Limitations: what the index does NOT say

• The reference price is not your shop's price

  material

  The index uses one reference price per year from heterogeneous sources, adjustments and estimates. It is not a measured national median for every year. The price you pay depends on the area, shop, product and purchasing channel.

  What we do: Every year carries an uncertainty range, and a separate city map shows visitor-reported prices.

• The wrap is not the same product across the series

  material

  Historical meat weights and changes in composition are model assumptions, not a series of market measurements. The assumed peak and decline in portion size need verification. The present-day weight defines the reference product. The era price refers to the wrap sold at the time; the classic price and weight-adjusted index depend on these assumptions.

  What we do: The headline index is expressed in "classic gyros", rescaled to a fixed 88 g baseline. The era price stays visible next to it everywhere.

• A price ratio does not measure grams of meat

  notable

  The wrap-to-portion price ratio compares prices within a menu. It does not measure how much meat goes into a wrap: sides, size and pricing policy also affect it. The saved pilot contains 4,581 rows, but 1,528 repeat exactly the same fields. Grouping them leaves 3,053 distinct catalogue descriptions. Weight appears in 66 of 1,068 portion descriptions (6.2%) and 3 of 1,985 wrap descriptions. These are not verified product identities or a representative market sample.

  What we do: Among portion declarations, 300 g appears 52 times, with a 200-500 g range. Wrap descriptions include XL and different meats, without clarifying whether weight refers to raw or cooked meat. They are NOT enough to calibrate historical portions. The conclusion that more expensive cities do not sell larger portions has been withdrawn: selected weight declarations and city median prices do not measure the same products. Geographic coefficients remain price coefficients.

• 1976-1989: regulated prices, not a free market

  material

  Prices in this period come from official price orders (maximum allowed prices). They are not free-market prices and often sit below what was actually charged. Regulation does not even stop in 1989: Market Order 14/1989 replaces the fixed ceiling with a formula tied to the cost of meat, which remained in force past 1992.

  What we do: We apply a +11.0% uplift, the mean of TWO measurements: the 1976 Makedonia report, which prints a 7.50 GRD ceiling next to an 8 GRD market price, and the 1989 report, which prints 65 against 70-80. The band stays wide at +/-35%, and those years are marked with a separate shaded zone on the chart.

• Minimum wage is not average wage

  notable

  The index tracks the statutory minimum wage of a single employee with no seniority increments. It does not represent average pay, freelancers, part-time work, undeclared work or benefits.

  What we do: The wage calculator lets you enter your own net pay and see your personal index. The mean and median full-time wage are now measured separately, for the years their sources cover. See the mean and median wage in the same unit.

• The series combines price inputs and calculations

  notable

  Price inputs graded A or B exist for 20 of the 51 years (1977-1978, 1980, 1983-1984, 1989, 1998, 2000-2002, 2011, 2015, 2019-2026). A grade does not establish a national measurement: inputs include local sources, regulated prices and estimates. Other years use different interpolation or derivation rules; the 2012-2019 period also applies price pass-through and rounding rules. Each year identifies the rule it actually uses.

  What we do: Every year carries an A-D confidence grade and the table shows where each number came from.

• The catering index does not always move like souvlaki

  notable

  CPI baskets are not single-product prices. The May 1979 statement that prices remain unchanged did not establish a full-year freeze: June/July sources show a later increase. The zero historical interpolation weight has been corrected. At the euro changeover, the Rizospastis survey reported increases up to 46%, a reported maximum rather than a national median.

  What we do: Interpolation remains labelled as a calculation. Local price ceilings and individual purchases are not presented as national samples.

• Classic-equivalent rescaling rests on an assumption

  notable

  We assume meat accounts for 70% of a wrap's cost. What that assumption costs is measurable: the 1976 index is 314 on the current assumption, becomes 327 if the real share is 60% and 296 if it is 80%. That is roughly ±5% at the oldest end of the series, and less the closer you get to today.

  What we do: The calculations page includes a sensitivity analysis: move the assumption and watch the result change.

• We do not measure quality, taste or sourcing

  minor

  A 4.30 EUR wrap with fresh pork and a 4.30 EUR wrap with reconstituted frozen meat count identically in the index.

  What we do: We only record measurable attributes (grams, composition), not ingredient quality.

• Counter price only, no delivery

  minor

  The reference basket is the in-store price. Delivery prices (and platform fees) are markedly higher today and are excluded from the series.

  What we do: Observations record the sales channel and are explicitly normalised to counter prices.

• Visitor reports are not a statistical sample

  minor

  Prices submitted by visitors are a self-selected sample, sometimes recalled decades after the fact.

  What we do: They appear as a separate comparison layer, with median and trimmed mean. They do NOT enter the index: the prices page shows a simulation of what would happen if they were blended in at weight w, but the published series never uses them.

The rule: a figure that cannot survive scrutiny does not deserve to be published. That is why every year carries a range, a confidence grade and a source, and why the correction form below exists.