This article aims at measuring recommendation value on the Tunisian market and uses a hand-collected database of 6646 recommendations (2005–2009). We apply the methodology of calendar–time portfolio analysis. This consists of simulating a portfolio that would include stocks depending on the recommendations issued by financial analysts. In order to measure abnormal (or ‘excess’) returns, the raw return of the portfolio is then compared to the evolution of the stock index and to the prediction of the Capital Asset- Pricing Model. Some of the portfolios we build earn a positive significant excess risk-adjusted return of 1.19% per month. Beyond the results that are in line with the literature, we provide two original results. First, ‘sell’ signals are informative, whereas ‘buy’ signals are not. We suggest that it is related to large (small) firms having more ‘buy’ (‘sell’) recommendations and to the direction of the market trend over the period. Second, the fact that recommendation levels have more impact than recommendation changes is explained by the specific informational context on that market, which is that recommendations are systematically disclosed each month, whereas on other markets, recommendations are produced only when the analyst has some new information to disclose.