This is an appeal from an order dismissing Robert Lloyd's petition for a writ of certio-rari to the City of Stoughton's Board of Review.1 We conclude that the board did not act according to law because it refused to consider the purchase price of Lloyd's real estate, and because Stoughton's 1990 reassessment was conducted using the "blanket, sweeping, meat cleaver approach" disapproved in State ex rel. *36Kaskin v. Board of Review of Kenosha, 91 Wis. 2d 272, 280, 282 N.W.2d 620, 623-24 (Ct. App. 1979). We therefore reverse, with instructions.
In 1986, Robert Lloyd purchased a small two-flat in Stoughton for $57,000. A 1990 reassessment of all Stoughton real estate set the value at $66,600. Lloyd objected, but the Stoughton Board of Review declined to change the assessment. After the circuit court dismissed his petition for a writ of certiorari, Lloyd appealed. He argues that the board erred by failing to consider the price he paid for his two-flat. He also claims that the method the assessor used to value his two-flat did not individualize assessments, but instead increased the value of all Stoughton properties by set percentages.
STANDARD OF REVIEW
Our power to review a board of review's real property assessment is limited. Flood v. Village of Lomira, Bd. of Review, 153 Wis. 2d 428, 433, 451 N.W.2d 422, 424 (1990). We consider only whether the evidence is sufficient, whether the board kept within its jurisdiction, whether it acted according to law, or acted arbitrarily or in bad faith. Id. We do so independently of the circuit court's conclusions. Steenberg v. Town of Oakfield, 167 Wis. 2d 566, 571, 482 N.W.2d 326, 327 (1992).
PURCHASE PRICE OF REAL ESTATE
In the presence of an arm's-length sale, it is error for an assessor to use other means to assess the value of real estate. Id. at 573,482 N.W.2d at 328. Lloyd argues that the board erred because it refused to use his con-cededly arm's-length purchase of the two-flat as the *37assessed value of the property. The city responds that it is entitled to ignore a 1986 purchase price in a 1990 reassessment because the assessment is over three years from the sale. It cites State ex rel. Wisconsin Edison Corp. v. Robertson, 99 Wis. 2d 561, 567-68, 299 N.W.2d 626, 629 (Ct. App. 1980), as authority for its position. In Wisconsin Edison, we held that a three-year period between the purchase of a dam and a reassessment of the property permitted the assessor to consider facts other than the purchase price of the dam, particularly in inflationary times. We said: "The assessor was not required to use the sale price of $1.00 as the property's fair market value on January 1,1978, though that amount is one of the factors having a bearing on the value of the property." Id. (footnote omitted).
The assessor was not required to use the price Lloyd paid for the two-flat in 1986 as his 1990 assessment.2 But, because Lloyd's purchase price was one of the factors having a bearing on the value of the property, the assessor made an error of law by refusing to consider that price. Of course, when the assessor considers Lloyd's purchase price on remand, he must also consider other factors, such as inflation, which may have affected that price since 1986. We therefore reverse the trial court's order on this issue, and direct the trial court to grant Lloyd's petition and remand to the board for further consideration in light of this opinion.
*38REASSESSMENT BASED UPON PERCENTAGE INCREASES
In State ex rel. Kaskin v. Board of Review of Kenosha, 91 Wis. 2d 272, 280, 282 N.W.2d 620, 623-24 (Ct. App. 1979), we held that a system which increased assessments by a predetermined percentage was contrary to law. We said:
[W]e have scattered sales of residential properties in a tax district used as the sole basis for a district-wide percentage increase of all assessments of all residential properties in such district. That substitutes a blanket, sweeping, meat cleaver approach for the property-by-property consideration of "important considerations" as mandated by Rosen [v. City of Milwaukee, 72 Wis. 2d 653, 665, 242 N.W.2d 681, 686(1976)].
Kaskin, 91 Wis. 2d at 280, 282 N.W.2d at 623-24.
The method employed by Stoughton to reassess both land and improvements is related to the method used in Kaskin. All sales of land in Stoughton during 1988 and 1989 were identified. There were thirty-nine such sales. The sales were compared with the then-assessed values of the land. The assessments ranged from 23% of the sales price to 197% of the sales price. The total value of the sales prices of the thirty-nine properties was $539,093, and the total assessed value of these properties was $337,300. Dividing the assessed value by the sales price, Stoughton determined that, on average, the thirty-nine properties were assessed at only 63% of their fair market value, i.e., their sales price. The reciprocal of .63 is 1.5873 (1 divided by .63).
The assessor then determined a land value for Lloyd's property of $7,138, which it obtained from "land *39data and comp." Lloyd does not dispute this figure. But the assessor then multiplied $7,138 by 1.5873 to value Lloyd's land at $11,330. Lloyd argues that the use of this multiplier and a similar "modifier" (1.18) for the improvements to his land is, in effect, the same "meat cleaver" approach disapproved in Kaskin. We will examine Lloyd's assertion.
Of the thirty-nine vacant land sales in Stoughton in 1988 and 1989, only one would have been assessed at its fair market value if the thirty-nine properties had been reassessed using the 1.5873 multiplier. That was the property for which the ratio between assessed value and sales price was .63. The other thirty-eight properties would have been inaccurately assessed. The problem, however, is not the assessment of the thirty-nine sold properties, but the assessment of the remaining properties in Stoughton that were not sold in 1988 and 1989. Stoughton had a population of8,786 in 1990. Wisconsin Legislative Reference Bureau, State of Wisconsin Blue Book 1991-1992, at 739 (1991). In addition to the thirty-nine land sales, there were 261 sales of improved properties, for a total of 300 real estate sales in Stoughton in 1988 and 1989.3 This suggests that only a small portion of Stoughton's real estate changed hands in those two years. If that figure is 10%, then 90% of Stoughton's real estate was not sold in 1988 and 1989. Lloyd's real estate is within that 90%.
The properties that were sold were presumably reassessed at their sales prices. There is no way to know if the properties not sold in 1988 and 1989 suffer from the same inaccuracies as thirty-eight of the thirty-nine land sales. But an exhibit provided by *40Stoughton suggests that these inaccuracies persisted. Of 261 sales of improved land, only four were assessed at a value which reflected the average ratio of assessed value to sales price.
The result of using a multiplier derived from the average ratio of assessed value to sales price is that, for unsold properties, inequities will continue. Two examples follow.
Record No. 2383, a vacant land sale, showed a sale price of $7,500. The Stoughton assessor had assessed the land at $1,700. After its sale, No. 2383 would be reassessed at its sale price of $7,500. But an unsold piece of land identical to No. 2383 would have its assessment increased by 1.5873, and would therefore be reassessed at $2,700, despite its value of $7,500.
Record No. 983, a vacant land sale, showed a sale price of $3,500. The property had been assessed at $6,900. An unsold piece of land identical to No. 983 would have its assessment increased by 1.5873, and would be reassessed at $10,950, despite its value of $3,500.
These examples are the most egregious variations from fair market value of the thirty-nine land sales in Stoughton in 1988 and 1989. But to the extent that the assumed 90% of Stoughton's real estate not sold in 1988 and 1989 deviated from a .63 ratio between assessed value and fair market value, the inequities presented by the two examples would exist. Based on the examples and an assumption that the assessed values of the unsold properties deviated from their fair market values in a manner similar to the thirty-nine land sales and 261 sales of improved property, we conclude that, after the 1990 reassessment, most *41properties in Stoughton were not assessed at fair market value.
As in Kaskin, scattered sales of properties were used as the sole basis to increase all assessments. By using multipliers derived from averages, inequities existing in Stoughton's assessments prior to 1990 were perpetuated or worsened by the 1990 reassessment. Thus, much of Stoughton's real estate is valued either higher or lower than its fair market value. This is the result of the "meat cleaver" approach we disapproved in Kaskin.
Section 70.32, Stats., requires that real property be assessed at its fair market value. The Stoughton assessor made an error of law remediable upon certio-rari. We therefore reverse the circuit court's order, and direct that it remand this matter to the board for further proceedings consistent with this opinion.
Lloyd also challenges the use of a computer modifier based on a replacement factor, and the method the assessor used to calculate the land value in a sale of an improved property. Because we cannot predict the methods by which the assessor will calculate Lloyd's assessment on remand, we do not consider these issues.
By the Court. — Order reversed and cause remanded with directions.
Because this case raised issues of statewide importance, we certified it to the supreme court. Our certification was denied.
There is nothing inherent in the three-year time lapse between purchase and reassessment. An arm's-length sale in the year of assessment is conclusive as to the value of the property. The sales price is highly significant for a few years, absent unusual events. But eventually, a sales price becomes of little help in determining a property's value.
Although the schedule of sales of improved land shows 276 such sales, fifteen were reclassified as sales of land only.