A research team from the University of Tsukuba analyzed the regional pricing of PC games sold on Steam and classified publishers' pricing settings into four patterns. Even though Valve provides recommended prices for each market, companies do not uniformly adopt them. Comparing 268 companies and 7,375 titles across 8 markets in 2023, the researchers found a range—from companies that closely followed the recommended prices to companies with large differences depending on country or title.
However, this is an observational study that captures the state of pricing at a single point in time. It is neither a study that proves corporate affiliation causes pricing differences, nor one that tested which pricing strategy boosts sales. It's important to read separately what the four patterns do explain and what they do not.
How were 7,375 titles compared?
Tomohiro Murakami and Mitsuo Yoshida used price data collected from Steam in August-September 2023. The target markets were 8 in total: in the Americas, the United States and Brazil; in Europe, France and the United Kingdom. In Asia, they covered China and India, and also included Japan and South Korea. To avoid the influence of temporary sales, they focused only on regular prices before discounts. Furthermore, to stably capture trends by publisher, they retained only companies handling 10 or more titles. As a result, the analysis covered 268 companies and 7,375 titles.
By excluding publishers with fewer than 10 titles, the pricing behavior of operators selling only a small number of works was not included in the analysis. The figure of 7,375 refers to the number of game titles, not the number of buyers or transactions. When extending the research results to all sellers on Steam or to consumer behavior in general, this selection criterion serves as a boundary.
The benchmark for comparison was the regional recommended prices that Valve provided at the time in 2023. For markets other than the United States, the research team calculated a "price ratio" by dividing the actual selling price by the recommended price. A ratio of 100% means the two match, while 130% means that when the recommended price is set to 100, the actual selling price was 130. This is not a rate of increase tracking price changes before and after a price hike.
Next, for each company, they used features such as the average, median, maximum, and minimum values by market, and applied hierarchical clustering. This is a statistical method that groups objects with similar characteristics in sequence; it is not a method that measures responses by varying experimental conditions. Nor is it a simulation result from a theoretical model. It is an analysis that searched for groups of companies with similar settings based on the distribution of published prices.
The unit of analysis is the publisher, aggregating characteristics of their pricing settings. It is not an experiment that assigned consumers as subjects, nor a survey that asked about purchase intentions or willingness to pay. While the placement of prices can be observed, whether consumers who saw those prices actually bought the products cannot be determined from this design.
How should the four patterns be read?
The most common pattern found through hierarchical clustering was the "adjusted type," with 116 companies. These companies made small adjustments from the recommended price in each market, with relatively small variation between markets. Next was the "emerging market premium type" with 69 companies, which tended to exceed the recommended price in Brazil and India, followed by the "benchmark-following type" with 54 companies, which stayed close to the recommended price. The "high-variance portfolio type," with large differences by country or title, had 29 companies. Adding up the four groups gives the total of 268 companies analyzed.
| Type | Number of Companies | Characteristics from Published Materials |
|---|---|---|
| Benchmark-Following Type | 54 companies | Close to recommended price in each market, with small variation |
| Adjusted Type | 116 companies | Small adjustments from recommended price, relatively small variation between markets |
| Emerging Market Premium Type | 69 companies | Tends to exceed recommended price especially in Brazil and India |
| High-Variance Portfolio Type | 29 companies | Large variation between markets and titles |
This table represents a typology comparing how companies used a common benchmark. Since the research did not analyze sales volume, revenue, or profit for each group, it cannot be determined which of the four types is superior. There is also no guarantee that the four groups obtained through clustering would appear in the same form at a different time or on another platform.
What Brazil's 128% and India's 130% represent
In terms of average price ratios by market, China was approximately 111%, Brazil approximately 128%, and India approximately 130%. For Brazil, this is a comparison meaning that when the recommended price is set to 100, the average actual selling price was about 128. India, on the same scale, was about 130. This difference cannot be read as a rate of price increase, since it does not measure price changes over time.
In France and the United Kingdom, the average price tended to be relatively close to the recommended price, and the variation in price ratios between titles also tended to be small. Japan and South Korea showed the same tendency. On the other hand, the published press materials do not include the average ratio, absolute prices, or confidence intervals for each of the four markets. The name of the statistical test used for between-group differences, or the effect size, also cannot be confirmed from the published materials. What can be read without supplementing detailed values is the difference that in the four mature markets, prices were close to the recommended price, while in China, variation was somewhat larger. Brazil and India showed large variation in both price ratio and dispersion.
The reasons for regional differences were also not measured directly. In addition to purchasing power and exchange rates, regulations and cross-border purchase risks form the background for considering international pricing, but the factors that produced the ratios in Brazil and India were not identified. Without surveys of the personnel who set the prices, or experiments that manipulated conditions, it is not possible to determine corporate intent from the observed differences alone.
The bias among 19 affiliated companies and the high-variance type
The research team also examined which of the four groups included the 19 companies confirmed to have affiliations or relationships with globally large-scale game companies. Of these, 11 companies were included in the high-variance portfolio type, accounting for 37.9% of the 29 companies in that group. Among all companies analyzed, affiliated/related companies numbered 19, or 7.1% of the 268 companies.
In the other groups, the benchmark-following type had 1 company (1.9% of 54), the adjusted type had 4 companies (3.4% of 116), and the emerging market premium type had 3 companies (4.3% of 69). This difference, backed by actual numbers, supports the observation that affiliated/related companies were more frequently observed in the high-variance type.
Even so, a causal relationship in which corporate affiliation caused pricing variation cannot be derived. The research team did not directly measure companies' internal pricing decision-making structures or pricing-setting capabilities, and this is not an intervention study that controlled for factors other than company size. It also does not lead to the assessment that larger companies have more sophisticated pricing strategies. What could be confirmed was a bias in affiliation.
2023 data and the 2026 price table
On March 27, 2026, Valve updated Steam's price conversion data to align with global market conditions. At the same time, it prepared three methods for converting from U.S. dollar prices to 37 currencies and 4 regional groups: a method that converts using exchange rates alone, a method that uses only purchasing power, and a method that combines purchasing power (or the price of equivalent entertainment products) with exchange rates.
Even now, it is publishers who ultimately decide on prices. Valve's official documentation explains that developers can choose which conversion method to use, or enter their own prices. The "coexistence of common standards and corporate discretion" captured by the research also applies to the current system, but the approximately 128% and 130% figures are results using the pre-revision recommended prices as the denominator. They cannot be repurposed as comparison values against the current recommended prices.
The paper, co-authored by Murakami and Yoshida, was published online ahead of print on August 21, 2026, in the Journal of International Marketing by SAGE Publications. However, as this is a single study immediately after publication, whether the four patterns would be reproduced at a different time or on another distribution platform remains unconfirmed.
To advance verification, it would be necessary to repeat the same analysis after the March 2026 revision of the conversion methods and track how companies chose among the three conversion methods. If sales volume and revenue could also be linked, it would be possible to move beyond merely establishing the existence of the four patterns and, for the first time, evaluate how each pricing setting relates to business outcomes.