Art Indexes for Art Investments
The value changes of heterogeneous goods such as real estate, art, computers, used cars, etc. can not be assessed using ordinary index calculation methods. As each object is unique, the differences in prices of a heterogeneous basket of goods can rather be the result of the different properties of the goods, than their effective changes in value.
Traditional art market indices attempt to correct the problem by using simple average calculations or by merely considering multiple sales of the same objects. Average calculations (with the corresponding year of sale as the sole determining factor) are often too vague as artists usually produce very different works of art. This means that heterogeneity strongly influences the result and only some of the transactions can be included if the estimated index is to be reasonably reliable. The number of multiple sales of the same objects by certain artists is often too low to calculate a reliable performance. The main problem is that more successful artists with higher priced works have a much better chance of getting into the index (sample selection bias); thus the index is no longer representative.
The basic idea of hedonic regression is to measure the influence that the various characteristics of a work (e.g. size, technique, condition of a painting, etc.) have on its price. The determinant sales year remains in the equation. Its coefficient reflects the influence of the sales year on a standardised object (i.e. a work without the aforementioned heterogeneous characteristics). Using the time coefficients, a robust index can be calculated. As a result, the main disadvantage of average calculations - unreliable indices due to the high degree of heterogeneity of works of art - is eliminated. However, even with this methodology, a certain number of transactions, including the most important determinants, must be available to allow for accurate calculations of return on investments.
Hedonic art market price indexes are calculated using variables from regression models. The coefficients of the variables - numerous characteristics describing the artist and the works of art - show how the prices of the works are influenced by those characteristics.
The significance of the hedonic model - the explanatory power R² - depends on the variables included in the equation. Generally, the more precise artists and their works of art are described by the variables, the more accurate is the resulting art market price index. Thus, the goal is to include as much relevant and independent information on an art object as possible.
Neither the significant nor the independent variables for a particular artist are known in advance. As a result, each artist or art market category requires a specific hedonic regression model consisting of an individual set of variables. Subsequently, the best set - the fit of the model - can be found only by trial and error.
To sum up, the best hedonic model is the one which provides the highest explanatory power and the smallest uncorrelated standard errors of the coefficients; the most stable implicit prices attached to each characteristic. The best fit may require including a high or low number of relevant variables.
Below are selected variables which are included in our hedonic models to calculate artist and art market indexes. They contain information on:
Artist’s personal data
- Life Span
- Years active
- Peer performance
Size and dimensions of the art object
Technique and materials used
Signature and date
- Signed by the artist
- Dated by the artist
Various categories including
- Painting themes
- Design themes
- Photography themes
- Sculpture themes
- Art schools
- Art categories
Importance of the art object within the artist’s oeuvre
Condition of the work of art
Provenance of the work of art
Information on the place of sale
The Sales history
- Date of sale
- Season of sale
- Repeat sales
- Pre-sales estimates
Art price categories
High to low price segments
Standard indexes – fixed base indices – use an initial figure from a base period. In future periods, the relevant figure is expressed as a percentage of that used in the base period. For example, if for the first three years the performance being tracked rose from 200 to 250 to 320, the index would start at 100, then rise to 125 and then to 160.
In the chain based method, there is no fixed base period. The year immediately preceding the one for which the index has to be calculated is assumed as the base year. Thus, in the 200/250/320 example, the first two year's index figures would remain as 100 and 125, but the third year would now be based on the second year, changing the index to 128.
Our standard art and artist market performance indexes are calculated using the second method with 1986 and 1998 as the first base years. The chief advantages of this method are that the price relatives of a year can be compared at a glance with the performance level of the immediately preceding year, and that the method also allows for changes to the model used to calculate the performance. The latter is an important point since taste changes over time. Those changes make it necessary to adjust the variables used for calculating art market performance.
The hedonic function analyzes the relationship between certain aspects (attributes or variables of the hedonic model) of a product and its price, by applying regression techniques. As a result, it is possible to measure separately the changes of the values of those aspects and the changes of the prices. The attributes of an artwork are valued; additionally, having introduced time dummies, the pure price effects can be calculated.
The most important underlying assumptions are: first, art works are all different, even if they stem from the same artist; second, the price of an object, the dependent variable, can be determined reliably by quantifiable characteristics; and lastly, there is a time-stable relationship between these characteristics and the resulting values attached to them.
Our standard price indexes are calculated directly from the series of time dummies’ coefficients for a single overall regression equation, the pooled approach of the time dummy variable method. Nevertheless, if the results show that the coefficients are not time-stable multiple equations are calculated (adjacent period approach).
Furthermore, the problem of selection bias can arise, since only executed sales transactions are considered in our database. In public and private auctions however, not all lots offered are subsequently sold successfully. This is why our models measure the price impact of lots actually not sold (bought-in) as well; the coefficient of this variable is significant for some, but not all artists.
All indices have the same set of variables for the calculation of the initial regression equation. The standard set is similar to the ones used in the studies published in our Academic Studies section. It includes such variables as: artists’ fame, volume of the works, places of sale, techniques, supports, liquidity and price metrics.
The final set of variables used for the calculations changes for each index, since many artists have a different set of price determinants.
Functional Form and Calculation
We are using linear regression models and apply OLS (Ordinary Least Squares), WLS (Weighted Least Squares) and GLS (Generalized Least Squares) calculations. The best model fit decides on the final functional form and calculation method chosen.
Our standard sales transaction prices are inflation adjusted prices net of any commissions. The price indexes calculated in Swiss Francs (CHF) are adjusted for inflation and currency fluctuations.
To be selected in our sample, all artists must meet:
Performance screens: all artists must have displayed a positive price development based on average calculations of past auction sales.
Liquidity screens: all artists must meet a minimum threshold of works sold per year (public and private sales).
All index changes are clearly reproducible and communicated. Changes can include: art categories, art schools, price segments, and calculation methodologies in general.
Art Category Indexes
The explanatory power (adjusted R²) of our art market category indices usually ranges between 0.75 and 0.9. This means, that most of our models can explain between around 75 % and 90 % of the price variances; the remaining unexplained variances are basically due to external factors such as changing taste or bidder fights. The volatility of our indexes, measured as standard deviation of the annual net return rates, is normally between 15 and 20 percentage points.
The explanatory power (adjusted R²) of our single artist indices ranges normally between 0.7 and 0.85. This means, that most of our models can explain between around 70 % and 85 % of the price changes. The volatility of our artist indexes, measured as standard deviation of the annual net return rates, is mostly between 20 and 30 percentage points.
All results explained here, depend on the investment horizon and currency chosen. Furthermore, our indexes (categories and artists) are calculated based on past transactions and can not guarantee future art returns.
Our database consists of data from private transactions and auction results, comprising some 1'500’000 art transactions. The private transactions originate from collectors, galleries and insurance companies. By far the largest share stems from auctions, because their results are publicly available and most of the characteristics of an object (determinants) are published. According to the estimates of experts, the assumed overall ratio of private and public art market transactions is approximately 60/40 in favour of private transactions. This would mean that the database used may quite often represent the smaller part of the market. In our opinion, however, this is not an issue with respect to the validity of the indices and the calculated performance numbers. Comparable financial market indices such as the S&P 500, for example, consist of less than 10% of the tradable shares at U.S. stock exchanges.
Our standard set of art category and artist indexes for subscribers is updated partially on a monthly basis. Depending on the liquidity of the artists, these updates consist currently of annual- and semi-annual sales transaction periods. More frequent sales transaction updates (e.g. monthly or quarterly) are available as individual reports.
The range of artists and categories is continuously expanded. The starting point is modern artists and their paintings, watercolours and drawings. Artists are assigned to their respective art schools based on our expertise and experts’ literature. Obviously, we cannot select artists from the point of view of art historical importance or financial attractiveness alone.
Our services include the creation of robust art market indices that have a comparable quality and validity as financial market indices and their subsequent performance calculations. For this reason, statistical considerations (especially liquidity and homogeneity) also play a decisive role in the selection of artists and categories. You can find a list of the most important artists represented in the respective categories and segments (Categoryperformance) by requesting the corresponding indices.
The majority of artists in our database belong to two main fine arts: painting and sculpture. Additionally, we have also included artists doing: photography, conceptual art and design (furniture etc.).
All artists currently available in our database can be found here: Artist Price Indexes A - Z
Our database consists of artists from various disciplines: painting, sculpture, photography, conceptual art and design (furniture etc.).
The art market indices consist mainly of paintings, watercolors, drawings, original prints, mixed media, reliefs, tapestry, sculptures, and installations from the same art category or art school.
All art category indexes currently available in our database can be found here: Art Market Price Indexes A - Z
The art categories and schools from our indexes can be found here: Art Categories & Schools
Artists and their works from the same art category or art school form our art category indexes. If you click on the name of one of the art category indexes (see link below) you will get the names of the artists included in the respective index.
All art category indexes currently available in our database can be found here: Art Market Price Indexes A-Z
The Swiss Painting Performance Index is an inflation-adjusted price index, based exclusively on auction transaction data of modern Swiss artists in Swiss Francs (CHF, also currency-adjusted). The annual data covers the years from 1986 to 2015. As all country indices, it is published annually as of 2011. The valuation of paintings is based on the hedonic method. The index includes a fixed portfolio of artists from different art schools. Only artists that have demonstrated a reasonable level of liquidity (i.e. with a minimum number of transactions) and a date of death after 1900 are included.
The paintingpricing module is basically a search function. If a user enters the search criteria in the Painting Calculator, the engine looks for matching painting transactions in our database. If several matches can be found, an average price is calculated as a first step. Subsequently, the present value of the average price is computed, using the inflation rate and the return rate. The output given is the estimated price range of the painting, the annual rates of return, the annualized rate of return (geometric annualized return), and the risk measured by the standard deviation of the returns of the artist. If a user enters the historic purchase price in US-Dollars, the present value range of the painting is estimated based on this input. Please note, that our database is not corrected for statistical outliers. This means that they may unduly influence the calculated range on some occasions.
The artistperformance module indicates the indices of the artists chosen. Multiple selections of artists are possible. We calculate the annual rates of return, the annualized rate of return, and the risk measured by the standard deviation directly from these indices. The indices are calculated net of inflation. The CHF indices are also adjusted for currency fluctuations. Some benchmark indices are also given (the selection will be expanded as well). The risk and return of artists can be compared directly to the benchmarks. Moreover, the correlation matrix indicates whether they could be used successfully in a diversified portfolio, as a hedge against inflation, or in times of economic downturns. Please note, that some long-term indices are smoothed exponentially in order to detect trends more clearly. As the numbers (rates of return, risk and average returns) are indicated for the calculated original index, they do not match the smoothed index anymore. As a result, comparing original and smoothed indices may be misleading if the respective numbers are not consulted.
The categoryperformance module indicates the index of the categories chosen. The indices can be selected using a Main Category or an Index Name (country index). The segments offered are the ones for low price paintings (LM, up to CHF 15’000), middle price paintings (MM up to CHF 50’000), high price paintings (PP, above CHF 50’000), and combinations of the three (Various). All artists are assigned to the respective segments based on average values of their paintings. The calculation methods, benchmarks, and output are the same as for artistperformance.
The artists, segments, years, etc. that can be selected in the free demo and membership section (drop down lists) are partly written in grey. This means that the corresponding data is not yet available, or only available to customers with membership subscription. The currently available full range of research results is only visible in the membership section.
You can arrange for paintings, artists and categories to be individually estimated. For the necessary information and input screens, see our menu point Membership Section. In the case of performance calculations (Artistperformance and Categoryperformance), your enquiries are checked and you will receive an e-mail within 48 hours advising whether the report can be produced for you, and if so, by what date and at what cost.
In our membership section, all analyses can be printed as a PDF document. No printouts are provided in the demo section.
All subscriptions from AFP are paid in advance, have a fixed expiry date, and are not renewed automatically. You will be informed by e-mail before your subscription runs out; there are no recurring charges. Cancellations before the expiry date are only possible without a refund.
Refunds are only possible within one month from the date of payment. Refunds are only granted if AFP charged a customer’s credit card by mistake. Unused subscriptions (or parts of them) expire; they cannot be accumulated, carried forward, or refunded.
For further inquiries please contact us by e-mail at firstname.lastname@example.org or
call +41 44 201 25 10
You can reach our customer service department during office hours (CET):
Monday - Friday 9:00 - 17:00