Sylvester Vaughns and Patricia Allen, representatives of two certified classes, brought this appeal from the district court’s decision that Prince George’s County (hereafter sometimes the county) did not discriminate against blacks in hiring or promotions from 1972 until 1981. The suit was brought under Title VII of the Civil Rights Act of 1964, §§ 701-718, as amended, 42 U.S.C. §§ 2000e et seq. (1976 & Supp. IV 1980) (hereafter the Act), which became applicable to county governments on 24 March 1972. Because we cannot find that the district court erred, we affirm its opinion, published at 538 F.Supp. 833 (D.Md. 1982).
The district court opinion fully sets out the facts. Briefly, Prince George’s County has approximately 4000 employees, hired through a central personnel office. The county charter requires that a merit system be used for all county employment practices. From 1973 on, the county’s employment practices have also been subject to an Equal Employment Opportunity plan. County jobs are divided into approximately 400 categories, called classes. Written specifications for each class include minimum standards for education and prior work experience.
The county’s hiring procedures call for all entry level openings to be available to both external applicants and people already employed by the county. ' Most of these positions are filled by externals. For openings above éntry level, the county personnel law requires that where there are at least six county employees qualified to fill the position, the opening be considered a promotion, and be available only to internal applicants. Between 1974 and 1980, approximately 50% of all positions were open to internals only.
The personnel office receives applications and “rates and ranks” them. The rating and ranking is done by comparing the specifications for each job with the information provided on the application form, and then determining whether the applicant meets or exceeds the requirements. A Register of Eligibles listing all qualified applicants for each vacant position is then prepared. Thereafter, a list of the top five candidates is sent to the head of the department with the vacancy. If more than one applicant is tied for fifth place, then the list includes all tied applicants. The department head, or a panel including the department head, then conducts interviews of at least five of the candidates, and chooses one of them. Each of the 20 department heads has discretion to conduct the interviews in his or her own way. No county-wide interview guidelines exist, although certain departments have their own established interview questions or procedures. Some departments use panels for interviewing, and include an EEO or minority representative.
*1302The appellant Alien represents unsuccessful black applicants for employment with the county who completed all required pre-employment tests and met all minimum qualification requirements and whose claims of racial discrimination were rejected by Prince George’s County before 19 December 1972. The appellant Vaughns represents blacks hired by the county after 24 March 1972, the effective date of the Act, and still employed after 2 July 1974. The appellants raise five issues on appeal. We will address each one separately. We note that the district court’s findings of fact may not be overset unless clearly erroneous, Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 1789, 72 L.Ed.2d 66 (1982).
I.
The appellants argue that the county’s system granting preference to internal applicants for county jobs is not a seniority or merit system within the meaning of § 703(h) of Title VII, 42 U.S.C. § 2000e-2(h). Under § 703(h), so long as there is no intent to discriminate, terms or conditions of employment which have a discriminatory impact may be imposed pursuant to a bona fide seniority or merit system adopted before the effective date of Title VII, American Tobacco Co. v. Patterson, 456 U.S. 63, 65, 102 S.Ct. 1534, 1535, 71 L.Ed.2d 748 (1982). We agree with the district court’s finding that Prince George’s County’s preference system was not intended to discriminate, and was a valid seniority system under § 703(h) of Title VII.
The first question to be decided is whether the preference system for internals constitutes a “bona fide seniority or merit system” within the meaning of the Act. We have found no case involving the applicability of § 703(h) to a system in which current employees are given preference over external applicants for job openings.1 However, we believe that common sense dictates that the county’s system be called a seniority system. It effectively grants “seniority” to all current employees, regardless of race or sex. Support for such conclusion is provided by California Brewers’ Association v. Bryant, 444 U.S. 598, 100 S.Ct. 814, 63 L.Ed.2d 55 (1980), where the Supreme Court reversed a Ninth Circuit decision that a clause in a collective bargaining agreement giving preference to permanent employees over temporary ones was not a seniority system within the meaning of the Act. The court noted that:
Title VII does not define the term “seniority system,” and no comprehensive definition of the phrase emerges from the legislative history of § 703(h). Moreover, our cases have not purported to delineate the contours of its meaning ... the principal feature of any and every “seniority system” is that preferential treatment is dispensed on the basis of some measure of time served in employment. Id., at 605-06, 100 S.Ct. at 819 [footnotes omitted].
The county’s preference system does accord preferential treatment on the basis of “some measure” of time employed — in fact, on the basis of any time in the county’s employ. We thus believe that the county’s system meets the requirements for a bona fide seniority system.
The court in Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971) held that where a seniority system has a discriminatory effect, business necessity for the practice must be shown. Here appellants claim that since the county’s current employees were white in a larger percentage than the percentage of whites in the geographic area from which the employees are drawn, the seniority system operated to perpetuate discrimination that *1303occurred prior to the effective date of the Act.
First, appellants have not satisfied us that in fact the county discriminated in its hiring prior to the effective date of the Act, but even if it did, the Act does not provide a remedy for the continuing effects of such discrimination, Trout v. Lehman, 702 F.2d 1094, 1104 (D.C.Cir.1983). If after the effective date of the Act, an employer “made all of its employment decisions in a wholly nondiscriminatory way [it] would not violate Title VII even if it had formerly maintained an all-white workforce by purposefully excluding Negroes,” Hazelwood School District v. U.S., 433 U.S. 299, 309, 97 S.Ct. 2736, 2742, 53 L.Ed.2d 768 (1977). Since Prince George’s County’s black population grew explosively after the effective date of the Act, the disparity between the number of blacks in the geographic area and the number employed by the county might be legitimate. During the 1970’s, the county’s black population grew approximately 2.3% per year. However, budget cuts during that time may have meant that there were fewer “new” employees hired, so that the population shift would not have been reflected in the proportion of black county employees. Second, the county’s preference system measures in a race-neutral way, and moreover, what it measures — the state of being a county employee or not — is neither a factor subject to manipulation for prejudicial ends, nor does it lend itself to use for the purpose of racial discrimination. As the district court found, the system was used legitimately, to assist county employees in upward mobility.
Since Griggs the Supreme Court has established that discriminatory intent must be found for a pre-Act seniority system such as this one is to be invalidated. Griggs held that:
[T]he touchstone is business necessity. If an employment practice which operates to exclude Negroes cannot be shown to be related to job performance, the practice is prohibited, 401 U.S. at 431, 91 S.Ct. at 853.
But later, in Trans World Airlines v. Har-dison, 432 U.S. 63, 97 S.Ct. 2264, 53 L.Ed.2d 113 (1977), the Supreme Court considered a case in which an employee, Hardison, had requested certain time off to observe his religion. Accomodating his requests would have required TWA to violate its seniority system by forcing an employee senior to Hardison to work during Hardi-son’s less desirable shift. In rejecting Har-dison’s Title VII claim, the Supreme Court noted that seniority systems are accorded “special treatment” under Title VII, (under § 703(h)), and concluded:
Thus, absent a discriminatory purpose, the operation of a seniority system cannot be an unlawful employment practice even if the seniority system has some discriminatory consequences, 432 U.S. at 82, 97 S.Ct. at 2276.
We find that under California Brewers’ Association, Hardison, and Patterson, Prince George’s County’s preference system was not in violation of Title VII.
II.
The appellants allege that the district judge erred in according no weight at all to their multiple regression, applicant flow, and static work force data. The court found that the appellants had failed to make out a prima facie case because their statistics, which were offered to prove their prima facie case, were fatally flawed. We note that the district court’s determination of what weight to accord the parties’ respective statistical evidence was “undoubtedly factual,” Trout v. Lehman, 702 F.2d at 1101, and thus subject to reversal only for clear error.
Appellants have argued extensively that since statistics may be used to prove a prima facie case, somehow it follows that a district court is not free to accord weight to only those statistics he finds reliable. Appellants’ argument is a non-sequitur. Certainly the burden is on the defendants to rebut a plaintiff’s statistics in. a Title VII case, when the plaintiff’s statistics establish a prima facie case. The county has not *1304contested this tenet. However, the proper allocation of the burden of proof to rebut a prima facie case is entirely unrelated to the ability of the finder of fact to accord whatever weight he considers proper to the evidence offered. Because the appellants offered statistical proof does not mean they are automatically entitled to a presumption that they have proved a prima facie case, without any scrutiny of their proffered statistics. “[Pjlaintiffs must demonstrate to the court’s satisfaction that their statistical comparisons are meaningful,” id,., 702 F.2d at 1101, although the statistics need not be perfect. The district court was manifestly, and we think justifiably, unsatisfied with the proffered statistics here, and we find no error in its decision.
In this case, the appellants’ only statistics included information about employees hired prior to the effective date of the act. Apparently appellants believed that these statistics should be acceptable, even if they could have been better, because evidence of continuing discrimination could render the county liable. Recently the District of Columbia Circuit addressed this very issue, in Trout v. Lehman, 702 F.2d at 1104. Commenting upon the decision of the district court to admit evidence including pre-Act hires, the court said:
The District Court concluded that plaintiffs need not factor out time-barred discrimination, because defendants may be held liable for the continuing effects of that discrimination. This theory was once in vogue, but it is flatly inconsistent with the Supreme Court’s pronouncements in [citations omitted].
The static work force, analysis of the appellants here compared the racial composition of the county’s work force and the labor market of the surrounding geographical area. By comparing the percentage of blacks in the work force to that of the surrounding area, in theory it is possible to ascertain whether there has been discrimination in hiring. Static work force analy-ses, however, are subject to so many flaws that they are often not reliable. The labor market statistics were census-based, and census data is gathered once every ten years. As we have seen, Prince George’s County’s black population grew almost 25% in one decade. It is also difficult to establish the proper bounds for the geographic area to be used. The work force statistic is subject to employer, turnover, which distorts the figure as a reflection of hiring practices. The appellants’ labor-market statistics were derived from both the 1970 and 1980 censuses, and their workforce data, from the county’s own sketchy and incomplete records. Thus there were flaws in both the workforce and the geographic area parts of the data. Moreover, the rate at which minorities accept positions offered to them is not always within the employer’s control.
We note that Title VII imposes no requirement that the workforce mirror the general population, International Brotherhood of Teamsters v. U.S., 431 U.S. 324, 339-40 n. 20, 97 S.Ct. 1843, 1856-1857 n. 20, 52 L.Ed.2d 396 (1977). We are also aware that any assumption that a workforce hired in a completely nondiscriminatory way will mirror the general population is “questionable” due largely to the vagaries of supply and demand in the labor market, Smith & Abram, “Quantitative Analysis and Proof of Employment Discrimination,” 1 U. of Ill.L.Rev. 33, 52 (1981).
For these reasons, the danger of relying on static work force analysis in any circumstances has been recognized by both cases and commentators. See, e.g., Smith & Abram, supra, at 59-61 nn. 111-116; see also, opinion below. Where, as here, the work force statistics have been skewed by the inclusion of employees hired before the applicable date of the Act, we cannot say that the district court erred in excluding the statistics.
The appellants also introduced a multiple regression salary analysis comparing earnings of black and white county employees with similar qualifications and experience. Since it included pre-1972 hires, it too was rejected as evidence by the district court. (The analysis showed that in 1980, black employees were paid approximately $1500 *1305less than similarly qualified white employees.) The variables used in this regression analysis included education, length of service with the county, salary in the job immediately preceding county employment, EEO category of the county job, and department within the county. Originally including all county employees, the analysis as offered in rebuttal would have excluded evidence of pre-1972 hires.
Multiple regression analyses are extremely complicated, and may be flawed because any one of the many variables used is improper, or because the sample size is too small, Smith & Abram, supra, at 67-69.2 Thus they are generally less favored than applicant flow analyses which the district court chose to rely on instead. Courts have not always admitted multiple regression analyses, and when admitted, they have sometimes been used as an aid to the understanding of the parties’ position, rather than as probative of what they purport to show. See, e.g., Greenspan v. Automobile Club of Michigan, 495 F.Supp. 1021, 1064-65 (E.D.Mich.1980). The district court was not under any necessity to accept the multiple regression salary analyses — other, better, evidence in the form of applicant flow analyses was available, and as we stated above, the finder of fact may decide what weight, if any, to accord to proffered evidence. Given the flaws in the appellants’ salary analysis, we agree that the court could exclude it properly.
The appellants also introduced an applicant flow analysis, which was excluded by the district court. The appellants’ applicant flow statistics did not distinguish between those positions filled by externals, and those which were never even advertised to the public, but were available only to internals. Moreover, the appellants’ applicant flow statistics failed to distinguish between the eight EEO categories of jobs available with the county in comparing unsuccessful and successful applicants.
Ill.
The appellants also allege as error the failure of the district court to allow them to introduce in rebuttal new data after the county’s evidence was introduced. When the district court rejected the appellants’ statistics including pre-1972 hires, and accepted the county’s statistics excluding pre-1972 hires, the appellants wanted to introduce a “new” set of statistics which excluded also employees hired prior to 1972. The court refused to allow these new statistics to be introduced in rebuttal. Ordinarily, rebuttal evidence may be introduced only to counter new facts presented in the defendant’s case in chief, 6 Wigmore, Evidence § 1873 (Chadbourne Rev.Ed.1976). Such new facts might include “surprise” evidence presented by the defendants. Permissible rebuttal evidence also includes evidence unavailable earlier through no fault of the plaintiff. Here, the district court reasoned that the appellants had had a chance to decide before trial what statistics to use, that they had elected to include pre-1972 hires, and that they should not be allowed to change their litigation strategy on rebuttal. A change in litigation strategy is not normally permitted on rebuttal, and here:
the Court was not presented with an offer of new factual evidence discovered during the trial, but rather with a new interpretation of physical evidence which had always existed during the pendency of these lawsuits, Russo v. Peikes, 71 F.R.D. 110, 113 (E.D.Pa.1976).
We therefore agree with the district court’s ruling on this point. This is especially true since the appellants’ expert, Dr. Haber, testified that he would have liked to exclude the pre-1972 hires from the beginning, but simply could not figure out how to do so. Dr. Haber’s testimony suggests strongly that he was not “surprised,” as he claimed, by the defendant’s objecting to the inclusion of pre-Act hires, since he himself *1306thought that excluding such hires would bé preferable.
The rebuttal evidence which the appellants allege was improperly refused comprised both static work force and multiple regression analysis which excluded evidence of pre-1972 hires. We cannot agree with the appellants that they should have been allowed to introduce new work force/labor market data excluding the pre1972 hires. Rebuttal was too late for the appellants to come up with this data. Had they wanted to introduce such statistics, the time to do so was when they presented their case. They did not, perhaps because they thought that the statistics which included pre-1972 hires were more likely to support their allegations that the county had hired disproportionately few blacks. But since their expert testified that he would have liked to exclude the pre-Act hires, they cannot be allowed to claim that he was surprised by the county’s introduction of data excluding such hires. And, we cannot believe that appellants could not have come up with data excluding the preAct hires before rebuttal, if they had wanted to do so. For Title VII cases, there are two special circumstances where rebuttal evidence is admissible: (1) when the defendant’s case asserts that there are special qualifications for the job at issue, E.E. O.C. v. Radiator Specialty Co., 610 F.2d 178, 185 (4th Cir.1979): and (2) when the rebuttal would challenge the veracity of the defendant’s evidence, Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 256, 101 S.Ct. 1089, 1095, 67 L.Ed.2d 207 (1981). Neither situation was present here. Thus we affirm the district court’s denial of permission to introduce new evidence at rebuttal.
IV.
The county introduced its own applicant flow data excluding pre-1972 hires, and the district court relied entirely on these statistics, to the exclusion of all statistical evidence offered by the appellants. The appellants contend that reliance solely on the county’s applicant flow data constituted clear error. We disagree.
Applicant flow data has been found more convincing than work force/labor market comparisons in several cases, see, e.g., U.S. v. County of Fairfax, 629 F.2d 932 (4th Cir.1980), cert. denied, 449 U.S. 1078, 101 S.Ct. 858, 66 L.Ed.2d 801 (1981); on remand, 25 F.E.P. 662 (E.D.Va.1981), and has been preferred by courts as well as commentators in Title VII cases, Reynolds v. Sheet Metal Workers Local 102, 498 F.Supp. 952, 965 (D.D.C.1980): aff'd., 702 F.2d 221 (D.C.Cir.1981). The county’s applicant flow analysis considered entry level hiring separately from the filling of non-entry level positions, since only external applicants competed for entry level positions as a rule, while the preference system meant that internal applicants were favored in hiring for non-entry level positions. Indeed, the county's expert testified that 50% of all county positions were opened only to internals, a statistic justifying the differentiation between internals and externals in the applicant flow analysis. The county’s applicant flow analysis also divided the applicants and employees according" to EEO category, a breakdown not attempted by the appellants’ expert in the applicant flow data he prepared.
For the same reasons that the district court had the right to decide what weight to accord appellants’ statistical evidence, the court as fact-finder could decide whether the county’s applicant flow analysis was reliable evidence. We recognize that the county's applicant flow analysis had shortcomings. Due largely to the county’s incomplete recordkeeping, it did not include all applicants for jobs during the relevant time period. But absent any more compelling evidence of clear error than appellants’ mere claim that the county’s applicant flow analysis was fatally flawed, we will not second-guess the factfinder’s decision on the relative weights to be accorded to the parties’ respective evidence. Although applicant flow analysis is not of course totally reliable, see, Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977), we find the county’s applicant flow *1307evidence persuasive, because unlike the appellants’ it was carefully structured to avoid overbreadth.
The results of the county’s applicant flow analysis are shown in detail in the district court opinion. No statistically significant differences appear for those jobs filled by internals in any of the eight EEO job categories into which all county jobs were broken down for the purposes of the statistical analysis in this litigation. For those jobs in which the successful applicant was an external, there was a standard deviation in excess of three in one of the eight EEO categories. Standard deviations between two and three were present in three more categories, but this court cautioned against reliance on standard deviations between one and three in E.E.O. C. v. American National Bank, 652 F.2d 1176, 1192 (4th Cir.1981); cert. denied, 459 U.S. 923, 103 S.Ct. 235, 74 L.Ed.2d 186 (1982); E.E.O.C. v. Federal Reserve Bank of Richmond, 698 F.2d 633, 648 (4th Cir.1983). The district court thus found the county not liable under Title VII on the basis of an applicant flow analysis showing discrimination significant beyond the cutoff point established in Federal Reserve Bank of Richmond in one of eight EEO categories. The district court considered the other evidence attested to by the county’s witnesses to have satisfactorily explained the slight evidence of discrimination shown by the raw statistics. We note that this statistically significant showing of discrimination was less than one full standard deviation over the threshold of three set by Federal Reserve Bank of Richmond, and was far less than the double-digit figures found significant in the seminal case for the use of statistics in Title VII cases, Castaneda v. Partida, 430 U.S. 482, 97 S.Ct. 1272, 51 L.Ed.2d 498 (1977). We also note that district courts are under a duty to evaluate statistics in light of “all other relevant facts and circumstances,” Dothard, 433 U.S. at 338, 97 S.Ct. at 2731 (Rehnquist, J., concurring); and in light of “all surrounding facts and circumstances,” Teamsters, 431 U.S. at 340, 97 S.Ct. at 1857. Since we are persuaded by his detailed opinion that the district judge did precisely this, we find no clear error on this point, either.
V.
The individual plaintiffs here claimed disparate treatment. Allen alleged that because of her race she was not hired; Vaughns, that because of his race he was not transferred or promoted when budget cuts eliminated his job. For the reasons stated by the district court’s careful and thorough opinion, we find no merit whatsoever to appellants' individual claims.
AFFIRMED.
. We note, however, that Gantlin v. Westvaco Corp., 526 F.Supp. 1356, 1364 (D.S.C.1981), aff’d., 734 F.2d 980, 986 (4th Cir.1984), was a Title VII case where an employer's seniority system was found to pass muster under § 703(h). The court noted that current employees received preference over applicants "from the street,” under the employer’s system allowing employees to request departmental transfers. While the court did not specifically discuss whether this aspect of the transfer system could have stood alone as a § 703(h) seniority system, it seems to have presumed that it was at least compatible with § 703(h).
. For a comprehensive discussion of multiple regression analyses, and statistics in general in Title VII cases, see Vuyanich v. Republic National Bank, 505 F.Supp. 224, 225 (N.D.Tex.1980); reconsidered and affirmed, 521 F.Supp. 656 (N.D.Tex.1981). .