Appellant, Zachary Fromberg, appeals to us from an order of the Circuit Court for Baltimore City (the Hon. Thomas Ward, presiding) affirming the action of the Insurance Commissioner permitting appellee, Allstate Insurance Company (“Allstate”), to “non-renew” appellant’s automobile liability insurance policy.
We are asked to consider whether:
I. The lower court erred in refusing to consider evidence not presented before the Insurance Commissioner.
II. The lower court erred in affirming the order of the Insurance Commissioner.
*238 Facts and Proceedings
Zachary Fromberg was an insured of Allstate for thirty-three years, from 1956 through 1989. He was considered by Allstate to be an acceptable risk during that time. On May 10, 1989, Allstate notified Fromberg of its intention to refuse to renew his policy unless Fromberg chose to be excluded from that policy. The grounds for the non-renewal were three claims that Allstate had paid under the policy and a speeding ticket that Fromberg had received. Allstate had paid $542 when Fromberg hit a fixed object on 12/5/88; it paid $181 when Fromberg struck another vehicle while backing up on 12/24/86; and it paid $333 when Fromberg struck a second parked vehicle on 7/16/88. Fromberg had received the speeding ticket on 1/15/87.
Allstate’s non-renewal notice explained that [i]n determining whether a particular risk should be continued, Allstate relies on certain recognized independent research studies [that] have shown that people with accident and/or motor vehicle violation involvement are more likely to have future accidents than people without this involvement.
The notice then listed one such study, entitled “Basic California Traffic Conviction and Accident Record Facts, December 1987” (“the California Study”). It cited a table containing accident and violation data developed in the California Study. The notice informed Fromberg that “[a]pplying this table to your policy record indicates a significantly greater chance of future accidents.”
Fromberg protested the notice of non-renewal, pursuant to Article 48A, Section 240AA(d) of the Insurance Code.1 On September 21, 1989, a hearing was held before a Hear*239ing Examiner. Appellant was not represented by an attorney. Allstate produced a sole witness, Steve Kreseski, who testified that Allstate, due to its “effort to provide preferred coverage, seeks to insure those drivers that have a minimum amount of accident-involvement history.” He explained that under Allstate’s underwriting standards, “any driver who has two or more occurrences within a three-year period shall be excluded from a multiple-driver policy or non-renewed.” He added that given his driving record, Fromberg exceeded those standards. Kreseski further noted that Allstate’s ability to surcharge for accidents was limited to the last two at-fault accidents2 and that Allstate could not apply a surcharge for violations. Kreseski then referred to the California Study cited in Allstate’s non-renewal notice. He stated that Allstate relied on the study for the proposition that
[a] driver who has had either traffic convictions or accident involvement is more likely to have an increased frequency of accident involvement than a driver who has had no previous traffic convictions or accident involvement history within that three-year period.
He explained that a driver like Fromberg, “who has had three accidents has a 4.48 times greater chance of subsequent accident, whereas a driver with one traffic violation has a 1.62 times greater chance of accident involvement.”
Kreseski asked to introduce the affidavit of Matthew Stegle, an Allstate representative. This affidavit explained how statistical data drawn from the California Study was used by Allstate to estimate future accident involvement. It further described in-house studies that Allstate had performed to determine the frequency of accident involvement by the average Allstate insured. The affidavit included a three-page addendum that set forth data from the Califor*240nia study addressing the “increased chances of future accident involvement for drivers with prior accident/violation involvement compared with drivers without prior accident/violation involvement.” Stegle explained that this data was used in Allstate’s cancellation/non-renewal notices. The Hearing Examiner asked Fromberg if he wanted time to review the eight-page affidavit. Fromberg declined, and objected to the admission of the affidavit on the grounds of materiality. Fromberg’s objection was noted and overruled. The Hearing Examiner admitted the document, “with the caveat that it will be given the appropriate weight that it merits when I review the evidence at a later point in time.” Kreseski then concluded his testimony. Fromberg had no questions to ask him, and proceeded to give his own testimony.
The Hearing Examiner issued her Order on Hearing on behalf of the Insurance Commissioner on October 26, 1989. She ruled that Allstate had not violated sections 234A3 and 240AA4 of the Insurance Code, and had met its “burden of proof and production as set forth and required by Section 234A of Article 48A and Crumlish v. Insurance Commissioner, et al, 70 Md.App. 182, 520 A.2d 738 (1987).” She based these conclusions of law on the following findings of fact:
*241Licensee proposed to nonrenew Policy No. 18 065 495 for the reasons set forth in its notice dated May 10, 1989; or in lieu of said nonrenewal, Licensee proposed to exclude ZACHARY FROMBERG from coverage under said policy for the reasons set forth in its notice.
Complainant protested Licensee’s proposed action and upon due consideration of the testimony and exhibits, I find, as a fact, that the Licensee utilizes underwriting standards which provide that it will not continue to insure any driver who has a frequency of losses, convictions for motor vehicle violations, or combination of losses and convictions (frequency is defined as at least two losses and/or convictions by any one operator).
The Hearing Examiner finds that ZACHARY FROM-BERG has the following driving record:
12/05/88 At-fault accident in which Zachary struck a light pole and for which Licensee paid $542.00 under collision;
12/24/86 At-fault accident in which Zachary backed into Claimant’s vehicle and for which Licensee paid $81.00 under property damage;
07/16/88 At-fault accident in which Zachary struck Claimant’s parked vehicle in a parking lot and for which Licensee paid $333.00 under property damage; and 01/15/88 Violation for speeding.
The Hearing Examiner finds that the driving record of ZACHARY FROMBERG exceeds the Licensee’s underwriting standards; and that the driving record of ZACHARY FROMBERG presents additional exposure to the Licensee which is not contemplated by its rating plan. The Licensee produced statistical data from an independent research study which demonstrated that ZACHARY FROMBERG belongs to a group or class of drivers who by virtue of 3 accidents has a 4.48 times greater chance of future accident involvement as compared to a driver with zero accidents.
The Licensee produced statistical data from an independent research study which demonstrated that ZACHARY *242FROMBERG belongs to a group or class of drivers who by virtue of 1 conviction has a 1.62 times greater chance of future accident involvement as compared to a driver with zero convictions.
The Hearing Examiner finds that the statistical data presented is valid. The Licensee produced other statistical data which demonstrated there is a strong relationship between past accidents and violations and the probability for future accident involvement.
The Hearing Examiner finds that the Licensee’s rating plan does not provide for the imposition of an adequate rate for ZACHARY FROMBERG.
The Hearing Examiner finds the Licensee’s application of its underwriting standards to be reasonably related to its economic and business purpose. The Licensee’s rating plan does not contemplate coverage for the additional exposure presented by the driving record of ZACHARY FROMBERG. Continuation of coverage for ZACHARY FROMBERG under circumstances where the Licensee will not receive an adequate rate for such additional exposure will adversely affect the Licensee’s losses and expenses.
Appellant then appealed to the Circuit Court for Baltimore City, in accordance with the procedures for appeal outlined in Article 48A, sections 240AA and 40 of the Insurance Code. On May 10, 1990, a hearing was held before Judge Thomas Ward. Fromberg was represented by an attorney at the hearing. On May 16, 1990, Judge Ward affirmed the Insurance Commissioner. Fromberg then noted this appeal.
The Law
I.
Appellant contends that the circuit court erroneously refused to consider new evidence that was not before the Insurance Commissioner. Specifically, appellant is referring to the California Study, which was not offered into *243evidence by Allstate at the first hearing. It was, however, referred to in two pieces of evidence that were admitted, namely the notice of non-renewal and the Stegle affidavit, and was cited in testimony by Allstate’s representative, Kreseski. At the circuit court hearing, appellant proffered a copy of the California Study to show that (1) through the Stegle affidavit, Allstate had misrepresented mathematical computations contained within the study, and (2) Allstate in general had misapplied the California Study to appellant, the insured. Judge Ward refused to receive the Study into evidence. He stated,
I’m sympathetic with you, but you can’t now rechange the record or put new evidence in. You can tell me what the existing record has that’s inaccurate, but you’ve got to use the record itself to show it.
Appellant argued that in reviewing tables reproduced from a study without seeing the actual study, the Hearing Examiner had committed an error that substantially prejudiced appellant’s rights, and had arrived at a decision that “was not supported by competent material and substantial evidence in view of the entire record” and involved an error of law. The circuit court rejected appellant’s argument that this alleged error should compel the court at least to remand the case to the Hearing Examiner.5
Appellant now argues that the court statutorily was required to consider the study. Pursuant to Article 48, section 40(4) of the Insurance Code, which addresses the de novo nature of appeals from the Insurance Commissioner, it is provided that
*244[ujpon receipt of such transcripts and evidence the court shall hear the matter de novo as soon as reasonably possible thereafter. Upon hearing of the appeal, the court shall consider the evidence contained in the transcript, exhibits, and documents therein filed by the Commissioner, together with such additional evidence as may be offered by any party to the appeal.
Id. The Court of Appeals has determined that “any party to the appeal [from the Insurance Commissioner] may offer additional evidence as a matter of right. Taking Sections 40(4) and (5) together, as we must, it is clear that this is the only meaning of “de novo” in Section 40(4).” Nuger v. Insurance Comm’r, 238 Md. 55, 61, 207 A.2d 619 (1965). To reach that conclusion, the Nuger court reasoned:
that under the old law additional evidence could be offered in the judicial proceedings on the appeal only if the court were satisfied that the evidence was material and that there were good reasons for failure to present it in the administrative proceedings, whereas under the new law any party to the appeal may offer additional evidence as a matter of right.
Thus, it is clear that Fromberg had the right to introduce the California Study subject, of course, to general rules of evidentiary admissibility. The trial judge erred in denying that right. See also Government Employees Ins. Co. v. Ins. Comm’r, 40 Md.App. 201, 208, 389 A.2d 422 (1978).
II.
It is well settled that “the basic standard for reviewing an administrative finding by the Insurance Commissioner is whether the finding is supported by ‘substantial evidence.’ This means whether ‘a reasoning mind” reasonably could have reached the factual conclusion the agency reached.’ ” Lumbermen’s Mutual Casualty Co. v. Ins. Comm’r, 302 Md. 248, 266, 487 A.2d 271 (1985). See also Miller v. Ins. Comm’r, 70 Md.App. 355, 365, 521 A.2d 761 (1987); Nationwide Mutual Ins. Co. v. Ins. Comm’r, 67 Md.App. 727, 737, 509 A.2d 719 (1986). This is a narrow standard of review *245“and the decision of the administrative agency carries a presumption of validity.” Doctor’s Hosp. v. Maryland Health Resources, 65 Md.App. 656, 667, 501 A.2d 1324 (1986). Reviewing courts are urged not to substitute their judgment for that of the agency from which the appeal arose. See Bulluck v. Pelham Wood Apartments, 283 Md. 505, 513, 390 A.2d 1119 (1978).
With the foregoing concepts in mind, we address appellant’s contention that the circuit court erred in not reversing the Commissioner’s decision. Appellant urges that his rights were substantially prejudiced because the Commissioner received the Stegle affidavit into evidence without the California Study. If the Hearing Examiner had reviewed this study, appellant insists, she would have been aware that Allstate misrepresented data from the study, and she would not have found in favor of the insurance company.
To justify its decision not to renew Fromberg, Allstate had to prove that the underwriting standards that warranted this decision were “reasonably related to the insurer’s economic and business purposes.” Md.Ann.Code art. 48A, § 234A(a) (1986 & Supp.1990). “Conclusory statements that the standard does relate are not enough to comply.” Crumlish v. Ins. Comm’r, 70 Md.App. 182, 190, 520 A.2d 738 (1987). In Crumlish we formulated a three question test that must be satisfied in order to sustain the burden of proof imposed by Section 234A.
Facts must be produced which answer at least the following questions:
1. What is the statistical basis for the supposition that a person who has two or more chargeable losses within a 24 month period is more likely to have a chargeable accident within the next 12 months than a person who has had no accidents, one chargeable accident, or two or more nonchargeable accidents?
2. How valid is any such statistical evidence?
*2463. If there is statistical validity to the supposition, what direct and substantial adverse effect would it have upon [the insurer’s] losses and expenses in light of its current approved rating plan?
Id. at 190, 520 A.2d 738. See also Mirkin v. Medical Mutual Liability Ins. Soc. of Maryland, 82 Md.App. 540, 551, 572 A.2d 1126 (1990).
Appellant argues that the Crumlish test was not met because the data from the California Study was misused and misrepresented by Allstate and is not a valid statistical basis for Allstate’s underwriting standard.6 The particular misrepresentation that appellant charges to Allstate is its use in the Stegle affidavit of a different correlation coefficient7 than that contained in the study. In the affidavit’s addendum, Stegle first reproduced two charts from the California Study, that respectively showed the likelihood (“times-as-many factor”) that those drivers with convictions during a two-year period would have accidents in the third year,8 and that those drivers who had had accidents during *247a two-year period would have accidents during the third year.9 Stegle then created a chart that showed the “correlation of prior conviction history and subsequent accident frequency based on California Study data.” He specifically stated that he had computed the correlation coefficient by using the actual third-year accident frequency ratio rather than the “times-as-many” factor used in the California Study. This correlation coefficient was 0.994. The correlation coefficient for the first two charts, as computed by the study, was .079 and .045, respectively. Thus, the correlation coefficient computed by Stegle was higher than that arrived at by the California Study.
Appellant argues that the use of Stegle’s coefficient negated the validity of the statistics underlying Allstate’s underwriting standard and that if the Hearing Examiner had known that the correlation coefficient differed from that used in the study, it would not have found that Allstate satisfied its burden of proof. The Hearing Examiner’s findings of fact demonstrated that she was persuaded that Allstate produced sufficient statistical evidence to support its underwriting standard. Her focus was not on the correlation coefficient, to which she did not even refer, but rather on the statistical data that showed that Fromberg belonged to a class of drivers “who by virtue of 3 accidents has a 4.48 times greater chance of future accident involvement as compared to a driver with zero accidents,” and “who by virtue of 1 conviction has a 1.62 times greater chance of future accident involvement as compared to a *248driver with zero convictions.” This data satisfied Allstate’s underwriting standards, which are based on the supposition that “people with accident and/or motor vehicle violation involvement are more likely to have future accidents than people without this involvement.” 10
*247ACCIDENTS
NUMBER OF ACCIDENTS FIRST TWO YEARS ACCIDENTS TIMES-AS-MANY FACTOR (THIRD YEAR)
0 1.00
1 1.73
2 2.86
3 4.48
4+ 3.50
*248The study apparently would have provided the correlation coefficient arrived at through use of a different method than Allstate’s. We really do not know what effect this would have had upon the Hearing Examiner. Nor did the trial court. Thus, we do not know whether the appellant’s rights were substantially prejudiced because the trial court and the Hearing Examiner did not review the actual study.
In Muhl v. Magan, 313 Md. 462, 479, 545 A.2d 1321 (1988), Judge Lawrence Rodowsky observed for the Court of Appeals:
Under § 40 the appeal from an order on hearing or from denial of a hearing is de novo. Nuger established early on that de novo review under § 40 does not permit a court to substitute its judgment for the expertise of the Commissioner. It is also settled that ‘if an administrative function remains to be performed, a reviewing court may not modify the administrative agency’s action even when *249a statute provides that the court may ‘affirm, modify or set aside’ because a court may not usurp administrative functions’____
Although the trial court is not permitted to substitute its judgment for that of the administrative agency, where, as here, additional evidence is to be submitted that might have an impact on the decision of the administrative agency, the trial court is constrained to “consider the evidence contained in the transcript, exhibits, and documents therein filed by the Commissioner, together with such additional evidence as may be offered by any party to the appeal.” Md.Ann.Code art. 48A, § 40(4) (1986). Upon considering the California Study together with the other evidence of record, the trial court must affirm the Commissioner unless it finds that “the substantial rights of the petitioned ] may have been prejudiced,” see Md.Ann.Code art. 48A, § 40(5) (1986), for any of the reasons set forth in Article 48A, section 40(5), supra. Accordingly, Fromberg is entitled to a further hearing, wherein the court will consider the California Study along with the other evidence presented in the record. Should the trial court find that the Commissioner’s “findings, inferences, conclusions, or decisions,” see Md. Ann.Code art. 48A, § 40(5) (1986), that may have prejudiced the petitioner might have been different if the “additional evidence” had been before the Commissioner, then the trial court should remand. It must not substitute its judgment for the expertise of the Commissioner. Muhl v. Magan, 313 Md. at 479, 545 A.2d 1321.
JUDGMENT REVERSED; CASE REMANDED TO THE CIRCUIT COURT FOR BALTIMORE CITY FOR FURTHER PROCEEDINGS CONSISTENT WITH THIS OPINION; COSTS TO BE PAID BY THE APPELLEES.
. This section provides that
[a]n insured shall have the right to protest the proposed action of the insurer by signing 2 copies of the notice and sending them to the Commissioner within 30 days after the date of mailing of the notice. The Commissioner shall, upon receipt of a protest, notify the insurer of the filing of the protest.
Md.Ann.Code art. 48A, § 240AA(d) (Supp.1990).
. The first accident, occurring on 12/24/86, for which Allstate paid $151, could not be surcharged because it did not meet the $300 threshold necessary in order for Allstate to surcharge an at-fault accident.
. Section 234A, which addresses “[u]nfairness or discrimination in underwriting, provides in relevant part that
[n]o insurer ... shall cancel or refuse to underwrite or renew a particular insurance risk or class of risk ... for any arbitrary, capricious, or unfairly discriminatory reason____ No insurer ... may cancel or refuse to underwrite or renew a particular insurance risk or class of risk except by the application of standards which are reasonably related to the insurer’s economic and business purposes. At any hearing to determine whether there has been a violation of this section, the burden of persuasion shall be upon the insurer to demonstrate that the cancellation, or refusal to underwrite or renew is justified under the standards so demonstrated.
Md.Ann.Code art. 48A, § 234A(a) (1986 & Supp.1990).
. Section 240AA dictates the proper procedure to be followed by the insurer in cases of "cancellation, nonrenewal, increase in premium in reduction of coverage under [a] motor vehicle liability insurance policy." Md.Ann.Code art. 48A, § 240AA (1986 & Supp.1990).
. Such action by the court would be mandated by the Insurance Code, which provides that:
[t]he court may affirm the decision of the Commissioner or remand the case for further proceedings; or it may reverse or modify the decision if the substantial rights of the petitioners may have been prejudiced because the administrative findings, inferences, conclusions, or decisions are ...
(v) [u]nsupported by competent, material, and substantial evidence in view of the entire record as submitted____
Md.Ann.Code art. 48A, § 40(5) (1986).
. As noted by appellee, Allstate, appellant has attacked the evidence produced by Allstate for the first two prongs of the Crumlish test, but has not challenged whether Allstate proved the "direct and substantial adverse effect" of continuing to carry Fromberg.
. The correlation coefficient is an index of how closely, and in what direction, two variables are related. The index can vary from 0 to 1 and can be negative or positive. A correlation of +1.0 indicates perfect association____ A correlation of 0 indicates the complete absence of association.
Calif. Dept, of Motor Vehicles, Basic California Traffic Conviction and Accident Record Facts 12 (1987).
. This chart provides:
CONVICTIONS
NUMBER OF CONVICTIONS FIRST TWO YEARS ACCIDENTS TIMES-AS-MANY FACTOR (THIRD YEAR)
0 1.00
1.62
2.08 2
2.51 3
3.36 4+
. This chart provides:
. The basis for this supposition derives from studies such as the California Study that have developed data correlating accidents and violations with future accident. In its non-renewal notice, Allstate has reproduced the following table from the California Study:
No. of Accidents 1980-1982 Subsequent Accident Risk No. of Violations 1980-1982 Subsequent Accident Risk
0 1.00 0 1.00
1 1.73 1 1.62
2 2.86 2 2.08
3 4.48 3 2.51
4+ 3.50 4+ 3.36
* Basic California Traffic Conviction and Accident Record Facts, December 1987, cited by the State of Maryland in case # 11383. (The category of Subsequent Accident Risk represents the multiplier for a driver’s expected accident involvement based on previous accident involvement or violation as compared to a driver who has none.)