delivered the opinion of the Court,
joined by Justice HECHT, Justice OWEN, Justice ABBOTT, and Justice JEFFERSON.The sole issue in this condemnation case is whether the condemnee’s valuation evidence is admissible. The condemnor, the City of Harlingen, contends that the trial court erred in admitting testimony from the landowner’s expert based on an appraisal method described as subdivision development analysis. Under this approach, the witness estimated the land’s gross value as if it were subdivided for residential development, then discounted this value for the estimated costs of development. Based on the expert’s testimony, the trial court, sitting without a jury, awarded the landowner $232,000 for the condemned land. The court of appeals affirmed. 1 S.W.3d 282. Because we conclude that the trial court erred in admitting this evidence, we reverse the judgment of the court of appeals and remand the case to the trial court for further proceedings.
I.
Between 1972 and 1979, Lois Sharbo-neau, individually and as representative of her deceased husband’s estate, purchased five adjoining parcels of land that formed a 9.85 acre tract. In 1996, the City condemned the entire tract to expand an adjacent city park. When the condemnation occurred, the tract was zoned as open land. The special commissioners, appointed to assess damages under Texas Property Code section 21.014, determined the property to be worth $98,500. Mrs. Sharbo-neau, dissatisfied with this award, appealed to the statutory county court.
Before trial, the parties stipulated that the highest and best use for the condemned property was as a residential subdivision. Mrs. Sharboneau called one expert witness, Joseph Patterson, a licensed, professional real estate appraiser with more than twenty years of experience, mostly in the Rio Grande Valley area. In his testimony at trial and in an appraisal report admitted into evidence, Patterson used the subdivision development method to appraise the condemned land. This method values an undeveloped tract by calculating what a developer could expect to realize from sales of individual lots, taking into account the costs of development and discounting future revenues to present value.
Patterson first determined how many lots could be carved out of the Sharboneau land. Based on lot sizes in surrounding neighborhoods, Patterson assumed that Mrs. Sharboneau’s tract could be divided into 44 lots of 7,700 square feet each. He then estimated the gross revenues for sales of these theoretical lots. Patterson reviewed recent sales of three comparable, unimproved residential lots in one nearby subdivision, which had sold for an average of $2.17 per- square foot. Based on this figure, he calculated an initial sales price of $16,709 per lot on the prospective Shar-boneau “subdivision.” Estimating that it would take three years to sell all the lots, and including an estimated 5 percent price increase for lots sold in both the second and third years of sale, Patterson determined that the total gross sales for lots on the Sharboneau land would be $772,727.
After calculating this figure, Patterson next subtracted ongoing sale and development expenses to reach the developer’s expected net sales proceeds. He itemized ongoing expenses as general overhead and sales expenses (which Patterson estimated at four percent of annual gross sales), clos-*181mg costs and attorney’s fees (one percent), and advertising costs (one percent). Patterson subtracted as an ongoing expense the developer’s liability for property taxes, which would decrease each year as individual buyers took ownership of the lots. Patterson then subtracted an “entrepreneurial/coordination expense,” perhaps better described as the developer’s profit. Based on Patterson’s experience with area developers, he estimated the normal profit at twenty-five percent of gross annual sales. Altogether, he calculated that the total of these ongoing expenses for the three-year absorption period would be $266,347, leaving total net sales proceeds of $508,380 for a developer of the property.
Patterson next applied a discount rate to the annual net sales proceeds to reflect the interest rate necessary to attract debt and equity capital for the development. After applying this discount rate, Patterson determined that the present value of the net sales proceeds was $413,770. But before arriving at a final estimate of the value of Mrs. Sharboneau’s land, Patterson also subtracted the developer’s costs for making the forty-four lots suitable for new home construction. Without offering any details about the nature of this work, Patterson estimated total costs of $123,150 for the entire tract. It is not clear whether he considered this cost an initial expense or a discounted amount for several years of work. With these costs subtracted, Patterson appraised the tract’s value as $290,620, or about $29,000 per acre, and thus offered this figure as fair market value. On cross-examination, however, Patterson admitted that he had never heard of undeveloped land in Harlingen being sold for such a high amount. The trial court overruled several timely objections to the admissibility of the subdivision development evidence.1
The City also offered an expert witness, Jesse Watson, a professional appraiser from Harlingen with additional experience in real estate sales. Watson used only the comparable sales approach to determine the land’s value. Watson identified three comparable sales of vacant land in Harlin-gen. The first was an eight-acre tract in an industrial area about 1.5 miles from the Sharboneau site, which had sold for $6,000 per acre eight months before the appraisal. The second site was a five-acre tract three miles away, located next to an older residential area, which had sold for $7,000 per acre thirteen months before the appraisal. This site was suitable for residential development, but was somewhat inferior in location to the Sharboneau land. The third site, a seventeen-acre tract only a half mile away, was located near an existing residential area. At the time of trial, it was being developed as a residential subdivision. This tract had been sold eighteen months before the appraisal for $10,500 per acre. After making adjustments to the three comparable properties to compensate for their varying characteristics, Watson concluded that the Sharboneau land would sell for between $9100 and $10,500 per acre. Without further explanation, he gave $10,000 per acre as the reasonable sales price, thus giving the Sharboneau property a value of $98,500.
The court concluded that the property’s value should be based on its highest and *182best use. After finding that the City’s comparable sales appraisal failed to provide evidence of the property’s value at its highest and best use, while Mrs. Sharbo-neau’s subdivision development appraisal did so, the court awarded Mrs. Sharboneau $232,000 as just compensation for the condemned property.
II.
Both the United States and Texas Constitutions require governments to compensate landowners for takings of their property for public use. U.S. Const, amend. V (requiring “just compensation”); Tex. Const, art. 1, § 17 (“adequate compensation”). When a government condemns real property, the normal measure of damages is the land’s market value. Tex. PROP. Code Ann. § 21.042(b); United States v. 50 Acres of Land, 469 U.S. 24, 29, 105 S.Ct. 451, 83 L.Ed.2d 376 (1984); Brunson v. State, 444 S.W.2d 598, 602 (Tex.1969). Thus, the central damage issue in the typical condemnation case is how to measure the market value of the condemned property.
Market value is “the price the property will bring when offered for sale by one who desires to sell, but is not obliged to sell, and is bought by one who desires to buy, but is under no necessity of buying.” State v. Carpenter, 126 Tex. 604, 89 S.W.2d 979, 979, modifying 126 Tex. 604, 89 S.W.2d 194, 202 (1936); accord Kirby Forest Indus., Inc. v. United States, 467 U.S. 1, 10, 104 S.Ct. 2187, 81 L.Ed.2d 1 (1984). The three traditional approaches to determining market value are the comparable sales method, the cost method, and the income method. Religious of the Sacred Heart v. City of Houston, 836 S.W.2d 606, 615-17 & n. 14 (Tex.1992). Mrs. Sharboneau asks us to join several other states in recognizing the subdivision development method as- a fourth approach.2
Courts have long favored the comparable sales approach when determining the market value of real property. See, e.g., Bauer v. Lavaca-Navidad River Auth., 704 S.W.2d 107, 110 (TexApp.—Corpus Christi 1985, writ refd n.r.e); County of Bexar v. Cooper, 351 S.W.2d 956, 958 (Tex.Civ.App.—San Antonio 1961, no writ); accord United States v. 8.11 Acres of Land, 680 F.2d 388, 395 (5th Cir.1982). If the goal of an appraisal is to ascertain market value, then logically there can be no better guide than the prices that willing buyers and sellers actually negotiate in the relevant market. Under a comparable sales analysis, the appraiser finds data for sales of similar property, then makes upward or downward adjustments to these sales prices based on differences in the subject property.
Comparable sales must be voluntary, and should take place near in time to the condemnation, occur in the vicinity of the condemned property, and involve land with similar characteristics. U.S. v. 83.90 Acres of Land, 709 F.2d 1012, 1014 (5th Cir.1983); Southwestern Bell Tel. Co. v. Ramsey, 542 S.W.2d 466, 476 (Tex.Civ.App.—Tyler 1976, writ refd n.r.e.). Comparable sales need not be in the immediate vicinity of the subject land, so long as they meet the test of similarity. See City of Austin v. Cannizzo, 153 Tex. 324, 267 S.W.2d 808, 815 (1954). But if the comparison is so attenuated that the appraiser and the fact-finder cannot make valid adjustments for these differences, a court should refuse to admit the sale as comparable. See, e.g., Holiday Inns, Inc. v. State, 931 S.W.2d 614, 623-24 (Tex.App.—Amarillo 1996, writ denied) (too remote in *183time); Urban Renewal Agency of Austin v. Georgetown Sav. & Loan Ass’n, 509 S.W.2d 419, 421-22 (Tex.Civ.App. — Austin 1974, writ ref'd n.r.e.) (dissimilar neighborhood).
When comparable sales figures are lacking or the method is otherwise inadequate as a measure of fair market value, courts have accepted testimony based on the cost approach and the income approach. The cost approach, which looks to the cost of replacing the condemned property, is best suited for valuing.improved property that is unique in character and not frequently exchanged on the marketplace. Religious of the Sacred Heart, 836 S.W.2d at 616 (citing American Institute of Real Estate Appraisers, The Appraisal of Real Estate 62, 349 (9th ed.1987)). While the cost method takes the property’s depreciation into account, it still “tends to set the upper limit of true market value.” Polk Cty. v. Tenneco, Inc., 554 S.W.2d 918, 921 (Tex.1977). Finally, the income approach to value is appropriate when property would, in the open market, be priced according to the income that it already generates. Id. By estimating this future income and applying a capitalization rate, the income approach allows the appraiser to arrive at a present value for the income-producing property. Id.
No matter what appraisal method an expert uses, however, the goal of the inquiry is always to find the fair market value of the condemned property. An appraisal method is only valid if it produces an amount that a willing buyer would actually pay to a willing seller. We must therefore determine whether the subdivision development analysis is a competent means of proving the fair market value of Mrs. Sharboneau’s property.
III.
The City argues, and Justice Baker agrees, that Patterson’s subdivision development analysis is invalid because it requires the appraiser to examine sales of dissimilar property. Texas law recognizes that sales of subdivided lots do not meet the test of similarity when compared to an undivided tract of land. State v. Willey, 360 S.W.2d 524, 525 (Tex.1962); City of Austin v. Cannizzo, 153 Tex. 324, 267 S.W.2d 808, 816 (1954). But the trial judge did not receive Patterson’s lot sales evidence as being comparable to Mrs. Sharboneau’s land. Instead, the trial court received the estimated values of individual lots as a single step in the expert’s mental process of arriving at a value for the undivided property. Because Mrs. Sharboneau did not offer evidence of individual lot sales as comparable to her own undivided property, Patterson’s testimony is not precluded by Willey and Cannizzo. The proper inquiry is not whether Patterson made any use of sales that were dissimilar to the condemned property; it is whether Patterson’s appraisal method as a whole was relevant and reliable evidence of market value.
In affirming the trial court’s judgment, the court of appeals reasoned that the subdivision development method was merely a hybrid of the sales comparison and income approaches to property appraisal. 1 S.W.3d at 288. Because the two methods are admissible individually, the court held that this “hybrid-classical” methodology must also be valid. Id. One court has even held that the subdivision development method is identical to the classic income approach. Board of County Comm’rs v. Kiser Living Trust, 250 Kan. 84, 825 P.2d 130, 137 (1992).
However, we believe that the subdivision development method is distinct from both comparable sales analysis and the income *184method of appraisal. Although subdivision development analysis requires the appraiser to examine the market for ready-to-build lots, such properties are not comparable to the larger, unsubdivided property actually being appraised. And the traditional income approach measures the value of property based on its known ability to produce income in its current state. See, e.g., State v. Schaefer, 530 S.W.2d 813, 815 (Tex.1975) (describing income approach testimony based on the condemned property’s actual yearly receipts). In contrast, the “income” part of a subdivision development analysis is based solely on the speculative, piecemeal sale of the property. Moreover, the subdivision development method includes a number of steps — such as estimates of absorption periods and a developer’s profit — that are present in none of the classic appraisal methods. For these reasons, we conclude that Patterson’s subdivision development analysis is a distinct method of appraisal that we must examine on its own merits.
The three classic approaches to real estate appraisal are relatively uncomplicated methods of arriving at the fair market value of condemned property. In contrast, Patterson’s subdivision development appraisal takes more than a dozen analytical steps, most involving assumptions and estimates, any one of which could seriously affect the appraisal’s accuracy. This wide margin for error counsels against using Patterson’s approach to value undeveloped land in ordinary circumstances. Other courts have criticized appraisals similar to Patterson’s as conjectural and speculative. State ex rel. Dep’t of Transp. v. Panell, 853 P.2d 244, 246 (Okla.Ct.App.1993); Fruit Growers Express Co. v. City of Alexandria, 216 Va. 602, 221 S.E.2d 157, 161-62 (1976); see also Travis Cent. Appraisal Dist. v. FM Properties Operating Co., 947 S.W.2d 724, 729 (Tex.App.—Austin 1997, writ denied) (stating that the subdivision development method is best suited for land that is already subdivided and being marketed, because development cost estimates will be more reliable). We share this skepticism.
Patterson’s appraisal is also suspect because it fails to account for basic marketplace realities. In essence, his testimony calculates what a developer can afford to pay for the right to subdivide, improve, and market the land. Patterson even testified that his appraisal represented “what a developer should pay for the right to receive the benefits” of the raw land. This may sometimes be equivalent to market value, but many times it will not be. In the real world, what a developer “should” pay and what the developer “will” pay may be quite different. Nowhere in the many steps of his analysis did Patterson take into account any characteristics of the relevant marketplace that would affect what price a willing buyer would pay to a willing seller. When demand for land is high, the seller can force the buyer to pay more for the property. But if the buyer has many opportunities to purchase roughly equivalent tracts, the price will inevitably fall.
In addition, Patterson’s subdivision development analysis made little or no adjustment for the buyer’s risk that the subdivision might fail.3 Patterson merely *185assumed that it would take three years to sell all the lots in the hypothetical subdivision. This prediction is insufficient to account for unexpected competition, political opposition to the development, economic stagnation, or other risks that the subdivision could turn out to be a bad investment. These risks too would tend to lower the price a developer would be willing to pay for the land.
Patterson’s appraisal oversimplifies the problem of finding market value in one crucial respect: it assumes that a willing buyer will value the land at the highest price that still allows a reasonable return on the investment. But a competitive market does not ordinarily guarantee that willing buyers will pay the highest price they can afford, for they will often have the option of purchasing comparable property for less money elsewhere. As the United States Supreme Court has recognized:
The highest and most profitable use for which the property is adaptable and needed or likely to be needed in the reasonably near future is to be considered, not necessarily as the measure of value, but to the full extent that the prospect of demand for such use affects the market value while the property is privately held.
Olson v. United States, 292 U.S. 246, 255, 54 S.Ct. 704, 78 L.Ed. 1236 (1934). In Texas condemnation law, market value properly reflects all factors that buyers and sellers would consider in arriving at a sales price. Cannizzo, 267 S.W.2d at 815 (instructing the fact-finder to consider all uses to which the land “is reasonably adaptable and for which it either is or in all reasonable probability will become available within the reasonable future”); Boyer & Lucas v. St. Louis, S.F. & Tex. Ry. Co., 97 Tex. 107, 76 S.W. 441, 441 (1903); All Am. Pipeline Co. v. Ammerman, 814 S.W.2d 249, 252 (Tex.App. — Austin 1991, no writ) (“The trial judge could consider all evidence related to such matters as suitability and adaptability, surroundings, ... and all circumstances which would tend to increase or diminish the pre-condemnation market value of the property”).
Just compensation does not require the government to eliminate the risks that the condemnee landowner would otherwise face in an uncertain marketplace. By starting with the value of ready-to-build lots in successfully completed subdivisions, Patterson’s subdivision development analysis bypassed all of the problems that could appear during an actual development, substituting instead the best possible outcome. Unless an appraisal gives a value based on the land’s condition at the time of condemnation — taking into account all relevant factors that affect its valuation, including the market for its possible future use — it is not relevant to the issue of market value.
IV.
Courts across the nation have considered whether to admit subdivision development method evidence. In some jurisdictions, where valuation evidence is freely admitted, courts almost automatically accept subdivision development method appraisals. E.g., City of Wichita v. Eisenring, 269 Kan. 767, 7 P.3d 1248, 1255 (2000); County of Ramsey v. Miller, 316 N.W.2d 917, 921-22 (Minn.1982). In other states, Mrs. Sharboneau’s appraisal would have been per se inadmissible. Contra Costa Water Dist. v. Bar-C Properties, 5 Cal.App.4th 652, 7 Cal.Rptr.2d 91, 93-95 (1992); Department of Transp. v. Benton, 214 Ga.App. 221, 447 S.E.2d 159, 161 *186(1994); Fruit Growers Express Co., 221 S.E.2d at 160-62. Still other courts have permitted subdivision development method evidence when the particular analysis and the facts of the case demonstrate that the method can produce a useful estimate of market value. Travis Cent. Appraisal Dist., 947 S.W.2d at 728-32 (approving subdivision development appraisal for land that was already being subdivided and sold, which made the underlying estimates more reliable); see also United States v. 819.98 Acres of Land, 78 F.3d 1468, 1471 (10th Cir.1996) (“[I]n the absence of comparable sales, other methods of valuation— such as the capitalized income approach— may be appropriate to determine the market value of condemned property.”); Seravalli v. United States, 845 F.2d 1571, 1574-75 (Fed.Cir.1988); Tamburelli Properties Ass’n v. Cresskill Borough, 15 N.J.Tax 629 (N.J.Tax Ct.1996); cf. El Paso Natural Gas Co. v. Federal Energy Regulatory Comm’n, 321 U.S.App.D.C. 12, 96 F.3d 1460, 1462 (D.C.Cir.1996) (discounted cash flow analysis admissible for valuation of an oil- and gas-producing field); Capital Properties, Inc. v. State, 636 A.2d 319, 321-22 (R.I.1994) (trial court may depart from comparable sales analysis when the condemned property is unique or special purpose). We agree with this more nuanced approach.
Evidence must be relevant to be admissible. Tex.R. Evid. 402. But Patterson’s subdivision development analysis determined only what a developer could hypothetically afford to pay to profitably subdivide the property, not what a developer would pay in the competitive, risk-filled marketplace of the real world. Because the appraisal did not account for these forces, it was not relevant to establishing the market value ■ of Mrs. Sharboneau’s property.
It may be that in some cases involving undeveloped land, expert opinions based on the subdivision development method would be reliable, relevant, and admissible. We cannot make that determination on the record before us. What the record does show is that Patterson’s testimony did not demonstrate what a willing buyer would pay to a willing seller in the relevant market. Because it did not do so, the trial court abused its discretion by admitting the evidence. For this reason, we reverse the judgment of the court of appeals and remand the case to the trial court for further proceedings consistent with this opinion.
Justice BAKER concurred in the judgment only, joined by Justice ENOCH, Justice HANKINSON and Justice O’NEILL.. In an amicus brief, the Appraisal Institute, a leading organization in the field of real estate appraisal, describes the steps of a subdivision development analysis somewhat differently from Patterson’s procedure. See Appraisal Institute, The Appraisal of Real Estate 328-31 (11th ed.1996). Nonetheless, its basic approach is broadly similar to Patterson’s: estimate the gross sales of lots from a hypothetical subdivision of the subject land, subtract the costs of marketing and development, and discount cash flow to arrive at the present value of the property to a willing developer-buyer.
. See infra Part IV.
. The Appraisal Institute’s description of subdivision development analysis states that "The discount rate applied, which is derived from and supported by the market, should reflect the risk involved.” Appraisal Institute, supra, at 329. Another publication from the Institute describes many marketplace uncertainties for which the appraisal must account. Douglas D. Lovell & Robert S. Martin, Subdivision Analysis 33-40 (1993). Patterson’s discount rate represented only "fináncial carrying cost for the debt service and return on equity,” with no adjustment for risk. No*185where else in his analysis did Patterson account for marketplace uncertainties.