OPINION
STRINGER, Justice.We determine whether statutory immunity under the discretionary function exception of Minn.Stat. § 466.03, subd. 6 (1994) applies to a determination by the City of St. Paul as to the sequence of traffic control signals at an intersection within the city limits. Following an accident at an intersection controlled by the city, respondent Darlene Zank sued respondent Gary Larson and appellant the City of St. Paul, alleging that Larson was negligent in failing to yield the right of way at the intersection and that the city was negligent in maintaining defective traffic control signals at the intersection. The city asserted various defenses, including immunity from liability, and moved for summary judgment arguing that its determination regarding signal interval timing at the intersection was protected by statutory immunity. The district court denied the city’s motion for summary judgment, the city sought an immediate appeal, and the court of appeals affirmed the district court’s denial of summary judgment in a divided opinion. We reverse, concluding that the city’s determination as to the timing of traffic control signals is protected by statutory immunity, and the city is therefore entitled to summary judgment.
On January 16, 1990, a motor vehicle accident occurred at the intersection of three streets: Dale Street running in a north/south direction, Front Street running in an east/ west (Erection, and Como Avenue running in a northwest/southeast direction. Zank was traveling southbound on Dale Street and intended to make a right turn onto Front Street and proceed westbound. To make her turn onto Front Street, she had to cross Como Avenue. Zank claims that she had a green light as she was coming south on Dale but, as she entered the intersection, the light changed to yellow. As she crossed Como Avenue, her ear was struck on the right side by Larson’s truck, traveling southeast on Como. Larson maintains that he had a green light when he entered the intersection and that he was following a truck that entered the intersection before he did. Larson apparently did not see Zank’s car until immediately before the impact and alleges that he was unable to avoid the accident. As a result of the accident, Zank suffered severe and permanent injuries to her head, neck, and back.
At the time of the accident, vehicles headed south on Dale Street toward the intersection had a green light for 22 to 25 seconds followed by a yellow light change interval of 3.5 seconds, and finally a clearance interval of one second when all lights at the intersec*721tion simultaneously indicate red.1 The purpose of the yellow and red lights is to clear the intersection of vehicles before conflicting traffic movements can proceed into the intersection. Zank alleges that the city is maintaining a nuisance and that the design of the intersection and the timing of the traffic control signals were contributing factors to her accident. David Daubert, a professional engineer, opined in an affidavit that when Zank entered the intersection on the yellow light, the red clearance interval did not give her sufficient time to clear the intersection before the traffic on Como could proceed on a green light into the intersection.
The city, on the other hand, submitted an affidavit by the city’s traffic signal timing engineer asserting that the one-second duration of the all red clearance interval at this intersection was sufficient for traffic to safely clear the intersection. The engineer further stated that the determination of how to time traffic control signals is made when the intersection is initially designed and involves the traffic design engineer, the traffic signal timing engineer, and the traffic maintenance engineer. State and federal guidelines, the volume of traffic, vehicle approach speed, the impact of signal timing on traffic flow, the configuration of the intersection, driving behavior, changing climate conditions, and the uniformity of signal timing throughout the city are taken into account, with the primary concern the “safe and efficient movement of people and goods.”
Ordinarily, interlocutory appeal is not available from the denial of summary judgment, McGowan v. Our Savior’s Lutheran Church, 527 N.W.2d 830, 832 (Minn.1995), but an exception to this rule arises when the trial court denies summary judgment on the basis of statutory immunity. See Minn. R.Civ.App.P. 103.03; McGovern v. City of Minneapolis, 475 N.W.2d 71, 73 (Minn.1991); Anderson v. City of Hopkins, 393 N.W.2d 363, 364 (Minn.1986). On appeal from denial of summary judgment, this court must determine whether any genuine issues of material fact exist and whether the district court erred in applying the law. State by Cooper v. French, 460 N.W.2d 2, 4 (Minn.1990). The determination of whether governmental action is protected by statutory immunity is a question of law. Snyder v. City of Minneapolis, 441 N.W.2d 781, 786 (Minn.1989).
Minnesota Statutes section 466.03, subdivision 6 (1994), provides that a municipality is immune from tort liability for “[a]ny claim based upon the performance or the failure to exercise or perform a discretionary function or duty, whether or not the discretion is abused.” In determining whether particular government activity is protected by statutory immunity, we have interpreted the discretionary function exception narrowly and have focused on its underlying purpose'— to preserve the separation of powers by preventing courts from passing judgment “on policy decisions entrusted to coordinate branches of government.” Holmquist v. State, 425 N.W.2d 230, 231 (Minn.1988); accord Rico v. State, 472 N.W.2d 100, 104 (Minn.1991). The issue is not whether the government action involved the exercise of discretion in a general sense, because almost every government function does involve some exercise of discretion, but rather whether the challenged activity “involved a balancing of policy objectives.” See Nusbaum v. Blue Earth County, 422 N.W.2d 713, 722 (Minn.1988). As we stated in Steinke v. City of Andover:
Discretionary immunity protects the government only when it can produce evidence its conduct was of a policy-making nature involving social, political, or economic considerations, rather than merely professional or scientific judgments.
525 N.W.2d 173, 175 (Minn.1994) (citations omitted).
In reaching our conclusion that the city is immune from liability for its decision *722relating to the timing of the traffic control signals, we look to the roots of the initial decision by the city as to what kind of traffic control device should be employed at the intersection based upon its configuration and the flow of pedestrian and vehicular traffic — clearly a policy determination for which the city has immunity. To control traffic with semaphores as opposed to other methods involved consideration of a variety of policy-based factors and is an activity that respondents admit is immunized. Respondents would draw the line on policy here, arguing that actions taken by the city in determining the traffic control signal sequence in implementation of the policy to install traffic control signals are not immune. The line between the adoption of policy and the implementation of policy is not as bright as respondents would have it, however, as we noted in Pletan v. Gaines:
The distinction between “making” and “implementing” policy is not so simple as plaintiffs would have it. Until people carry out a governmental policy, by doing or not doing something, the policy is a dead letter. Discretionary function immunity would afford little comfort if it did not extend to some of the consequences of the policy itself. Whether .certain consequences are immune depends, as we have noted, on whether the consequential conduct itself involves the balancing of public policy considerations in the formulation of policy.
494 N.W.2d 38, 44 (Minn.1992); see also Smith v. Johns-Manville Corp., 795 F.2d 301, 308 (3d Cir.1986) (warning that decision making should not be “broken down into component parts” and isolated from the context of the overall plan).
Here we deal with the consequences of the policy to implement traffic control with a traffic control signal. One of those consequences is that because of the size and eon-figuration of this particular intersection— three streets coming together as spokes into one intersection — it would require an all red sequence of over five seconds, according to the affidavit of respondents’ expert witness, to assure that all traffic — even the slowest traffic entering the intersection at the last possible moment during the yellow light change interval — would clear it before conflicting traffic movements begin. But an all red sequence of this duration creates its own risk of “traffic hazard”: that of causing confusion among drivers who mistakenly believe that the traffic control signal is malfunctioning and proceed into the intersection against the red signal. For this reason, according to the city’s traffic signal timing engineer, no intersection in the city has an all red signal longer than two seconds in duration:
Longer all red period[s] have been considered and rejected by the Traffic Engineering Division because longer all red intervals give drivers the perception the signal is stuck and this often leads to driver confusion and vehicles entering the intersection prior to being given the green indication.
Thus, in determining the duration of the all red clearance signal here, the city balanced competing safety considerations and selected a one-second interval, perhaps on the not unreasonable assumption that drivers entering this complex intersection would at all times exercise caution and be alert to conflicting traffic movements.2 In any event, we conclude that the city’s decision to provide a shorter all red signal resulted from a balancing of the safety considerations stemming from and in consequence of the city’s initial determination as to the configuration of the intersection and the appropriate traffic control devices. The decision was therefore protected by statutory immunity and the city is entitled to summary judgment.
Reversed.
. The traffic lights at this intersection are activated by either pedestrians or vehicles resulting in many different sequencing combinations. If all phases are active, Dale Street is the first phase, followed by Como Avenue, and then Front Street:
All Red
Green Yellow Clearance
Dale Street 22-25 sec. 3.5 sec. 1.0 sec.
Como Avenue 7-24 sec. 3.5 sec. 1.5 sec.
Front Street 7-25 sec. 3.5 sec. 1.0 sec.
. The Minnesota Manual on Uniform Traffic Control Devices, which provides a uniform policy for all traffic control devices within the state, affords substantial discretion to local governmental units regarding all red clearance intervals. All red clearance intervals are permitted, but not required by the manual. The conclusion of respondents' expert, that it was "reasonable for [Zank] to expect the timing of the lights to be in such a manner and sequence as to protect her while making the turn” when she enters the intersection on a yellow signal, ignores the fact that under the manual, the city may decline to include an all red interval in traffic control signal sequences.