry/PSOcoated_v3
• The German federal government, colours for screen and print: https://styleguide.bundesregierung.gov.de/sg-de/basiselemente/farben
• DNSC, the warning about the website imitating the ANAF portal (20.03.2025): https://stirileprotv.ro/stiri/actualitate/dnsc-avertizeaza-atacuri-de-tip-phishing-care-vizeaza-utilizatorii-anaf-prin-e-mailuri-false.html

← All research documents